Create and answer a case study similar to the Visa Case (Chapters 7, 9, 10, 13) - Cases Available under Assignments tab on Blackboard that discusses network effects, consumer lock-in and cooperative relations (collusion?) between companies. When does cooperation stifle innovation and lead to anticompetitive outcomes?
Rachel Wentink
December 6, 2011
Final Paper
Overview of the x86 Microprocessor Industry
This paper will examine the nature of the x86 microprocessor industry and the competitive relationship between its two primary players, Intel and Advanced Micro Devices (AMD). It will touch on how firms may behave in a duopoly, especially in regards to innovation and anticompetitive behavior, as well as network effects and consumer lock-in.
Computers rely on what are called central processing units or CPU. The CPUs are the “brains” behind today’s desktop computers and laptops, performing the instructions provided by computer programs such as operating systems and applications. The CPUs built for desktop computers and laptops are called microprocessors because they are built onto a single silicon chip, or integrated circuit. Most of the microprocessors made for desktop computers and laptops are from a family named x86 after an early product, the 8086. Intel and Advanced Micro Devices (AMD) are the two primary competitors within this industry. According to a February, 2011 article in Information Week Intel holds an 80.7% share of microprocessor shipments worldwide. AMD’s share is 19% of the worldwide market.
Intel and AMD are publicly traded corporations, Intel on the NASDAQ, and AMD on the New York Stock Exchange. As would be expected from two competitors, they do not share any common board members. AMD has two board members from IBM, but Intel’s does not contain any PC manufacturers.
The Corporate Profile on Yahoo Finance describes Intel primarily as a firm that “. . . engages in the design, manufacture, and sale of integrated circuits for computing and communications industries worldwide. It offers microprocessor products used in notebooks, netbooks, desktops, servers, workstations, storage products, embedded applications, communications products, consumer electronics devices, and handhelds. “
According to Yahoo Finance, AMD is a U.S.-based semiconductor company selling microprocessors for servers and notebook PCs. It addition, it “ . . . provides embedded processor products for vendors in industrial controls, digital signage, point of sale/self-service kiosks, medical imaging, set-top box, and casino gaming machines, as well as enterprise class telecommunications, networking, security, storage systems and thin-clients, or computers. In addition, the company offers chipset products, including integrated graphics processor chipsets and discrete chipsets for desktop and notebook PCs, professional workstations, and servers.”
Early History and Competitors
The history of Intel and AMD is inextricably tied to a firm named Fairchild Semiconductor. Robert Noyce, considered the “father of the silicon chip”, was general manager at Fairchild, and among his direct reports was an engineer named Gordon Moore, who headed up research and development. As Fairchild became more bureaucratic and less innovative, many of the original founders moved on to other companies. In 1967, National Semiconductor moved from Connecticut to Santa Clara and began hiring many of Fairchild’s talented engineers. Believing that Fairchild was irreparably harmed by the loss of human capital, Noyce and Moore founded Intel in 1968. Their intent was to build integrated memory circuits and competing with Fairchild Semiconductor and National Semiconductor.
AMD’s founders also came from Fairchild Semiconductor, roughly nine months after Intel’s founding. By the time they decided to move, investors felt that the field was already full of viable competitors in the semiconductor industry. While Noyce was able to secure funding for Intel in one afternoon, AMD’s funding was much more difficult to secure, and up until the last moment the founders were unsure they would raise enough funds to move ahead. In fact, Noyce invested in AMD, even though he knew they could turn into a competitor of Intel’s in the future. There is no information to imply that Noyce’s connection led to any collusive activity, or that he had any influence whatsoever on AMD’s plans.
In addition to Fairchild Semiconductor and National Semiconductor, other competitors in the industry in those early days were made up of a mix of existing and newly founded companies. They included Motorola, Mostek, founded in 1969 by ex-employees of Texas Instruments, and Zilog, founded in 1974 by two ex-Intel employees, one of whom had invented the microprocessor. More recent competitors include Via, a small Taiwanese firm, and Transmeta, who sold x86 microprocessors for notebooks, but exited the market in 2006.
Initially, Intel began to build memory chips, but in 1970, the Nippon Calculating Machine Corporation contracted with Intel for a set of chips. Federico Faggin, was assigned to the project instead, and designed and built the chips. Intel’s history would have been dramatically different had the Japanese firm stuck to the original terms of the contract. But instead, the calculator market had become much more competitive, and they came back demanding a price cut. Robert Noyce arranged to refund some of the money to the client for the rights to the design of the chips, including the microprocessor. Intel launched its first microprocessor, the 4004, in November, 1971, gaining a first-mover advantage.
Investopedia defines a first-mover as “ . . . being the first to enter a specific market or industry. Being the first allows a company to acquire superior brand recognition and customer loyalty. The company also has more time to perfect its product or service. “
In Managerial Economics and Business Strategy, author Michael Baye ascribes multiple advantages to being a first-mover in an industry: “Being first may permit her to enjoy monopoly profits throughout the life of the patent. Due to learning curve effects, additional first-mover benefits may accrue in the form of lower costs. First-mover advantages due to learning curve effects can persist long after the patent expires, and also can be present even if there is not a patent in the first place.” (pg. 492). Learning curve effects were present in the semiconductor industry, and played a key role in barriers to entry.
Barriers to Entry
Barriers to entry were and are high in the x86 industry. Entering the market requires significant investment in fabrication plants and equipment, and in the human capital necessary to effectively design and manufacture the chips successfully. Human capital and learning effects have been extremely important in the industry because of the difficulty in manufacturing a high percentage of usable chips. With each new design, cramming many thousands more transistors onto a single silicon chip, the manufacturing challenge would re-emerge. Rates of saleable chips could be very low during the initial manufacturing runs for a new design.
Gordon Moore is famous in the computer industry for Moore’s Law. In a 1965 article in Electronics Magazine, he famously predicted that the number of components on a single chip would double each year. In 1975, he updated his prediction to once every two years. He also predicted that the cost to manufacture the chips would drop significantly. His prediction was dubbed Moore’s Law in 1970, and the industry results adhered to Moore’s Law through the 1990s and into the 2000s. The implications meant that Intel could not bring in high margins at the lower end of their product line. They had to constantly innovate in order to offer higher margin products, as other vendors would compete on price for the less powerful chips. The rapid, constant innovation also increased the need for learning.
Studies have been performed on the importance of “learning by doing” in the semiconductor industry, (Irwin and Klenow, 1994), (Hatch and Reichelstein, 1994), where considerable knowledge is built up and shared on how to resolve manufacturing problems with the sensitive chips. Hatch (1996), as cited in Salgado (1997) describes the challenge: “It is not uncommon for yields for new semiconductor processes to start as low as 10%. In such cases, the cost of scrapped output is initially very high but falls to low levels as yield rises to 90% or higher over time”. Also, according to Creating the Digital Future: The Secrets of Consistent Innovation at Intel by Albert Yu, wafer size and yields (the percentage of usable chips) dramatically impacts profitability in the semiconductor industry. (pp.85-86). Intel, recognizing the value human capital could bring in resolving these manufacturing challenges, exploited its first-mover advantage to hire the best and the brightest in the industry and in the academic world. With the benefit of the more famous name, it could court brilliant engineers more successfully than AMD or other semiconductor firms.
In addition to the investment in plant, equipment, and human resources, barriers to entry also include establishing relationships with the manufacturers who would buy the chips to use in their finished products. Another barrier to entry can be caused by network effects, as discussed in the next section of this paper. Consumer lock-in to a particular product or service can serve as a barrier to entry for other potential competitors who wish to establish their own standard.
Network Effects and Consumer Lock-In
The microprocessor industry makes an excellent case study for what economists like to refer to as positive indirect network externalities or network complementarities. According to Katz and Shapiro (1994), consumers benefit when they utilize a network such as a telecommunications network, and others are also using the technology. Translated, it means being the sole user of a telephone wouldn’t provide much value. The largest part of the value is in the number of users on the network. These are typically referred to as network effects, or direct network externalities as the users are on the same physical network. Katz and Shapiro also bring up a similar effect with durable goods, which they describe as the “hardware/software paradigm”. The consumer must not only determine which hardware they prefer, but also must determine how much software may be available for it. Along with computers, they mention credit cards networks as an example, in which the card is the hardware, and the merchant acceptance acts as the software. With computers, the benefit is a network complementarity, or indirect network externality because of the size and availability of complementary products and/or services.
These positive network complementarities make a product more attractive to a consumer. A classic example is the attractiveness of the iPhone. When is it considered as just a phone, it is not nearly as attractive as when it is considered as a device with over 500,000 available applications, including games. That positive complementarity can have a dark side, however, in that it can tie consumers to a particular good or service because of the benefits of the network, in a phenomenon known as “consumer lock-in”. The partnership between Microsoft, with its Windows operating system, and Intel as the de facto standard for personal computer microprocessors, is a type of network that locked consumers in. Called the “Wintel” platform, consumers remained with these types of computers because of the greater number of applications, the ability to share documents and other files with other computer users, and the ease of finding qualified support.
In the early 1970s Intel continued to innovate in microprocessors, but considered the memory chip their main line of business. Microprocessors weren’t bringing in that much money, as there weren’t that many products in which they could be used. Intel turned down opportunities to build the computers.Gordon Moore was not interested, as he couldn’t see the reason why someone would want one. He couldn’t envision the positive network complementarities.
Intel might not have seen the potential, but by 1977, external firms began to offer smaller computers using the Intel chips. MITS released the Altair computer kit, using Intel’s 8080 chip. Radio Shack also had a kit. Steve Wozniak and Steve Jobs were also innovating, with a portable computer that could actually be picked up by a person.
A program called Visicalc was written for the Apple II. It was the first commercial spreadsheet, and was designed for use by accountants and bookkeepers. It helped to answer Gordon Moore’s question of why someone would want a portable computer. It demonstrated a use for these computers in the business world, and demonstrated the potential market for creating software programs for the devices. Others began writing programs, including word processing programs which freed up secretaries from the tyranny of the typewriter. Network complementarities began to take off for the portable computer.
An almost accidental sales call on IBM secured Intel an order to provide the 8088 microprocessor to IBM for its launch of a personal computer. IBM insisted on a second source, and Intel turned again to AMD. A contract was negotiated in 1982 that would last for 10 years, in which AMD would act as a second source for Intel but also for Intel to second source for AMD as well. Ultimately they ended up in a court battle over the contract.
Motorola had supplied its 68000 processor for the Apple computer, but IBM’s PC, using Microsoft’s MS-DOS as the operating system and Intel’s microprocessor became the de facto standard for much of the personal computer industry by allowing “clones” to be made by other manufacturers. Because of the contract with IBM, Intel’s chips became the standard microprocessor for the clone industry. The 8088 was based upon the 8086 chip, hence the x86 name for the line of microprocessors. Intel swept AMD along in its wake as the second source.
Market Model and Margins
With the IBM release of the PC with Intel’s chips, the founders of Zilog realized their hopes to become a major factor in the microprocessor market were over. Both founders cashed out. Zilog continues today, but once the founders left, it was no longer a credible rival. Other rivals also either disappeared, or focused on other lines of business. As Microsoft released a new operating system, called Windows, the “Wintel” platform, as it is called, became the largest percentage of the portable computer market. While Apple succeeded early on, over time it became a niche player in the market, which meant suppliers for its processors were also relegated to niche status. Intel and Microsoft were now the standard platform. The combination of Windows and Intel, dubbed Wintel, held the majority share of the portable computer market from the point that IBM launched its PC. In fact, in July, 2011, the “Wintel” platform share of the market was declared to be at its lowest level in 20 years at just under 82% by IT analyst firm Canalys. The success of the Wintel platform did force a type of consumer lock-in: if businesspeople and home users wanted a wealth of applications to choose from for their personal computers, for many years they had to choose Wintel.
A pattern emerged, in which Intel would release a new chip which was faster and more powerful, Just as Moore’s Law had predicted. Intel would exercise monopoly power with that chip until AMD could bring to market a comparable, or in some cases, slightly more powerful chip. According to Andy Grove: The Life and Times of an American, the Microprocessor Report on November 28, 1990 carried a story titled “AMD Ends 386 Monopoly”. (pg. 256). At that point the market for those microprocessors acted as a duopoly, with falling prices and margins. Intel would already be at work on the next microprocessor model, and the pattern would repeat itself. The author states, “By 1992, as Intel was busily at work trying to move the world to the 486, AMD had achieved a market share in the 386 class microprocessor of greater than 50 percent.” (Pg. 257). Rarely did AMD bring a significantly innovative and better microprocessor to market before Intel, or at least, that’s how it appeared.
In the early years of the market, Intel could count on new PC buyers to increase its revenue. However, as pointed out, market share and prices would fall rapidly for microprocessors as soon as AMD could release a comparable product. According to the U.S. Department of Commerce’s Bureau of Economic Analysis “ . . . there is a high degree of turnover in this segment as new, faster chips are brought to the market. Second, prices fall steeply over the life of each chip; prices typically start at between $600 to $1000 at introduction--substantially higher than the prices of existing chips. By the time the chip exits the market, its price has fallen to under $100. The steepness of these profiles could reflect demand- or supply-driven forces. On the demand side, the profiles are consistent with the view that users are initially willing to pay high prices for new chips but as the introduction of the new (better) chip nears, they are less willing to do so and prices of the incumbent chips fall.” (pg.14).
From Intel’s perspective, rightly or wrongly, one of the primary reasons why the margins declined so rapidly was the existence of AMD.
Intel Potentially Restrains Competition
An oligopoly is defined by Baye as “A market structure in which there are only a few firms, each of which is large relative to the total industry.” (pg.314). A duopoly is defined on Investopedia as “A situation in which two companies own all or nearly all of the market for a given type of product or service.” This definition clearly fits Intel and AMD as the microprocessor industry gained traction. When one member of the duopoly is so much larger than the other, questions arise about competitive behavior.
In “A Dynamic Oligopoly With Collusion and Price Wars”, Fershtman and Pakes (2000) posit an industry with at most a handful of competitors and use game theory to determine the likelihood of collusion. If there are two firms, their conclusion was that “collusion breaks down when one of the two is near an exit state, but again just how near depends on where the larger competitor is.” (pg. 221). They point out two reasons for collusion being hard to sustain in such an industry. In large part, it has to do with the potential for the smaller firms to exit the industry in the foreseeable future. In theory, the potential for exiting soon takes the game from an infinitely repeatable one to one of a shorter, potentially definable period, as described by Baye on page 363. The closer a small firm is to exiting, the more definable the end period, and the greater the incentive for the small firm to cheat. However, there is also an incentive on the side of the larger firm, which is not covered in the game theory chapter in Baye. According to Fershtman and Pakes, there is incentive to predatory behavior on the part of the larger firm. “The large firm realizes that the future will be better if it can ‘force’ the small firm(s) to exit and then monopolize the market until the next entrant arrives. Since the smaller firm is more likely to exit if there is no collusion, the larger firm has an incentive to deviate from the collusion pricing in order to increase the likelihood of the smaller firm exiting.” (pp. 221-222).
Fershtman and Pakes suggest the potential for collusion is more likely when there are two firms who can be described as strong incumbents. While Intel can be characterized as a strong incumbent, AMD cannot. AMD more closely matches the description of a firm likely to exit the industry, especially if Intel could find a way to starve it of revenue. Intel’s predatory behavior didn’t result in lower prices for newer microprocessors in order to drive AMD from the industry. Instead, as Intel attained a significant portion of the market and built its reputation, it would appear that it took steps designed to reduce AMD’s revenue to an untenable level.
The relationship between Intel and AMD began to deteriorate from the point of their 1982 licensing agreement. Jerry Sanders, the CEO of AMD, could not resist describing the great deal AMD had achieved to the press, and characterizing Intel as having gotten the worst end of the bargain. Some say Andrew Grove took offense at his language. Whatever the reason, Intel began a campaign to try to undercut the agreement and force its early termination. The agreement called for the possibility of Intel accepting product designs from AMD and opting to act as a second source for them. Second sourcing was a standard practice in the semiconductor industry. It was followed by firms who would not risk a high profile project by relying on only one supplier of a critical input. Second sourcing required the leading firm to share its design with the second source, who would then manufacture a similar part, often for less money. Through second sourcing, there was a significant knowledge spillover in the industry.
If Intel accepted any of AMD’s product designs and agreed to second source them, it would reduce or remove the need for AMD to pay royalties for second sourcing Intel’s product designs. AMD wanted to second source the 80286 chips. AMD proposed two product designs Intel could second source. According to Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company, an internal memo within Intel has since come to light details Intel’s strategy at that point:
“1/ Assure AMD they are our primary source through regular management contact and formal meetings
2/ Take no more AMD products under the current agreement” (pg.233)
Intel had determined that it had enough of a hold on the market to sell the upcoming 80386 chip without a second source. AMD would no longer be able to undercut Intel’s margins. Intel’s management believed by not second sourcing the new chip, it could keep the price of the 80386 higher for a longer period of time. However, during this period, it needed to convince AMD it would give them the 80386 design for second sourcing while still using AMD as a second source for the 80286 chips.
Another memo outlining Intel’s policy states, “Maintain a second-source, business as usual posture in the market place. . . . We do not want them [AMD] to go to Hitachi or NEC, and should not stimulate them to do so.” (pg. 234).
As the situation dragged out, AMD determined to invoke the arbitration clause of their contract. Originally they thought it would be a short process, estimating roughly 12 weeks. The arbitration, which ultimately took three years, began a series of bitter legal battles which weren’t settled until 2009.
Halting second sourcing was one means of starving AMD of revenue. AMD was forced to then create its own designs, and for the most part, delivered competitive, value-priced microprocessors. Intel was then accused of taking additional steps to prevent AMD’s successful competition: they began paying rebates to PC manufacturers such as Dell and HP. The rebates required the manufacturer’s to buy less of AMD’s products. According to the European Union’s competition commissioner, during the period of October 2002 – December 2007, the rebates helped Intel maintain a share of at least 70 percent of microprocessor sales.
In 1991, AMD filed an antitrust case against Intel. Per an article in Business week, AMD claimed Intel was behaving in an anti-competitive manner, stating in court documents that it "used multibillion-dollar payments to coerce PC makers such as Dell (DELL) and Hewlett-Packard (HPQ) into using Intel products to the exclusion of AMD's chips."
At the same time, Intel had a long-standing patent dispute with AMD: it had an agreement with a subsidiary to allow for patent-sharing. AMD moved to spin off the subsidiary as a separate company, GlobalFoundries. When GlobalFoundries tried to take the patent agreement with it, Intel cried foul, stating that the agreement did not allow AMD to extend it to a third-party organization. Without the patents, AMD and GlobalFoundries would not be able to manufacture chips of any market value.
In April, 2009, Intel paid AMD $1.25 billion dollars to settle the allegations of antitrust violations. The agreement called for quarterly meetings between the companies and methods to mediate disagreements in the future. It also enabled the firms to cross-license each other’s patents for five years.
The settlement with AMD helped take some of the pressure off Intel, who had been ordered by EU regulators in May of 2009 to pay a fine of 1.45 billion for antitrust violations. As part of the determination, according to the New York Times, Neelie Kroes, the competition commissioner for the EU, "ordered Intel to stop offering rebates that were conditioned on buying less of a rival’s product, or not buying them at all..." In July of 2009, Intel appealed the verdict, but on average the court takes approximately two years to reach a verdict, and the determination has not yet appeared.
In the meantime, the FTC was also threatening to investigate Intel, and there was speculation that the agreement with AMD was intended to ward off a lawsuit. However, Intel's anti-competitive woes were not yet over; the FTC moved forward with its case. In addition to the problem of the rebates, the FCC also accused Intel of a more insidious method of undercutting AMD—that of changing its compiler, used by software developers, so that the performance of non-Intel microprocessors would degrade during comparison testing.
Benchmarking firms are widely used by IT publications to compare performance by selective computer components. Comparisons of performance across substitutable components act as a standard means of informing IT organizations and consumers on the potential quality and benefits of those components. By changing the compiler and not informing the development community of its act, the FTC says that Intel essentially stacked the deck in their favor.
In their public document outlining the reasons for pursuing Intel they stated, “Among the harms to consumers caused by Intel’s deceptive conduct was the harm to the credibility and reliability of industry benchmarks. Industry benchmarks are important tools for consumers to make informed purchasing choices. Informed consumer choice is a basic building block of competition.“(pg. 6).
In December, 2010 Intel settled with the FTC.
Did a Duopoly Lead to More Innovation?
In "Does AMD Spur Intel to Innovate more?" (2011), authors Ronald Goettler of the University of Chicago and Brett Gordon of Columbia University examine the question of whether AMD’s existence brought about greater social welfare through increased consumer surplus and innovation, or if consumers and society as a whole would have been better off with Intel operating solely as a monopoly. Their sample data spanned the period of 1993 to 2004.
One of the primary points of the paper is that the industry behaves differently when the product is durable, as are CPUs and PCs in general. They feel duopolies and monopolies in durable goods markets face different challenges than those in industries selling services or perishable products. A monopoly in a durable goods industry does face a competitor: themselves. Translated to the personal computer microprocessor industry, the challenge becomes one of convincing consumers and businesses that they need to upgrade to newer, faster chips, even though the older ones still function. The paper cites the 2005 Computer Industry Almanac to show that 82 percent of PC purchases in 2004 were replacements. For all practical purposes, once the PC market began to slow down, Intel would have faced competition in order to grow sales significantly, even if AMD did not exist.
As a result, their point was that durable product firms must continue to innovate to stimulate demand for replacement products. Had Intel been a monopoly, it would still have had to innovate to stimulate demand for its microprocessors. Although not mentioned in the paper, a potential positive network effect was the development of more complex personal computer software which required faster clock speeds in order to provide an effective user experience. Popular, processor-intensive software products stimulated demand for newer personal computers with more powerful hardware.
They also felt that duopoly innovation is more sensitive to consumer preferences. A monopoly does not have to focus as heavily on consumer desires—when two firms are competing for the market, they listen more to their potential customers than a single firm will. This point doesn’t necessarily mean that innovation is highest in the duopoly, however. The authors found that in their data sample, rapid market growth, low price sensitivity and high quality preferences favor a duopoly. Other conditions could favor a monopoly. In fact, the authors concluded that in the case of the personal computer microprocessor industry, for their sample data, innovation would be higher if Intel were a monopoly. “The finding that innovation by a monopoly exceeds that of a duopoly reflects two features of the model: the monopoly must innovate to induce consumers to upgrade, and the monopoly is able to extract much of the potential surplus from these upgrades because of its substantial pricing power. If competition with itself were reduced by a steady flow of new consumers into the market, the monopoly would reduce innovation below that of the duopoly."
Their conclusion was that consumer surplus is higher in a duopoly. While innovation would be higher if Intel were a monopoly, AMD drove prices down and as a result, margins dropped for the firms, and consumers benefited.
They also felt the percentage of innovation spillovers affect the pace of innovation and whether a duopoly or monopoly would innovate more. The shape is an inverted U: if innovation spillovers are very high, the leader cannot maintain the lead. If spillovers are very low, the smaller firm cannot glean enough information to keep it competitive, and it must exit the industry. The spillover effect is partly why a monopoly would innovate more -- the leader, concerned that the follower is copying their technology, limits their innovation more than they would if they had no competitor. It should be noted that Intel and AMD cross-license their technology, which muddies the knowledge spillover somewhat. During the period studied, both firms had healthy R&D budgets, with Intel at approximately 11% and AMD at approximately 20%.
One potential way to alter the market to Intel and AMD’s benefit, although not to the consumers’, would have been to reduce the durability of microprocessors. The authors found what appears to be a prisoner’s dilemma in the industry: Intel and AMD would like to create less durable products, but for a variety of reasons, they could not agree to do so.
Ultimately the authors concluded that consumer surplus was higher by 4.2% with the existence of AMD providing competition external to Intel, and driving margins down. At the same time, society would have benefited as a whole by greater innovation, and their sample data pointed to a potential 4.2% increase in innovation had Intel been a true monopoly.
The Future of the Industry
As for future competition between Intel and AMD, consumer tastes have changed. It has shifted demand. Consumers are moving in overwhelming numbers from personal computers to more mobile devices, which has affected the x86 microprocessor producers. The new growth industry is in developing chips for smartphones, electronic tablets and other mobile devices, and Intel is still trying to get a foothold in the new market.
According to eWeek in “A Chip for All Reasons”, when addressing Intel’s investors earlier this year, CEO Paul Otellini said: “There’s been so much written about tablets that I don’t know where to start, except to say we’re on track. “We’re tracking 35 designs on multiple operating systems. Some are shipping now with Windows. We’re demonstrating some Android devices now. The tablet race is nowhere near finished. . .You’ll see the first Intel-based phones [using new Medfield chips] in the market the first part of next year”.
Within the data center or PC server market, the microprocessor battle is moving to support multiple cores to handle the large capacity needed for cloud services. The same article quotes an analyst who follows Intel. “ . . . When you look at the cloud services world—meaning Google, Facebook, Amazon, Microsoft with Azure—they’re all running x86. There’s very little of non-x86 in all those operations.” At the time the article was published, in 2010, AMD appeared to be primarily focused on cloud services. eWeek quoted Rick Bergman, who headed up the company’s processor and computing platforms initiatives at the time, .“ … despite excitement among its competitors about serving handheld devices, [AMD] doesn’t have “any immediate clients to serve the mobile form factor. But it will ‘move into that going forward.’ “AMD’s core processor and GPU technology, he said, “play great in handheld devices as well, but we’ve got a lot of opportunity we’re going to focus on elsewhere.”
Conclusion
For the moment, there is still a duopoly market for x86 personal computer processors with relatively high barriers to entry. That market is less relevant to the future growth of Intel and AMD. There is little in the way of exciting new development for desktop microprocessors. Now roughly 40 years since the creation of the first computer microprocessor, it is a mature industry in which most purchases come from replacement.
In the Visa and Mastercard case presented by Baye, there was speculation that cooperation stifled competition between the two organizations, resulting in less innovation for consumers. The same member banks were represented on the board of each organization, and they refused to approve programs that would give either organization a competitive edge. In comparison, although the founders for Intel and AMD came from the same firm, Fairchild Semiconductor; although one of Intel's founders invested in AMD and enabled their startup; and although initially the two firms were forced to share knowledge and designs through the practice of second sourcing, they developed a bitter rivalry. From the material at hand, it would appear as though Intel deliberately pursued a program of anticompetitive actions against AMD that was anything but cooperative. The two firms spent many years in legal battles. The recent settlement may change the nature of their relationship--that remains to be seen.
According to Goettler and Gordon (2011), if Intel had operated as a pure monopoly consumers would have benefited from approximately 4.2% more innovation. On the other hand, their paper also stated that consumers received roughly the same amount of additional consumer surplus as a result of the duopoly--in this case meaning that prices were lower for microprocessors. From a consumer standpoint, given the tremendous amount of innovation already in the industry, lower prices might have been the better outcome. We'll never know what would have happened if Intel had remained the only x86 microprocessor manufacturer.
Other firms are now attempting to enter the x86 market, which doesn't bode well for either Intel or AMD. According to The Wall Street Journal in "AMD to Cut 1,400 Jobs as New Chief Revamps", " . . . chips based on designs from ARM Holdings PLC architecture not only dominate mobile devices but are working to expand into PCs and servers, the traditional market of Intel and AMD." The entry of other manufacturers will most likely end the duopoly in that portion of the microprocessor market.
The greater opportunities for growth now are in server microprocessors for the infrastructure required for cloud-based services, in which both AMD and Intel are competing, and in microprocessors for mobile technology. Neither will be duopolies. It remains to be seen how the market will evolve.
In a change from last year’s position, AMD is being taken to task by its board of directors for not getting into mobile technology more quickly. As it restructures in the coming months and lays off roughly 10% of its workforce, it promises it will take the savings and invest some of it in new development areas, such as tablet technology and cloud computing. However, both AMD and Intel lack the first mover advantage in mobile technology, and will have to play catch-up. While that role is familiar to AMD, in large part it will be new one to Intel.
Bibliography
Aizcorbe, Ana. (2006). Why Are Semiconductor Price Indexes Falling So Fast? Industry Estimates and Implications for Productivity Measurement. Economic Inquiry, Volume 44, Issue 3 (pp.485-496).
Federal Trade Commission. (2010, August 4). FTC Settles Charges of Anticompetitive Conduct Against Intel. [Press Release]. Retrieved from http://www.ftc.gov/opa/2010/08/intel.shtm
Fershtman, C., Pakes, A. (Summer, 2000). A Dynamic Oligopoly with Collusion and Price Wars. The RAND Journal of Economics, Volume 13, No. 2, pp. 207-236.
Katz, Michael. L. Shapiro, Carl. (Spring, 1994). Systems Competition and Network Effects . Journal of Economic Perspectives. Volume 8, Number 2, pp. 93-115.
Yu, Albert. (1998). Creating the Digital Future: The Secrets of Consistent Innovation at Intel. New York, NY. The Free Press.
End Final Paper
Quiz Questions
1. Which of the items below is a possible outcome of a positive network externality or network complementarity?
A) 1,079,002 available applications for the salesforce.com platform
B) Talking to your Aunt Rhoda by phone every Saturday via Apple’s FaceTime program.
C) Multiple Repair Shops for Hondas in every large to mid-size U.S. city.
D) One million views of that embarrassing video of you on YouTube.
E) All of the above
2. According to Fershtman and Pakes, what would prevent collusion in an oligopoly?
A) High barriers to entry
B) The likely exit of the industry by the smaller firm(s)
C) Strong rivalry between firms
D) The sale of durable goods.
3. According to Goettler and Gordon’s paper “Does AMD Spur Intel to Innovate More which firms were Intel’s competition for x86 microprocessors during the period from 1993-2004?
A) AMD
B) IBM
C) Intel
D) ARM Holdings
E) A and C
4. Which benefits below are not potential outcomes of first mover advantage?
A) Monopoly profits due to patents
B) Consumer Lock-In
C) Reduced production cost through greater “know-how”
D) Lower prices
5. Historically, what item(s) below did NOT contribute to the dramatically falling prices in the microprocessor industry in the 1980s and 1990s?
A) Cheaper inputs
B) Learning by doing
C) Continuous improvements in microprocessor power
D) All of the above
Answers
1. E – All of the above. You may not want to speak with Aunt Rhoda, but the fact that you both can do so on the same network (Apple’s FaceTime) makes it a positive direct network externality. And you may hate that video, but the person who posted it is enjoying the benefits of the positive network complementarity.
2. B– The likely exit of the industry by the smaller firm(s). The likely exit acts as an uncertain timeline. The closer the small firm is to exiting, the more certain the end date of the game, which causes both small and large firms to decide it is better to cheat than to collude. See Baye on Games With an Uncertain Final Period, pg. 370.
3. E – A and C. According to Goettler and Gordon’s theory, Intel’s primary competition was Intel because the microprocessors and the computers that used them were durable goods. According to their theory, in durable goods, even a monopoly firm has competition—it becomes its own competitor.
4. D – Lower prices. Baye cites potential monopoly profits, learning curve effects that drive down costs, and first-mover advantage due to consumer lock-in in Chapter 13.
5. A –Cheaper inputs. Historically, cheaper inputs may have lowered prices at some point, but they did not contribute to the dramatic drop in prices. For each new chip design, learning by doing improved manufacturing techniques, which resulted in an increased percentage of usable chips and a lower overall manufacturing cost. As new breakthroughs in upcoming chips were anticipated, consumers would either wait for the latest model before purchasing a PC, or would demand the older PCs with older microprocessors at a lower price.
The Future of the Industry
As for future competition between Intel and AMD, consumer tastes have changed. It has shifted demand, as described by Baye in Chapter Two of our text. Consumers are moving in overwhelming numbers from personal computers to more mobile devices, which has affected the x86 microprocessor producers. The new growth industry is in developing chips for smartphones, electronic tablets and other mobile devices, and Intel is still trying to get a foothold in the new market.
According to eWeek in “A Chip for All Reasons”, when addressing Intel’s investors earlier this year, CEO Paul Otellini said: “There’s been so much written about tablets that I don’t know where to start, except to say we’re on track. We’re tracking 35 designs on multiple operating systems. Some are shipping now with Windows. We’re demonstrating some Android devices now. The tablet race is nowhere near finished.” He later added “You’ll see the first Intel-based phones [using new Medfield chips] in the market the first part of next year”. ...
Within the data center or PC server market, the microprocessor battle is moving to support multiple cores to handle the large capacity needed for cloud services. The same article quotes an analyst who follows Intel. “ . . . When you look at the cloud services world—meaning Google, Facebook, Amazon, Microsoft with Azure—they’re all running x86. There’s very little of non-x86 in all those operations.” At the time the article was published, in 2010, AMD appeared to be primarily focused on cloud services. eWeek quoted Rick Bergman, who headed up the company’s processor and computing platforms initiatives at the time, .“ … despite excitement among its competitors about serving handheld devices, [AMD] doesn’t have “any immediate clients to serve the mobile form factor. But it will ‘move into that going forward.’ “AMD’s core processor and GPU technology, he said, “play great in handheld devices as well, but we’ve got a lot of opportunity we’re going to focus on elsewhere.”
Conclusion
For the moment there is still a duopoly market for x86 personal computer processors with relatively high barriers to entry. That market is less relevant to the future growth of Intel and AMD. There is little in the way of exciting new development for desktop microprocessors.Now roughly 40 years since the creation of the first portable computer microprocessor, it is a mature industry in which most purchases come from replacement.
In the Visa and Mastercard case presented by Baye, there was speculation that cooperation stifled competition between the two organizations, resulting in less innovation for consumers. The same member banks were represented on the board of each organization, and they refused to approve programs that would give either organization a competitive edge. In comparison, although the founders for Intel and AMD came from the same firm, Fairchild Semiconductor; although one of Intel's founders invested in AMD and enabled their startup; and although initially the two firms were forced to share knowledge and designs through the practice of second sourcing, they developed a bitter rivalry. From the material at hand, it would appear as though Intel deliberately pursued a program of anticompetitive actions against AMD that was anything but cooperative. The two firms spent many years in legal battles. The recent settlement may change the nature of their relationship--that remains to be seen.
According to Goettler and Gordon (2011), if Intel had operated as a pure monopoly consumers would have benefited from approximately 4.2% more innovation. On the other hand, their paper also stated that consumers received roughly the same amount of additional consumer surplus as a result of the duopoly--in this case meaning that prices were lower for microprocessors. From a consumer standpoint, given the tremendous amount of innovation already in the industry, lower prices might have been the better outcome. We'll never know what would have happened if Intel had remained the only x86 microprocessor manufacturer.
Other firms are now attempting to enter the x86 market, which doesn't bode well for either Intel or AMD. According to The Wall Street Journal in "AMD to Cut 1,400 Jobs as New Chief Revamps", " . . . chips based on designs from ARM Holdings PLC architecture not only dominate mobile devices but are working to expand into PCs and servers, the traditional market of Intel and AMD." The entry of other manufacturers will most likely end the duopoly in that portion of the microprocessor market.
The greater opportunities for growth now are in server microprocessors for the infrastructure required for cloud-based services, in which both AMD and Intel are competing, and in microprocessors for mobile technology. Neither will be duopolies. It remains to be seen how the market will evolve.
In a change from last year’s position, AMD is being taken to task by its board of directors for not getting into mobile technology more quickly. As it restructures in the coming months and lays off roughly 10% of its workforce, it promises it will take the savings and invest some of it in new development areas, such as tablet technology and cloud computing. However, both AMD and Intel lack the first mover advantage in mobile technology, and will have to play catch-up. While that role is familiar to AMD, in large part it will be new one to Intel.
Quiz Questions
1. Which of the items below is a possible outcome of a positive network externality or network complementarity?
A) 1,079,002 available applications for the salesforce.com platform
B) Talking to your Aunt Rhoda by phone every Saturday via Apple’s FaceTime program.
C) Multiple Repair Shops for Hondas in every large to mid-size U.S. city.
D) One million views of that embarrassing video of you on YouTube.
E) All of the above
2. According to Fershtman and Pakes, what would prevent collusion in an oligopoly?
A) High barriers to entry
B) The likely exit of the industry by the smaller firm(s)
C) Strong rivalry between firms
D) The sale of durable goods.
3. According to Goettler and Gordon’s paper “Does AMD Spur Intel to Innovate More which firms were Intel’s competition for x86 microprocessors during the period from 1993-2004?
A) AMD
B) IBM
C) Intel
D) ARM Holdings
E) A and C
4. Which benefits below are not potential outcomes of first mover advantage?
A) Monopoly profits due to patents
B) Consumer Lock-In
C) Reduced production cost through greater “know-how”
D) Lower prices
5. Historically, what item(s) below did NOT contribute to the dramatically falling prices in the microprocessor industry in the 1980s and 1990s?
A) Cheaper inputs
B) Learning by doing
C) Continuous improvements in microprocessor power
D) All of the above
Answers
1. E – All of the above. You may not want to speak with Aunt Rhoda, but the fact that you both can do so on the same network (Apple’s FaceTime) makes it a positive direct network externality. And you may hate that video, but the person who posted it is enjoying the benefits of the positive network complementarity.
2. B– The likely exit of the industry by the smaller firm(s). The likely exit acts as an uncertain timeline. The closer the small firm is to exiting, the more certain the end date of the game, which causes both small and large firms to decide it is better to cheat than to collude. See Baye on Games With an Uncertain Final Period, pg. 370.
3. E – A and C. According to Goettler and Gordon’s theory, Intel’s primary competition was Intel because the microprocessors and the computers that used them were durable goods. According to their theory, in durable goods, even a monopoly firm has competition—it becomes its own competitor.
4. D – Lower prices. Baye cites potential monopoly profits, learning curve effects that drive down costs, and first-mover advantage due to consumer lock-in in Chapter 13.
5. A –Cheaper inputs. Historically, cheaper inputs may have lowered prices at some point, but they did not contribute to the dramatic drop in prices. For each new chip design, learning by doing improved manufacturing techniques, which resulted in an increased percentage of usable chips and a lower overall manufacturing cost. As new breakthroughs in upcoming chips were anticipated, consumers would either wait for the latest model before purchasing a PC, or would demand the older PCs with older microprocessors at a lower price.
Barriers to Entry
Barriers to entry were and are high in the x86 industry. It required significant investment in fabrication plants and equipment, and in the human capital necessary to effectively design and manufacture the chips successfully. Human capital and learning effects have been extremely important in the industry because of the difficulty in manufacturing a high percentage of usable chips. With each new design, cramming many thousands more transistors onto a single silicon chip, the manufacturing challenge would re-emerge. Rates of saleable chips could be very low during the initial manufacturing runs for a new design.
Gordon Moore is famous in the computer industry for Moore’s Law. In a 1965 article in Electronics Magazine, he famously predicted that the number of components on a single chip would double each year. In 1975, he updated his prediction to once every two years. He also predicted that the cost to manufacture the chips would drop significantly. His prediction was dubbed Moore’s Law in 1970, and the industry results adhered to Moore’s Law through the 1990s and into the 2000s. The implications meant that Intel could not bring in high margins at the lower end of their product line. They had to constantly innovate in order to offer higher margin products, as other vendors would compete on price for the less powerful chips. The rapid, constant innovation also increased the need for learning.
Studies have been performed on the importance of “learning by doing” in the semiconductor industry (Irwin and Klenow, 1994), (Hatch and Reichelstein, 1994), where considerable knowledge is built up and shared on how to resolve manufacturing problems with the sensitive chips. Hatch (1996), as cited in Salgado (1997) describes the challenge: “It is not uncommon for yields for new semiconductor processes to start as low as 10%. In such cases, the cost of scrapped output is initially very high but falls to low levels as yield rises to 90% or higher over time”. Also, according to Creating the Digital Future: The Secrets of Consistent Innovation at Intel by Albert Yu, wafer size and yields (the percentage of useable chips) dramatically impacts profitability in the semiconductor industry. (pp.85-86).
In addition to the investment in plant, equipment, and human resources, barriers to entry also included establishing relationships with the manufacturers who would buy the chips to use in their finished products.
December 4, 2011
Overview of the x86 Microprocessor Industry
Computers rely on what are called central processing units or CPU. The CPUs are the “brains” behind today’s desktop computers and laptops, performing the instructions provided by computer programs such as operating systems and applications. The CPUs built for desktop computers and laptops are called microprocessors because they are built onto a single silicon chip, or integrated circuit. Most of the microprocessors made for desktop computers and laptops are from a family named x86 after an early Intel product, the 8086. Intel and Advanced Micro Devices (AMD) are the two primary competitors within this industry, between them earning 95% of the x86 microprocessors revenue worldwide.
Early History of the Microprocessor Industry
Long before desktop computers were even considered possible, engineers were building large, bulky computers. Vacuum tubes were used in the earliest computers, as well as in other devices. The tubes were fragile, took up a lot of space, and had heat requirements that required significant power consumption. They also needed to be changed out periodically as they wore out. One of the critical steps enabling the move from these room-sized computers to the desktop, laptop, and mobile computers we use today was discovering how to utilize transistors to perform the work the vacuum tubes performed in the early days of the industry. Transistors were much smaller, less fragile, and did not require heat. Also, they did not have to be changed periodically. The transistor still had a limitation however—they had to be connected to the electrical circuits of the computer, which was difficult work requiring the soldering of thousands of tiny wires. To make the transistor practical, they needed to be “pre-connected”.
Two men are credited with the creation of the integrated circuit that solved the pre-connection problem. Jack Kilby at Texas Instruments holds the patent for the integrated circuit. His first model was built with germanium. Robert Noyce developed a circuit with silicon which resolved some of the remaining technical shortcomings of Kilby’s chip, thus removing the need for vacuum tubes.
The history of Intel and AMD is inextricably tied to a firm named Fairchild Semiconductor. Robert Noyce was general manager at Fairchild, and among his direct reports was an engineer named Gordon Moore, who headed up research and development. As Fairchild became more bureaucratic and less innovative, many of the original founders moved on to other companies. In 1967, National Semiconductor moved from Connecticut to Santa Clara and began hiring many of Fairchild’s talented engineers. Believing that Fairchild was irreparably harmed by the loss of human capital, Noyce and Moore founded Intel in 1968. Their intent was to build integrated memory circuits and compete with Fairchild Semiconductor and National Semiconductor.
AMD’s founders also came from Fairchild Semiconductor, roughly nine months after Intel’s founding. By the time they decided to move, investors felt that the field was already full of viable competitors in the semiconductor industry. While Noyce was able to secure funding for Intel in one afternoon, AMD’s funding was much more difficult to secure, and up until the last moment the founders were unsure they would raise enough funds to move ahead. In fact, Noyce invested in AMD, even though he knew they could turn into a competitor of Intel’s in the future. There is no information to imply that Noyce’s connection led to any collusive activity, or that he had any influence whatsoever on AMD’s plans.
In addition to Fairchild Semiconductor and National Semiconductor, other competitors in the industry in those early days were made up of a mix of existing and newly founded companies. They included Motorola, Mostek, founded in 1969 by ex-employees of Texas Instruments, and Zilog, founded in 1974 by two ex-Intel employees, one of whom had invented the microprocessor.
Intel began initially to build memory chips, but in 1970, the Nippon Calculating Machine Corporation contracted with Intel for a set of chips. The chips were for a desktop calculator named the Busicom. The chips needed to carry out basic arithmetical calculations. However, an Intel engineer, Ted Hoff, thought of a better plan. He wrote up a plan for a central processing unit (CPU), a memory chip, a read-only memory chip, and another device to handle the input and output. However, he didn’t build it as he was pulled off onto other projects. A new engineer from Fairchild Semiconductor, Federico Faggin, was assigned to the project instead, and designed and built the chips. Intel’s history would have been dramatically different had the Japanese firm stuck to the original terms of the contract. But instead, the calculator market had become much more competitive, and they came back demanding a price cut. Robert Noyce, after speaking with the engineers on the project, knew Intel had created something they would want to keep, and arranged to refund some of the money to the client for the rights to the design of the chips, including the microprocessor. Intel launched its first microprocessor, the 4004, in November, 1971, gaining the first mover advantage.
Network Effects
The microprocessor industry makes an excellent case study for what economists like to refer to as positive indirect network externalities or network complementarities. According to Katz and Shapiro (1994), consumers benefit when they utilize a network such as a telecommunications network, and others are also using the technology. Translated, it means being the sole user of a telephone wouldn’t provide much value. The largest part of the value is in the number of users on the network. These are typically referred to as network effects, or direct network externalities as the users are on the same physical network. Katz and Shapiro also bring up a similar effect with durable goods, which they describe as the “hardware/software paradigm”. The consumer must not only determine which hardware they prefer, but also must determine how much software may be available for it. They liken the paradigm to computers, and among other examples, credit cards networks, in which the card is the hardware, and the merchant acceptance acts as the software. With computers, the benefit is a network complementarity, or indirect network externality because of the size and availability of complementary products and/or services.
These positive network complementarities make a product more attractive to a consumer. A classic example is the attractiveness of the iPhone. When is it considered as just a phone, it is not nearly as attractive as when it is considered as a device with over 500,000 available applications, including games. That positive complementarity can have a dark side, however, in that it can tie consumers to a particular good or service because of the benefits of the network, in a phenomenon known as “consumer lock-in”. The partnership between Microsoft, with its Windows operating system, and Intel as the de facto standard for personal computer microprocessors, is a type of network that locked consumers in. Called the “Wintel” platform, consumers remained with these types of computers because of the greater number of applications, the ability to share documents and other files with other computer users, and the ease of finding qualified support.
Like many products, microprocessors are not valuable in and of themselves; for them to provide value, they must be an input to a finished product. In the early 1970s Intel continued to innovate in microprocessors, but considered the memory chip their main line of business. Microprocessors weren’t bringing in that much money, as there weren’t that many products in which they could be used. Intel turned down opportunities to build the computers.Gordon Moore was not interested, as he couldn’t see the reason why someone would want one. He couldn’t envision the positive network complementarities.
Intel might not have seen the potential, but by 1977, external firms began to offer smaller computers using the Intel chips. MITS released the Altair computer kit, using Intel’s 8080 chip. Radio Shack also had a kit. Steve Wozniak and Steve Jobs were also innovating, with a portable computer that could actually be picked up by a person. A program called Visicalc was written for the Apple II. It was the first commercial spreadsheet, and was designed for use by accountants and bookkeepers. It helped to answer Gordon Moore’s question of why someone would want a portable computer. It demonstrated a use for these computers in the business world, and demonstrated the potential market for creating software programs for the devices. Others began writing programs, including word processing programs which freed up secretaries from the tyranny of the typewriter. Network complementarities began to take off for the portable computer.
An almost accidental sales call on IBM secured Intel an order to provide the 8088 microprocessor to IBM for its launch of a personal computer. IBM insisted on a second source, and Intel turned again to AMD. A contract was negotiated in 1982 that would last for 10 years, in which AMD would act as a second source for Intel but also for Intel to second source for AMD as well. Ultimately they ended up in a court battle over the contract.
Motorola had supplied its 68000 processor for the Apple computer, but IBM’s PC, using Microsoft’s MS-DOS as the operating system and Intel’s microprocessor became the de facto standard for much of the personal computer industry by allowing “clones” to be made by other manufacturers. Because of the contract with IBM, Intel’s chips became the standard microprocessor for the clone industry. The 8088 was based upon the 8086 chip, hence the x86 name for the line of microprocessors. Intel swept AMD along in its wake as the second source.
Intel Potentially Restrains Competition
In “A Dynamic Oligopoly With Collusion and Price Wars”, Fershtman and Pakes (2000) posit an industry with at most a handful of competitors and use game theory to determine the likelihood of collusion. If there are two firms, their conclusion was that “collusion breaks down when one of the two is near an exit state, but again just how near depends on where the larger competitor is.” (pg. 221). They point out two reasons for collusion being hard to sustain in such an industry. In large part, it has to do with the potential for the smaller firms to exit the industry in the foreseeable future. In theory, the near term potential for exiting takes the game from an infinitely repeatable one to one of a shorter, potentially definable period, as described by Baye in our text on page 363. The closer a small firm is to exiting, the more definable the period, and the greater the incentive for the small firm to cheat. However, there is also an incentive on the side of the larger firm, which is not covered in the game theory chapter in Baye. According to Fershtman and Pakes, there is incentive to predatory behavior on the part of the larger firm. “The large firm realizes that the future will be better if it can ‘force’ the small firm(s) to exit and then monopolize the market until the next entrant arrives. Since the smaller firm is more likely to exit if there is no collusion, the larger firm has an incentive to deviate from the collusion pricing in order to increase the likelihood of the smaller firm exiting.” (pp. 221-222).
According to Fershtman and Pakes, the potential for collusion is more likely when there are two firms who can be described as strong incumbents. While Intel can be characterized as a strong incumbent, AMD cannot. AMD more closely matches the description of a firm likely to exit the industry, especially if Intel could find a way to starve it of revenue. Intel’s predatory behavior didn’t result in lower prices for newer microprocessors in order to drive AMD from the industry. Instead, as Intel attained a significant portion of the market and built its reputation, it appears that it took steps designed to reduce AMD’s revenue to an untenable level.
The relationship between Intel and AMD began to deteriorate from the point of their 1982 licensing agreement, which was mentioned in a previous posting. Jerry Sanders, the CEO of AMD, could not resist describing the great deal AMD had achieved to the press, and characterizing Intel as having gotten the worst end of the bargain. Some say Andy Grove took offense at his language. Whatever the reason, Intel began a campaign to try to undercut the agreement and force its early termination. The agreement called for the possibility of Intel accepting product designs from AMD and opting to act as a second source for them. Second sourcing was a standard practice in the semiconductor industry. It was followed by firms who would not risk a high profile project by relying on only one supplier of a critical input. Second sourcing required the leading firm to share its design with the second source, who would then manufacture a similar part, often for less money. Through second sourcing, there was a significant knowledge spillover in the industry.
If Intel accepted any of AMD’s product designs and agreed to second source them, it would reduce or remove the need for AMD to pay royalties for second sourcing Intel’s product designs. AMD wanted to second source the 80286 chips. AMD proposed two product designs Intel could second source. According to Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company, an internal memo within Intel has since come to light details Intel’s strategy at that point:
"1/ Assure AMD they are our primary source through regular management contact and formal meetings
2/ Take no more AMD products under the current agreement"
Intel had determined that it had enough of a hold on the market to sell the upcoming 80386 chip without a second source. AMD would no longer be able to undercut Intel’s margins. Intel’s management believed by not second sourcing the new chip, it could keep the price of the 80386 higher for a longer period of time. However, during this period, it needed to convince AMD it would give them the 80386 design for second sourcing while still using AMD as a second source for the 80286 chips. Another memo outlining Intel’s policy states, “Maintain a second-source, business as usual posture in the market place. . . . We do not want them [AMD] to go to Hitachi or NEC, and should not stimulate them to do so.”
As the situation dragged out, AMD determined to invoke the arbitration clause of their contract. Originally they thought it would be a short process, estimating roughly 12 weeks. The arbitration, which ultimately took three years, began a series of bitter legal battles which weren’t settled until 2009.
Halting second sourcing meant AMD was forced to then create its own designs, and for the most part, it delivered competitive, value-priced microprocessors. In fact, for a period of time it became a stronger, more credible competitor. Intel was accused of taking additional steps to prevent AMD’s successful competition at about that time: they began paying rebates to PC manufacturers such as Dell and HP. The rebates required the manufacturers to buy less of AMD’s products. According to the European Union’s competition commissioner, during the period of October 2002 – December 2007, the rebates helped Intel maintain a share of at least 70 percent of microprocessor sales. In May, 2009 the European Union fined Intel $1.45 billion dollars and ordered them to stop paying the rebates.
Aizcorbe, Ana. (September 1, 2005). Why Are Semiconductor Price Indexes Falling So Fast? Industry Estimates and Implications for Productivity Measurement. Bureau of Economic Analysis.
Baye, Michael R. Managerial Economics and Business Strategy. New York, NY: McGraw-Hill/Irwin.
Fershtman, C., Pakes, A. (2000). A Dynamic Oligopoly with Collusion and Price Wars. The RAND Journal of Economics, volume 13, No. 2 (Summer, 2000), pp. 207-236.
Jackson, Tim. (1997). Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company. New York, NY: the Penguin Group.
Kanter, James. Europe Fines Intel $1.45 Billion in Antitrust Case. (2009, May 13).
Retrieved from http://www.nytimes.com/2009/05/14/business/global/14compete.html?pagewanted=all,
Katz, Michael. L. Shapiro, Carl. Systems Competition and Network Effects . Journal of Economic Perspectives. Volume 8, Number 2 (Spring, 1994), pp. 93-115.
November 20, 2011
Since my last posting I’ve scanned a variety of scholarly articles on oligopolies, but the most pertinent one I’ve found is titled Does AMD Spur Intel to Innovate More? It is a working paper dated June 20, 2011, by Ronald Goettler of the University of Chicago and Brett Gordon of Columbia University. It has been conditionally accepted at the Journal of Political Economy.
In the Visa and Mastercard case, the question was whether cooperation between the two stifled competition. In the case of AMD and Intel, since Intel retains such a large share of the market, the question becomes whether innovation would be greater or consumer surplus would be higher if Intel were a monopoly. Does the existence of AMD, with such a small share of the market, cause Intel to innovate more or less?
While a large part of the mathematics in the paper was completely over my head, the authors made a series of points which can help us all better understand how duopolies and monopolies work in certain environments. I will summarize the main points here.
The industry behaves differently if the product is durable. This is one of the key points of the paper. Duopolies and monopolies in durable goods markets face different challenges than those in industries selling services or perishable products. A monopoly in a durable goods industry does face a competitor: themselves. Translated to the personal computer microprocessor industry, the challenge becomes one of convincing consumers and businesses that they need to upgrade to newer, faster chips, even though the older ones still function. The paper cites the 2005 Computer Industry Almanac to show that 82 percent of PC purchases in 2004 were replacements. For all practical purposes, once the PC market began to slow down, Intel would have faced competition in order to grow sales significantly, even if AMD did not exist.
As a result, durable product firms must continue to innovate to stimulate demand for replacement products. Had Intel been a monopoly, it would still have had to innovate to stimulate demand for its microprocessors. (Note, although not mentioned in the paper, a potential positive network effect was the development of more complex personal computer software which required faster clock speeds in order to provide an effective user experience. Popular, processor-intensive software products stimulated demand for newer personal computers with more powerful hardware).
Duopoly innovation is more sensitive to consumer preferences. A monopoly does not have to focus as heavily on consumer desires—when two firms are competing for the market, they listen more to their potential customers. This point doesn’t necessarily mean that innovation is highest in the duopoly, however. The authors found that in their data sample, rapid market growth, low price sensitivity and high quality preferences favor a duopoly. Other conditions could favor a monopoly. In fact, the authors concluded that in the case of the personal computer microprocessor industry, for their sample data, innovation would be higher if Intel were a monopoly. “The finding that innovation by a monopoly exceeds that of a duopoly reflects two features of the model: the monopoly must innovate to induce consumers to upgrade, and the monopoly is able to extract much of the potential surplus from these upgrades because of its substantial pricing power. If competition with itself were reduced by a steady flow of new consumers into the market, the monopoly would reduce innovation below that of the duopoly."
Consumer surplus is higher in a duopoly. While innovation would be higher if Intel were a monopoly, AMD drove prices down and as a result, margins dropped for the firms, and consumers benefited.
The percentage of innovation spillovers affect the pace of innovation and whether a duopoly or monopoly would innovate more. The shape is an inverted U: if innovation spillovers are very high, the leader cannot maintain the lead. If spillovers are very low, the smaller firm cannot glean enough information to keep it competitive, and it must exit the industry. The spillover effect is partly why a monopoly would innovate more -- the leader, concerned that the follower is copying their technology, limits their innovation more than they would if they had no competitor. It should be noted that Intel and AMD cross-license their technology, which muddies the knowledge spillover somewhat. During the period studied, both firms had healthy R&D budgets, with Intel at approximately 11% and AMD at approximately 20%.
The authors found what appears to be a prisoner’s dilemma in the industry: Intel and AMD would like to create less durable products, but for a variety of reasons, they cannot agree to do so.
Accelerated or increased depreciation could have an unexpected effect on the equilibrium behavior of the industry. While not a key point of the paper, I thought it was interesting enough to highlight. The faster personal computers can be depreciated, the shorter the life cycle of the equipment. Therefore, less utility will be received through each purchase, and the more price sensitive purchasers will become. While depreciation is not a factor with the majority of home PC users, it would affect business purchases. The firms would no longer need to innovate as much in order to stimulate replacement purchases. Faster depreciation would stimulate an increase in demand without substantial product improvements. Therefore, innovation would decline.
Based upon the sample data spanning from 1993 to 2004, the authors concluded that consumer surplus was 4.2% higher with AMD in the market, but innovation would have been 4.2% higher had AMD not been a competitor.
Up until now I have been exploring questions surrounding the duopoly of Mastercard and Visa. From this point on, I'll compare and contrast their situation with that of another duopoly, Intel and AMD.
Here are descriptions of both Intel and AMD from the New York Times:
The Intel Corporation (INTC) is the leading maker of semiconductor chips and is known for its quest to make computer chips ever smaller, faster and cheaper. It also makes circuit boards and other semiconductor products, which are the building blocks of computers, servers, consumer electronics and communications devices. Intel supplies about 80 percent of the PC microprocessor chips worldwide.
In an effort to expand beyond its core chip-making business, Intel said in August 2010 that it had agreed to buy McAfee, the computer antivirus software maker, for about $7.7 billion in cash. With its McAfee purchase, Intel is gaining an entrance into the security tech sector, one that is expected to continue growing quickly.
Intel is also pushing hard in the highly competitive race to put chips in smartphones and televisions and other consumer devices that are gaining computer power. In October 2010 it reported a flurry of sales of the company’s chips in TV and set-top boxes, figures that seemed to bode well for Intel’s consumer aspirations. The company sold one million chips aimed at smart TVs last quarter, and its only prominent loss was getting shut out of Apple TV.
Advanced Micro Devices, Inc. (AMD) is a global semiconductor company. The Company primarily offers x86 microprocessors, for the commercial and consumer markets, embedded microprocessors for commercial, commercial client and consumer markets and chipsets for desktop and notebook personal computers (PCs), professional workstations and servers. It also offers graphics, video and multimedia products for desktop and notebook computers, including home media PCs, professional workstations and servers and technology for game consoles. As of December 25, 2010, the Company had two segments: Computing Solutions and Graphics. Computing Solutions segment includes microprocessors, chipsets and embedded processors and related revenue. The Graphics segment includes graphics, video and multimedia products and related revenue, as well as revenue received in connection with the development and sale of game console systems that incorporate its graphics technology.
Unlike Visa and Mastercard, who ended up in court because they appeared to be collusive, intending to prevent other credit card companies from competing with them, there is no doubt that in this duopoly, Intel and AMD are bitter rivals. Intel has faced a number of legal battles over anti-competitive behavior against AMD. In April, 2009, Intel paid AMD 1.25 billion dollars to settle allegations of antitrust violations. Initially, per an article in Business Week, AMD claimed Intel was behaving in an anti-competitive manner, stating in court documents that it "used multibillion-dollar payments to coerce PC makers such as Dell (DELL) and Hewlett-Packard (HPQ) into using Intel products to the exclusion of AMD's chips."
At the same time, Intel had a long-standing patent dispute with AMD: it had an agreement with a subsidiary to allow for patent-sharing. AMD moved to spin off the subsidiary as a separate company, GlobalFoundries. When GlobalFoundries tried to take the patent agreement with it, Intel cried foul, stating that the agreement did not allow AMD to extend it to a third-party organization. Without the patents, AMD and GlobalFoundries would not be able to manufacture chips of any market value.
The agreement called for quarterly meetings between the companies and methods to mediate disagreements in the future. It also enabled the firms to cross-license each other’s patents for five years. The bigger question is whether the agreement will change the relationship between the two companies and the nature of the duopoly. AMD is known for lower priced chips, and up until the past few years, provided good "value for money". However, they have often lagged behind Intel in delivering transformative technology.
The settlement with AMD helped take some of the pressure off Intel, who had been ordered by EU regulators in May of 2009 to pay a fine of 1.45 billion for antitrust violations. As part of the determination, according to the New York Times, Neelie Kroes, the competition commissioner for the EU, "ordered Intel to stop offering rebates that were conditioned on buying less of a rival’s product, or not buying them at all..." In July of 2009, Intel appealed the verdict, but the court takes approximately two years to reach a verdict.
In the meantime, the FTC was also threatening to investigate Intel, and there was speculation that the agreement with AMD was intended to ward off a lawsuit. However, Intel's anti-competitive woes were not yet over; the FTC moved forward with its case and in December, 2010 Intel settled with them.
This week I'll examine whether Mastercard and Visa are poised to lose their overwhelming control of the payment industry due to current technology changes in the market.
On April 13, 2011, Senator Dick Durbin of Illinois wrote an open letter to Jamie Dimon, the CEO of JP Morgan Chase, regarding the profits the banks, Visa, and Mastercard were enjoying through their interchange fees.
"For years, card-issuing banks like Chase have agreed to let the Visa and MasterCard duopoly fix the interchange fee rates that banks receive from merchants each time a debit card is swiped. The banks get the fees but they do not set the fees...
Visa and MasterCard have incentive to constantly increase interchange rates and there is no countervailing market force to temper these fee increases. Visa and MasterCard want as many of their debit cards to be swiped as possible because they are paid a network fee by merchants each time a card is swiped. By raising interchange rates, Visa and MasterCard can entice banks to issue more of their cards. Because Visa and MasterCard have enormous market power and control around 80 percent of the debit cards in consumers' wallets, merchants cannot realistically say no to accepting Visa and MasterCard and have no leverage to negotiate fee rates with them."
Of interest is the statement that there is no countervailing market force to temper the interchange fee increases. However, the market may have already produced a new, competitive means of payment to be carried on an alternative network: smartphones. Nicknamed in some articles as "m-commerce", mobile phones are a big growth area for digital payments. Multiple organizations are in a struggle to see who will win in this new digital age. As profiled in a Bloomberg article, AT&T, Verizon to Target Visa, MasterCard With Smartphones, dated August 2, 2010, the carriers have banded together to take on Visa and Mastercard. The payments would be carried by the Discover network.
Analysts of the payment card industry felt that merchants would be willing to embrace the new payment method as they have been battling with Visa and Mastercard over fees for many years. This type of payment system is already effective in several countries outside the U.S., including Japan and the U.K. As cited in the article, an industry consulting firm, Mercatus, states "More than half of U.S. consumers, and almost 80 percent of those between the ages of 18 and 34, will use mobile financial services within five years". If Visa and Mastercard do not find a way to embrace the newer technology, they stand to lose their positions of power in the payment industry.
To ensure they'd don't lose control, Visa and Mastercard have been developing their own plans for mobile phone payments. In contrast to earlier actions cited in the case in Baye's Managerial Economics and Business Strategy, this time they don't appear to be perfectly aligned. Visa has multiple initiatives underway, including joining the network AT&T and Verizon started, now named the Isis Network, as well as developing a concept called the "digital wallet". Described as a "wallet in the cloud", it will be open to competing card vendors such as American Express. Mastercard has in turn developed a partnership with Google and acquired DataCash, which handles online payments.
The rush to develop market dominance in digital payments doesn't stop with the mobile carriers, Discover, Visa, and Mastercard. American Express is developing partnerships with Facebook and others to promote its own method of supporting digital payments. PayPal is also introducing electronic payment capability via mobile phones. Their offering doesn't require a special near field communications (NFC) chip in the phone, or a special NFC-enabled reader in the store. They are betting that merchants will drag their feet in re-equipping their stores, and therefore a non-NFC method of payment will gain the largest market share.
While the switch to mobile phone payments won't happen overnight, it seems to be the wave of the future. Only time will tell if the Visa and Mastercard duopoly will be broken by the disruptive changes coming with m-commerce. Some industry analysts feel the mobile carriers can mount an effective challenge, given their expertise in handling large scale billing and payments. In addition to the firms who will process the payments, there is uncertainty surrounding the technology to enable mobile phone payment, with multiple alternatives being backed by different players in the market. The winner will also be determined by the mobile phone vendors, who must provide the phone hardware, and the massive network of merchants who must decide which technology they'll adopt. It's too soon to tell who will win. Only one thing seems certain: at some point, carrying credit cards will seem as quaint as carrying checks or cash feels now.
Oct.14, 2011
Last week I examined whether the dual management structure of Visa and Mastercard inhibited competition and therefore innovation, using the newer chip technology as the representative example of innovation in the industry. This week I'll focus on how Mastercard and Visa's duality may affect the evaluation of performance by the member banks, and therefore affect overall market share and innovation, as discussed in the paper, On Nonexclusive Membership in Competing Joint Ventures, by Jerry A. Hausman, Gregory k. Leonard, and Jean Tirole, and published in spring, 2003.
As mentioned below,Visa and MasterCard have roughly 75% of the dollar volume of transactions, and 86% of the general purpose cards issued. In the paper, the authors note that "over 6000 entities issue credit cards". They state that the HHI of the issuers is quite low. However, the issuers must work within the services provided by Visa and Mastercard, and duality gives them an unusual level of insight into the performance for both. Objective measurements, such as percentage of the market, are available to everyone, including non-members. But detailed comparisons in terms of overall system performance, management of the associations, and incentives to provide one card vs. another are available only to member banks. According to Hausman et. al, that evaluation can affect which cards they offer their customers, thereby driving and possibly changing market share. According to their research, the sensitivity of the members would differ if their membership was exclusive vs. non-exclusive, as it is today. If membership were exclusive, meaning a member bank could only issue either Visa or Mastercard, the banks would be more sensitive to large quality issues, and there would be an incentive for member banks to switch all of their customers from one card to another if quality were poor enough. With the current system of duality, member banks are more sensitive to small changes in quality, as they find it less expensive to switch customers incrementally. This is most often achieved by convincing a new customer to take one card over the other.
What is the end result of this sensitivity? According to the authors, "we conclude that duality facilitates benchmarking for incremental innovations and inhibits benchmarking for large ones." While they find that the associations have opted for the economically efficient level of investment in innovation, they seem to feel that the type of innovation encouraged by duality is the slower, more incremental type rather than the larger, more disruptive changes encouraged by exclusive, and more competitive, arrangements.
Hausman, Jerry A., Leonard, Gregory K., and Tirole, Jean. On Nonexclusive Membership in Competing Joint Ventures, The RAND Journal of Economics, Vol.34, No. 1 (Spring, 2003), pp. 43-62.
Oct. 3, 2011
Visa and MasterCard are the two largest providers in the market for general purpose credit card network products and services. Together, Visa and MasterCard account for 75% of the dollar volume of transactions, and account for 86% of the number of general purpose cards issued. American Express, Discover/Novus, and Diners Club are the significant competitors in the general purpose card market. Both Visa and MasterCard are joint ventures (associations) that are owned and operated by the member banks that issue cards and provide card acceptance services. The member banks have an interest in both Visa and MasterCard (known as duality) and have representatives serving on the board of directors or important committees of both Visa and MasterCard.
Since the member banks of Visa and MasterCard have a significant interest in both Visa and MasterCard, they have little incentive or desire to compete directly against one another. The member banks have refused efforts by MasterCard management to develop its brand through marketing campaigns. The banks felt that this action would hurt the Visa brand, which the banks have a stake in. They have also not supported new product development, unless the new development was available to both Visa and MasterCard.
In addition to avoiding competition between brands, the member banks have prohibited its banks from issuing competing credit cards, such as American Express and Discover/Novus. They have also worked jointly to discourage merchant acceptance of competing cards and access to ATMs for cash advances. Related Textbook Material and Suggested Teaching Opportunities
The primary parties:
• Visa and MasterCard are joint ventures owned by the member banks that issue cards and provide card processing services. • Visa and MasterCard are separate entities with separate management and staff. However, the member banks provide representatives to the board of directors and important management committees. Therefore, both Visa and MasterCard have representatives of the same member banks setting their corporate policies and strategies.
The Competitors:
• American Express, Discover/Novus, and Japan Credit Bureau (JCB) are the primary competitors in the general purpose card market. Many retail companies issue their own cards for use at a single retail store, but these services do not compete directly with Visa and MasterCard.
Barriers to Entry:
• Barriers to entry are very high in the credit card market. In order to gather a large network of merchants, credit cards must establish a large consumer base. However, it is very difficult to gather a consumer base until there iswidespread merchant acceptance. This Catch-22 makes it extremelydifficult for new entry into the market.
The Primary Issues:
Dual ownership of Visa and Mastercard is believed to lessen competition between them. Restraint of competitors prevents them from competing in the market
• The member banks that own Visa and MasterCard do not want the management of one of the associations to compete in a way that would put the other at a disadvantage. Neither Visa nor MasterCard has made efforts to distinguish its brand from the other; because of the feeling this could harm the other brand. • Competition is also stifled in the development of new technologies. The member banks have resisted the development of new products unless the technology would be available to both Visa and MasterCard.
How have Visa and MasterCard worked to restrain its competitors from competing in the market?
• Visa adopted a rule that prohibited its member banks from issuing credit cards that it deems to be competitive, including Discover/Novus and American Express. This rule did not apply to banks issuing MasterCard credit cards. MasterCard adopted a similar rule to prevent its member banks from contracting with American Express.
• In the mid-1980's, Visa and MasterCard passed rules that prohibited American Express and Discover/Novus from processing credit cardtransactions through their terminals. This exclusion made it difficult for American Express and Discover/Novus to gain acceptance from merchants, since the merchants would have to operate separate terminals. Eventually they loosened these restrictions due to merchant demands, however,processing fees were higher for American Express than for other third party processors.
• Visa and MasterCard have also tried to exclude other competitors from the automated teller machine (ATM) market. Visa and MasterCard ownworldwide ATM networks, Plus and Cirrus respectively. Visa andMasterCard allow users to use both networks, but exclude American Express and Discover/Novus from using their ATM networks. As a result, American Express and Discover/Novus have had to negotiate anddevelop ATM network agreements with smaller fragmented networks.
What evidence is there that Visa and MasterCard would compete more aggressively without duality?
• Exclusionary rules do not apply outside the United States, so the banks have credit card companies have more incentive to compete. Several foreign banks have relationships with American Express and Visa orMasterCard. Therefore, MasterCard and Visa have an incentive to provideinnovative products and services,to ensure that the banks continue to promote their products.
I have downloaded a number of articles on network effects and collusion to begin my research. This week I'll briefly explore the basis for the claim that Mastercard and Visa's relationship restrained competitive behavior, which in turn limited the adoption of new technologies and services which could have benefited the consumer.
The smart card serves as a classic technology example highlighting how consumers can be affected by stifled competition. Per the case in Baye, "a smart card differs from the cards in the U.S.in that it can store information on an integrated circuit, instead of, or in addition to, a magnetic stripe. Integrated circuits are capable of storing significantly more information than magnetic stripes." They are often used in banking, healthcare, and government.
According to the case, in the 1980s, MasterCard developed and extensively tested smart cards.They determined in 1987 that it would give them a competitive advantage over Visa. However, bank representatives on MasterCard's executive committee wouldn't move ahead unless Visa would also do so, and Visa said no. The important point to note is that Visa said no in the U.S. Smart cards are the preferred choice in Europe, including cards issued by Visa and Mastercard. The reason, according to the case, is that outside the U.S., Visa and Mastercard do compete against one another.
A September 13, 2011 posting on Credit Newsline at http://creditnewsline.com/news/visa-and-mastercard-pushing-u.s.-towards-smart-cards/, noted “cards that use chips provide much more security compared to magnetic stripe cards, according to CBSNews.com. As a result of the use of magnetic stripe cards, the United States has higher fraud rates for counterfeit cards, which costs banks, merchants and consumers billions every year.” In addition to experiencing greater security problems at home, Americans are also inconvenienced as they travel outside the U.S., because in certain countries, it is becoming more and more difficult to use our magnetic stripe cards. Newer machines cannot read the information on our cards. Our technology is becoming obsolete.
In "United States v. Visa USA, Inc., 1999-2 Trade Cas. (CCH) ¶ 72,584, the Department of Justice alleged that the Visa and MasterCard corporations agreed upon a number of practices, including an agreement not to engage in certain types of product development. The firms allegedly agreed not to develop and market, or delayed development and marketing of smart cards, commercial cards, and methods for making Internet transactions more secure." (Robert H. Lande & Howard P. Marvel, The Three Types of Collusion: Fixing Prices, Rivals, and Rule, Wisconsin Law Review, 2000:941). In 2001, the District Court determined that the government had not provided enough evidence to prove its case that Visa and Mastercard's dual governance unreasonably restrained trade. However, the judge determined their rules preventing their member banks from issuing American Express or Discover cards did unreasonably restrain trade under Section 1 of the Sherman Act.
It may be difficult for the government to make the case that Visa and Mastercard's dual management structure has inhibited competition and therefore innovation. Fortunately for the American consumer, the technology finally will be adopted: Visa has reconsidered and according to Credit Newsline, is now requiring that processors in the United States be able to support merchant acceptance of chip transactions by April 2013. Needless to say, Mastercard also recently announced it would roll out the technology.
Rachel Wentink
December 6, 2011
Final Paper
Overview of the x86 Microprocessor Industry
This paper will examine the nature of the x86 microprocessor industry and the competitive relationship between its two primary players, Intel and Advanced Micro Devices (AMD). It will touch on how firms may behave in a duopoly, especially in regards to innovation and anticompetitive behavior, as well as network effects and consumer lock-in.
Computers rely on what are called central processing units or CPU. The CPUs are the “brains” behind today’s desktop computers and laptops, performing the instructions provided by computer programs such as operating systems and applications. The CPUs built for desktop computers and laptops are called microprocessors because they are built onto a single silicon chip, or integrated circuit. Most of the microprocessors made for desktop computers and laptops are from a family named x86 after an early product, the 8086. Intel and Advanced Micro Devices (AMD) are the two primary competitors within this industry. According to a February, 2011 article in Information Week Intel holds an 80.7% share of microprocessor shipments worldwide. AMD’s share is 19% of the worldwide market.
Intel and AMD are publicly traded corporations, Intel on the NASDAQ, and AMD on the New York Stock Exchange. As would be expected from two competitors, they do not share any common board members. AMD has two board members from IBM, but Intel’s does not contain any PC manufacturers.
The Corporate Profile on Yahoo Finance describes Intel primarily as a firm that “. . . engages in the design, manufacture, and sale of integrated circuits for computing and communications industries worldwide. It offers microprocessor products used in notebooks, netbooks, desktops, servers, workstations, storage products, embedded applications, communications products, consumer electronics devices, and handhelds. “
According to Yahoo Finance, AMD is a U.S.-based semiconductor company selling microprocessors for servers and notebook PCs. It addition, it “ . . . provides embedded processor products for vendors in industrial controls, digital signage, point of sale/self-service kiosks, medical imaging, set-top box, and casino gaming machines, as well as enterprise class telecommunications, networking, security, storage systems and thin-clients, or computers. In addition, the company offers chipset products, including integrated graphics processor chipsets and discrete chipsets for desktop and notebook PCs, professional workstations, and servers.”
Early History and Competitors
The history of Intel and AMD is inextricably tied to a firm named Fairchild Semiconductor. Robert Noyce, considered the “father of the silicon chip”, was general manager at Fairchild, and among his direct reports was an engineer named Gordon Moore, who headed up research and development. As Fairchild became more bureaucratic and less innovative, many of the original founders moved on to other companies. In 1967, National Semiconductor moved from Connecticut to Santa Clara and began hiring many of Fairchild’s talented engineers. Believing that Fairchild was irreparably harmed by the loss of human capital, Noyce and Moore founded Intel in 1968. Their intent was to build integrated memory circuits and competing with Fairchild Semiconductor and National Semiconductor.
AMD’s founders also came from Fairchild Semiconductor, roughly nine months after Intel’s founding. By the time they decided to move, investors felt that the field was already full of viable competitors in the semiconductor industry. While Noyce was able to secure funding for Intel in one afternoon, AMD’s funding was much more difficult to secure, and up until the last moment the founders were unsure they would raise enough funds to move ahead. In fact, Noyce invested in AMD, even though he knew they could turn into a competitor of Intel’s in the future. There is no information to imply that Noyce’s connection led to any collusive activity, or that he had any influence whatsoever on AMD’s plans.
In addition to Fairchild Semiconductor and National Semiconductor, other competitors in the industry in those early days were made up of a mix of existing and newly founded companies. They included Motorola, Mostek, founded in 1969 by ex-employees of Texas Instruments, and Zilog, founded in 1974 by two ex-Intel employees, one of whom had invented the microprocessor. More recent competitors include Via, a small Taiwanese firm, and Transmeta, who sold x86 microprocessors for notebooks, but exited the market in 2006.
Initially, Intel began to build memory chips, but in 1970, the Nippon Calculating Machine Corporation contracted with Intel for a set of chips. Federico Faggin, was assigned to the project instead, and designed and built the chips. Intel’s history would have been dramatically different had the Japanese firm stuck to the original terms of the contract. But instead, the calculator market had become much more competitive, and they came back demanding a price cut. Robert Noyce arranged to refund some of the money to the client for the rights to the design of the chips, including the microprocessor. Intel launched its first microprocessor, the 4004, in November, 1971, gaining a first-mover advantage.
Investopedia defines a first-mover as “ . . . being the first to enter a specific market or industry. Being the first allows a company to acquire superior brand recognition and customer loyalty. The company also has more time to perfect its product or service. “
In Managerial Economics and Business Strategy, author Michael Baye ascribes multiple advantages to being a first-mover in an industry: “Being first may permit her to enjoy monopoly profits throughout the life of the patent. Due to learning curve effects, additional first-mover benefits may accrue in the form of lower costs. First-mover advantages due to learning curve effects can persist long after the patent expires, and also can be present even if there is not a patent in the first place.” (pg. 492). Learning curve effects were present in the semiconductor industry, and played a key role in barriers to entry.
Barriers to Entry
Barriers to entry were and are high in the x86 industry. Entering the market requires significant investment in fabrication plants and equipment, and in the human capital necessary to effectively design and manufacture the chips successfully. Human capital and learning effects have been extremely important in the industry because of the difficulty in manufacturing a high percentage of usable chips. With each new design, cramming many thousands more transistors onto a single silicon chip, the manufacturing challenge would re-emerge. Rates of saleable chips could be very low during the initial manufacturing runs for a new design.
Gordon Moore is famous in the computer industry for Moore’s Law. In a 1965 article in Electronics Magazine, he famously predicted that the number of components on a single chip would double each year. In 1975, he updated his prediction to once every two years. He also predicted that the cost to manufacture the chips would drop significantly. His prediction was dubbed Moore’s Law in 1970, and the industry results adhered to Moore’s Law through the 1990s and into the 2000s. The implications meant that Intel could not bring in high margins at the lower end of their product line. They had to constantly innovate in order to offer higher margin products, as other vendors would compete on price for the less powerful chips. The rapid, constant innovation also increased the need for learning.
Studies have been performed on the importance of “learning by doing” in the semiconductor industry, (Irwin and Klenow, 1994), (Hatch and Reichelstein, 1994), where considerable knowledge is built up and shared on how to resolve manufacturing problems with the sensitive chips. Hatch (1996), as cited in Salgado (1997) describes the challenge: “It is not uncommon for yields for new semiconductor processes to start as low as 10%. In such cases, the cost of scrapped output is initially very high but falls to low levels as yield rises to 90% or higher over time”. Also, according to Creating the Digital Future: The Secrets of Consistent Innovation at Intel by Albert Yu, wafer size and yields (the percentage of usable chips) dramatically impacts profitability in the semiconductor industry. (pp.85-86). Intel, recognizing the value human capital could bring in resolving these manufacturing challenges, exploited its first-mover advantage to hire the best and the brightest in the industry and in the academic world. With the benefit of the more famous name, it could court brilliant engineers more successfully than AMD or other semiconductor firms.
In addition to the investment in plant, equipment, and human resources, barriers to entry also include establishing relationships with the manufacturers who would buy the chips to use in their finished products. Another barrier to entry can be caused by network effects, as discussed in the next section of this paper. Consumer lock-in to a particular product or service can serve as a barrier to entry for other potential competitors who wish to establish their own standard.
Network Effects and Consumer Lock-In
The microprocessor industry makes an excellent case study for what economists like to refer to as positive indirect network externalities or network complementarities. According to Katz and Shapiro (1994), consumers benefit when they utilize a network such as a telecommunications network, and others are also using the technology. Translated, it means being the sole user of a telephone wouldn’t provide much value. The largest part of the value is in the number of users on the network. These are typically referred to as network effects, or direct network externalities as the users are on the same physical network. Katz and Shapiro also bring up a similar effect with durable goods, which they describe as the “hardware/software paradigm”. The consumer must not only determine which hardware they prefer, but also must determine how much software may be available for it. Along with computers, they mention credit cards networks as an example, in which the card is the hardware, and the merchant acceptance acts as the software. With computers, the benefit is a network complementarity, or indirect network externality because of the size and availability of complementary products and/or services.
These positive network complementarities make a product more attractive to a consumer. A classic example is the attractiveness of the iPhone. When is it considered as just a phone, it is not nearly as attractive as when it is considered as a device with over 500,000 available applications, including games. That positive complementarity can have a dark side, however, in that it can tie consumers to a particular good or service because of the benefits of the network, in a phenomenon known as “consumer lock-in”. The partnership between Microsoft, with its Windows operating system, and Intel as the de facto standard for personal computer microprocessors, is a type of network that locked consumers in. Called the “Wintel” platform, consumers remained with these types of computers because of the greater number of applications, the ability to share documents and other files with other computer users, and the ease of finding qualified support.
In the early 1970s Intel continued to innovate in microprocessors, but considered the memory chip their main line of business. Microprocessors weren’t bringing in that much money, as there weren’t that many products in which they could be used. Intel turned down opportunities to build the computers.Gordon Moore was not interested, as he couldn’t see the reason why someone would want one. He couldn’t envision the positive network complementarities.
Intel might not have seen the potential, but by 1977, external firms began to offer smaller computers using the Intel chips. MITS released the Altair computer kit, using Intel’s 8080 chip. Radio Shack also had a kit. Steve Wozniak and Steve Jobs were also innovating, with a portable computer that could actually be picked up by a person.
A program called Visicalc was written for the Apple II. It was the first commercial spreadsheet, and was designed for use by accountants and bookkeepers. It helped to answer Gordon Moore’s question of why someone would want a portable computer. It demonstrated a use for these computers in the business world, and demonstrated the potential market for creating software programs for the devices. Others began writing programs, including word processing programs which freed up secretaries from the tyranny of the typewriter. Network complementarities began to take off for the portable computer.
An almost accidental sales call on IBM secured Intel an order to provide the 8088 microprocessor to IBM for its launch of a personal computer. IBM insisted on a second source, and Intel turned again to AMD. A contract was negotiated in 1982 that would last for 10 years, in which AMD would act as a second source for Intel but also for Intel to second source for AMD as well. Ultimately they ended up in a court battle over the contract.
Motorola had supplied its 68000 processor for the Apple computer, but IBM’s PC, using Microsoft’s MS-DOS as the operating system and Intel’s microprocessor became the de facto standard for much of the personal computer industry by allowing “clones” to be made by other manufacturers. Because of the contract with IBM, Intel’s chips became the standard microprocessor for the clone industry. The 8088 was based upon the 8086 chip, hence the x86 name for the line of microprocessors. Intel swept AMD along in its wake as the second source.
Market Model and Margins
With the IBM release of the PC with Intel’s chips, the founders of Zilog realized their hopes to become a major factor in the microprocessor market were over. Both founders cashed out. Zilog continues today, but once the founders left, it was no longer a credible rival. Other rivals also either disappeared, or focused on other lines of business. As Microsoft released a new operating system, called Windows, the “Wintel” platform, as it is called, became the largest percentage of the portable computer market. While Apple succeeded early on, over time it became a niche player in the market, which meant suppliers for its processors were also relegated to niche status. Intel and Microsoft were now the standard platform. The combination of Windows and Intel, dubbed Wintel, held the majority share of the portable computer market from the point that IBM launched its PC. In fact, in July, 2011, the “Wintel” platform share of the market was declared to be at its lowest level in 20 years at just under 82% by IT analyst firm Canalys. The success of the Wintel platform did force a type of consumer lock-in: if businesspeople and home users wanted a wealth of applications to choose from for their personal computers, for many years they had to choose Wintel.
A pattern emerged, in which Intel would release a new chip which was faster and more powerful, Just as Moore’s Law had predicted. Intel would exercise monopoly power with that chip until AMD could bring to market a comparable, or in some cases, slightly more powerful chip. According to Andy Grove: The Life and Times of an American, the Microprocessor Report on November 28, 1990 carried a story titled “AMD Ends 386 Monopoly”. (pg. 256). At that point the market for those microprocessors acted as a duopoly, with falling prices and margins. Intel would already be at work on the next microprocessor model, and the pattern would repeat itself. The author states, “By 1992, as Intel was busily at work trying to move the world to the 486, AMD had achieved a market share in the 386 class microprocessor of greater than 50 percent.” (Pg. 257). Rarely did AMD bring a significantly innovative and better microprocessor to market before Intel, or at least, that’s how it appeared.
In the early years of the market, Intel could count on new PC buyers to increase its revenue. However, as pointed out, market share and prices would fall rapidly for microprocessors as soon as AMD could release a comparable product. According to the U.S. Department of Commerce’s Bureau of Economic Analysis “ . . . there is a high degree of turnover in this segment as new, faster chips are brought to the market. Second, prices fall steeply over the life of each chip; prices typically start at between $600 to $1000 at introduction--substantially higher than the prices of existing chips. By the time the chip exits the market, its price has fallen to under $100. The steepness of these profiles could reflect demand- or supply-driven forces. On the demand side, the profiles are consistent with the view that users are initially willing to pay high prices for new chips but as the introduction of the new (better) chip nears, they are less willing to do so and prices of the incumbent chips fall.” (pg.14).
From Intel’s perspective, rightly or wrongly, one of the primary reasons why the margins declined so rapidly was the existence of AMD.
Intel Potentially Restrains Competition
An oligopoly is defined by Baye as “A market structure in which there are only a few firms, each of which is large relative to the total industry.” (pg.314). A duopoly is defined on Investopedia as “A situation in which two companies own all or nearly all of the market for a given type of product or service.” This definition clearly fits Intel and AMD as the microprocessor industry gained traction. When one member of the duopoly is so much larger than the other, questions arise about competitive behavior.
In “A Dynamic Oligopoly With Collusion and Price Wars”, Fershtman and Pakes (2000) posit an industry with at most a handful of competitors and use game theory to determine the likelihood of collusion. If there are two firms, their conclusion was that “collusion breaks down when one of the two is near an exit state, but again just how near depends on where the larger competitor is.” (pg. 221). They point out two reasons for collusion being hard to sustain in such an industry. In large part, it has to do with the potential for the smaller firms to exit the industry in the foreseeable future. In theory, the potential for exiting soon takes the game from an infinitely repeatable one to one of a shorter, potentially definable period, as described by Baye on page 363. The closer a small firm is to exiting, the more definable the end period, and the greater the incentive for the small firm to cheat. However, there is also an incentive on the side of the larger firm, which is not covered in the game theory chapter in Baye. According to Fershtman and Pakes, there is incentive to predatory behavior on the part of the larger firm. “The large firm realizes that the future will be better if it can ‘force’ the small firm(s) to exit and then monopolize the market until the next entrant arrives. Since the smaller firm is more likely to exit if there is no collusion, the larger firm has an incentive to deviate from the collusion pricing in order to increase the likelihood of the smaller firm exiting.” (pp. 221-222).
Fershtman and Pakes suggest the potential for collusion is more likely when there are two firms who can be described as strong incumbents. While Intel can be characterized as a strong incumbent, AMD cannot. AMD more closely matches the description of a firm likely to exit the industry, especially if Intel could find a way to starve it of revenue. Intel’s predatory behavior didn’t result in lower prices for newer microprocessors in order to drive AMD from the industry. Instead, as Intel attained a significant portion of the market and built its reputation, it would appear that it took steps designed to reduce AMD’s revenue to an untenable level.
The relationship between Intel and AMD began to deteriorate from the point of their 1982 licensing agreement. Jerry Sanders, the CEO of AMD, could not resist describing the great deal AMD had achieved to the press, and characterizing Intel as having gotten the worst end of the bargain. Some say Andrew Grove took offense at his language. Whatever the reason, Intel began a campaign to try to undercut the agreement and force its early termination. The agreement called for the possibility of Intel accepting product designs from AMD and opting to act as a second source for them. Second sourcing was a standard practice in the semiconductor industry. It was followed by firms who would not risk a high profile project by relying on only one supplier of a critical input. Second sourcing required the leading firm to share its design with the second source, who would then manufacture a similar part, often for less money. Through second sourcing, there was a significant knowledge spillover in the industry.
If Intel accepted any of AMD’s product designs and agreed to second source them, it would reduce or remove the need for AMD to pay royalties for second sourcing Intel’s product designs. AMD wanted to second source the 80286 chips. AMD proposed two product designs Intel could second source. According to Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company, an internal memo within Intel has since come to light details Intel’s strategy at that point:
“1/ Assure AMD they are our primary source through regular management contact and formal meetings
2/ Take no more AMD products under the current agreement” (pg.233)
Intel had determined that it had enough of a hold on the market to sell the upcoming 80386 chip without a second source. AMD would no longer be able to undercut Intel’s margins. Intel’s management believed by not second sourcing the new chip, it could keep the price of the 80386 higher for a longer period of time. However, during this period, it needed to convince AMD it would give them the 80386 design for second sourcing while still using AMD as a second source for the 80286 chips.
Another memo outlining Intel’s policy states, “Maintain a second-source, business as usual posture in the market place. . . . We do not want them [AMD] to go to Hitachi or NEC, and should not stimulate them to do so.” (pg. 234).
As the situation dragged out, AMD determined to invoke the arbitration clause of their contract. Originally they thought it would be a short process, estimating roughly 12 weeks. The arbitration, which ultimately took three years, began a series of bitter legal battles which weren’t settled until 2009.
Halting second sourcing was one means of starving AMD of revenue. AMD was forced to then create its own designs, and for the most part, delivered competitive, value-priced microprocessors. Intel was then accused of taking additional steps to prevent AMD’s successful competition: they began paying rebates to PC manufacturers such as Dell and HP. The rebates required the manufacturer’s to buy less of AMD’s products. According to the European Union’s competition commissioner, during the period of October 2002 – December 2007, the rebates helped Intel maintain a share of at least 70 percent of microprocessor sales.
In 1991, AMD filed an antitrust case against Intel. Per an article in Business week, AMD claimed Intel was behaving in an anti-competitive manner, stating in court documents that it "used multibillion-dollar payments to coerce PC makers such as Dell (DELL) and Hewlett-Packard (HPQ) into using Intel products to the exclusion of AMD's chips."
At the same time, Intel had a long-standing patent dispute with AMD: it had an agreement with a subsidiary to allow for patent-sharing. AMD moved to spin off the subsidiary as a separate company, GlobalFoundries. When GlobalFoundries tried to take the patent agreement with it, Intel cried foul, stating that the agreement did not allow AMD to extend it to a third-party organization. Without the patents, AMD and GlobalFoundries would not be able to manufacture chips of any market value.
In April, 2009, Intel paid AMD $1.25 billion dollars to settle the allegations of antitrust violations. The agreement called for quarterly meetings between the companies and methods to mediate disagreements in the future. It also enabled the firms to cross-license each other’s patents for five years.
The settlement with AMD helped take some of the pressure off Intel, who had been ordered by EU regulators in May of 2009 to pay a fine of 1.45 billion for antitrust violations. As part of the determination, according to the New York Times, Neelie Kroes, the competition commissioner for the EU, "ordered Intel to stop offering rebates that were conditioned on buying less of a rival’s product, or not buying them at all..." In July of 2009, Intel appealed the verdict, but on average the court takes approximately two years to reach a verdict, and the determination has not yet appeared.
In the meantime, the FTC was also threatening to investigate Intel, and there was speculation that the agreement with AMD was intended to ward off a lawsuit. However, Intel's anti-competitive woes were not yet over; the FTC moved forward with its case. In addition to the problem of the rebates, the FCC also accused Intel of a more insidious method of undercutting AMD—that of changing its compiler, used by software developers, so that the performance of non-Intel microprocessors would degrade during comparison testing.
Benchmarking firms are widely used by IT publications to compare performance by selective computer components. Comparisons of performance across substitutable components act as a standard means of informing IT organizations and consumers on the potential quality and benefits of those components. By changing the compiler and not informing the development community of its act, the FTC says that Intel essentially stacked the deck in their favor.
In their public document outlining the reasons for pursuing Intel they stated, “Among the harms to consumers caused by Intel’s deceptive conduct was the harm to the credibility and reliability of industry benchmarks. Industry benchmarks are important tools for consumers to make informed purchasing choices. Informed consumer choice is a basic building block of competition.“(pg. 6).
In December, 2010 Intel settled with the FTC.
Did a Duopoly Lead to More Innovation?
In "Does AMD Spur Intel to Innovate more?" (2011), authors Ronald Goettler of the University of Chicago and Brett Gordon of Columbia University examine the question of whether AMD’s existence brought about greater social welfare through increased consumer surplus and innovation, or if consumers and society as a whole would have been better off with Intel operating solely as a monopoly. Their sample data spanned the period of 1993 to 2004.
One of the primary points of the paper is that the industry behaves differently when the product is durable, as are CPUs and PCs in general. They feel duopolies and monopolies in durable goods markets face different challenges than those in industries selling services or perishable products. A monopoly in a durable goods industry does face a competitor: themselves. Translated to the personal computer microprocessor industry, the challenge becomes one of convincing consumers and businesses that they need to upgrade to newer, faster chips, even though the older ones still function. The paper cites the 2005 Computer Industry Almanac to show that 82 percent of PC purchases in 2004 were replacements. For all practical purposes, once the PC market began to slow down, Intel would have faced competition in order to grow sales significantly, even if AMD did not exist.
As a result, their point was that durable product firms must continue to innovate to stimulate demand for replacement products. Had Intel been a monopoly, it would still have had to innovate to stimulate demand for its microprocessors. Although not mentioned in the paper, a potential positive network effect was the development of more complex personal computer software which required faster clock speeds in order to provide an effective user experience. Popular, processor-intensive software products stimulated demand for newer personal computers with more powerful hardware.
They also felt that duopoly innovation is more sensitive to consumer preferences. A monopoly does not have to focus as heavily on consumer desires—when two firms are competing for the market, they listen more to their potential customers than a single firm will. This point doesn’t necessarily mean that innovation is highest in the duopoly, however. The authors found that in their data sample, rapid market growth, low price sensitivity and high quality preferences favor a duopoly. Other conditions could favor a monopoly. In fact, the authors concluded that in the case of the personal computer microprocessor industry, for their sample data, innovation would be higher if Intel were a monopoly. “The finding that innovation by a monopoly exceeds that of a duopoly reflects two features of the model: the monopoly must innovate to induce consumers to upgrade, and the monopoly is able to extract much of the potential surplus from these upgrades because of its substantial pricing power. If competition with itself were reduced by a steady flow of new consumers into the market, the monopoly would reduce innovation below that of the duopoly."
Their conclusion was that consumer surplus is higher in a duopoly. While innovation would be higher if Intel were a monopoly, AMD drove prices down and as a result, margins dropped for the firms, and consumers benefited.
They also felt the percentage of innovation spillovers affect the pace of innovation and whether a duopoly or monopoly would innovate more. The shape is an inverted U: if innovation spillovers are very high, the leader cannot maintain the lead. If spillovers are very low, the smaller firm cannot glean enough information to keep it competitive, and it must exit the industry. The spillover effect is partly why a monopoly would innovate more -- the leader, concerned that the follower is copying their technology, limits their innovation more than they would if they had no competitor. It should be noted that Intel and AMD cross-license their technology, which muddies the knowledge spillover somewhat. During the period studied, both firms had healthy R&D budgets, with Intel at approximately 11% and AMD at approximately 20%.
One potential way to alter the market to Intel and AMD’s benefit, although not to the consumers’, would have been to reduce the durability of microprocessors. The authors found what appears to be a prisoner’s dilemma in the industry: Intel and AMD would like to create less durable products, but for a variety of reasons, they could not agree to do so.
Ultimately the authors concluded that consumer surplus was higher by 4.2% with the existence of AMD providing competition external to Intel, and driving margins down. At the same time, society would have benefited as a whole by greater innovation, and their sample data pointed to a potential 4.2% increase in innovation had Intel been a true monopoly.
The Future of the Industry
As for future competition between Intel and AMD, consumer tastes have changed. It has shifted demand. Consumers are moving in overwhelming numbers from personal computers to more mobile devices, which has affected the x86 microprocessor producers. The new growth industry is in developing chips for smartphones, electronic tablets and other mobile devices, and Intel is still trying to get a foothold in the new market.
According to eWeek in “A Chip for All Reasons”, when addressing Intel’s investors earlier this year, CEO Paul Otellini said: “There’s been so much written about tablets that I don’t know where to start, except to say we’re on track. “We’re tracking 35 designs on multiple operating systems. Some are shipping now with Windows. We’re demonstrating some Android devices now. The tablet race is nowhere near finished. . .You’ll see the first Intel-based phones [using new Medfield chips] in the market the first part of next year”.
Within the data center or PC server market, the microprocessor battle is moving to support multiple cores to handle the large capacity needed for cloud services. The same article quotes an analyst who follows Intel. “ . . . When you look at the cloud services world—meaning Google, Facebook, Amazon, Microsoft with Azure—they’re all running x86. There’s very little of non-x86 in all those operations.” At the time the article was published, in 2010, AMD appeared to be primarily focused on cloud services. eWeek quoted Rick Bergman, who headed up the company’s processor and computing platforms initiatives at the time, .“ … despite excitement among its competitors about serving handheld devices, [AMD] doesn’t have “any immediate clients to serve the mobile form factor. But it will ‘move into that going forward.’ “AMD’s core processor and GPU technology, he said, “play great in handheld devices as well, but we’ve got a lot of opportunity we’re going to focus on elsewhere.”
Conclusion
For the moment, there is still a duopoly market for x86 personal computer processors with relatively high barriers to entry. That market is less relevant to the future growth of Intel and AMD. There is little in the way of exciting new development for desktop microprocessors. Now roughly 40 years since the creation of the first computer microprocessor, it is a mature industry in which most purchases come from replacement.In the Visa and Mastercard case presented by Baye, there was speculation that cooperation stifled competition between the two organizations, resulting in less innovation for consumers. The same member banks were represented on the board of each organization, and they refused to approve programs that would give either organization a competitive edge. In comparison, although the founders for Intel and AMD came from the same firm, Fairchild Semiconductor; although one of Intel's founders invested in AMD and enabled their startup; and although initially the two firms were forced to share knowledge and designs through the practice of second sourcing, they developed a bitter rivalry. From the material at hand, it would appear as though Intel deliberately pursued a program of anticompetitive actions against AMD that was anything but cooperative. The two firms spent many years in legal battles. The recent settlement may change the nature of their relationship--that remains to be seen.
According to Goettler and Gordon (2011), if Intel had operated as a pure monopoly consumers would have benefited from approximately 4.2% more innovation. On the other hand, their paper also stated that consumers received roughly the same amount of additional consumer surplus as a result of the duopoly--in this case meaning that prices were lower for microprocessors. From a consumer standpoint, given the tremendous amount of innovation already in the industry, lower prices might have been the better outcome. We'll never know what would have happened if Intel had remained the only x86 microprocessor manufacturer.
Other firms are now attempting to enter the x86 market, which doesn't bode well for either Intel or AMD. According to The Wall Street Journal in "AMD to Cut 1,400 Jobs as New Chief Revamps", " . . . chips based on designs from ARM Holdings PLC architecture not only dominate mobile devices but are working to expand into PCs and servers, the traditional market of Intel and AMD." The entry of other manufacturers will most likely end the duopoly in that portion of the microprocessor market.
The greater opportunities for growth now are in server microprocessors for the infrastructure required for cloud-based services, in which both AMD and Intel are competing, and in microprocessors for mobile technology. Neither will be duopolies. It remains to be seen how the market will evolve.
In a change from last year’s position, AMD is being taken to task by its board of directors for not getting into mobile technology more quickly. As it restructures in the coming months and lays off roughly 10% of its workforce, it promises it will take the savings and invest some of it in new development areas, such as tablet technology and cloud computing. However, both AMD and Intel lack the first mover advantage in mobile technology, and will have to play catch-up. While that role is familiar to AMD, in large part it will be new one to Intel.
Bibliography
Aizcorbe, Ana. (2006). Why Are Semiconductor Price Indexes Falling So Fast? Industry Estimates and Implications for Productivity Measurement. Economic Inquiry, Volume 44, Issue 3 (pp.485-496).
AMD Profile. Yahoo Finance. Retrieved from http://finance.yahoo.com/q/pr?s=AMD+Profile
Baye, Michael R. (2010). Managerial Economics and Business Strategy. New York, NY. McGraw-Hill/Irwin.
Canalys. (2011, July 29). Wintel Share Falls Under 82%. [Press Release]. Retrieved from http://www.canalys.com/newsroom/wintel-share-global-pc-industry-falls-under-82
Duopoly Definition. Investopedia. Retrieved from http://www.investopedia.com/terms/d/duopoly.asp#axzz1flT0WUFh.
Federal Trade Commission. (2010, August 4). FTC Settles Charges of Anticompetitive Conduct Against Intel. [Press Release]. Retrieved from http://www.ftc.gov/opa/2010/08/intel.shtm
Federal Trade Commission. (2010, August 4).Analysis of Proposed Consent Order to Aid Public Comment in the Matter of Intel. Retrieved from http://www.ftc.gov/os/adjpro/d9341/100804intelanal.pdf
Fershtman, C., Pakes, A. (Summer, 2000). A Dynamic Oligopoly with Collusion and Price Wars. The RAND Journal of Economics, Volume 13, No. 2, pp. 207-236.
First Movers Definition. Investopedia. Retrieved from http://www.investopedia.com/terms/f/firstmover.asp#axzz1flT0WUFh
Goettler, Ronald L. Gordon, Brett R. (2011, October 19). Competition and Product Innovation in Dynamic Oligopoly. Working Paper.
Gonsalves, Anton. (2011, February 24). Intel Up, AMD Down in Global Microprocessor Shipments. Information Week. Retrieved from http://www.informationweek.com/news/hardware/processors/229219340
Hesseldahl, Arik. (2009, November15). The Intel-AMD Settlement: A Play-by-Play. Bloomberg Businessweek. Retrieved from http://www.businessweek.com/technology/content/nov2009/tc20091115_692400.htm
Intel Profile. Yahoo Finance. Retrieved from http://finance.yahoo.com/q/pr?s=INTC+Profile
Jackson, Tim. (1997). Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company. New York, NY: the Penguin Group.
Kanter, James. Europe Fines Intel $1.45 Billion in Antitrust Case. The New York Times. (2009, May 13).
Retrieved from http://www.nytimes.com/2009/05/14/business/global/14compete.html?pagewanted=all,
Katz, Michael. L. Shapiro, Carl. (Spring, 1994). Systems Competition and Network Effects . Journal of Economic Perspectives. Volume 8, Number 2, pp. 93-115.
Moore, Gordon E. (1965, April 19). Cramming more components onto integrated circuits. Electronics, Volume 38, Number 8. Retrieved from ftp://download.intel.com/museum/Moores_Law/Articles-Press_Releases/Gordon_Moore_1965_Article.pdf
Preimesberger, Chris. (2011, October 3). Intel Wants to Be the Chip for All Reasons. eWeek. Retrieved from http://www.eweek.com/c/a/IT-Infrastructure/Intel-Wants-to-Be-the-Chip-for-All-Reasons-795129/
Salgado, Hugo H. (2007, April). Estimating Market Structure in a Dynamic Duopoly Model in the Personal Computer Processor Industry. Working Paper.
Tedlow, Richard S. (2006). Andy Grove: The Life and Times of an American. New York, NY. The Penguin Group.
Tibken, Shara. Clark, Don. (2011, November 4). AMD to Cut 1,400 Jobs as New Chief Revamps.The Wall Street Journal . Retrieved fromhttp://online.wsj.com/article/SB10001424052970203804204577016292499557570.html
Yu, Albert. (1998). Creating the Digital Future: The Secrets of Consistent Innovation at Intel. New York, NY. The Free Press.
End Final Paper
Quiz Questions
1. Which of the items below is a possible outcome of a positive network externality or network complementarity?
A) 1,079,002 available applications for the salesforce.com platform
B) Talking to your Aunt Rhoda by phone every Saturday via Apple’s FaceTime program.
C) Multiple Repair Shops for Hondas in every large to mid-size U.S. city.
D) One million views of that embarrassing video of you on YouTube.
E) All of the above
2. According to Fershtman and Pakes, what would prevent collusion in an oligopoly?
A) High barriers to entry
B) The likely exit of the industry by the smaller firm(s)
C) Strong rivalry between firms
D) The sale of durable goods.
3. According to Goettler and Gordon’s paper “Does AMD Spur Intel to Innovate More which firms were Intel’s competition for x86 microprocessors during the period from 1993-2004?
A) AMD
B) IBM
C) Intel
D) ARM Holdings
E) A and C
4. Which benefits below are not potential outcomes of first mover advantage?
A) Monopoly profits due to patents
B) Consumer Lock-In
C) Reduced production cost through greater “know-how”
D) Lower prices
5. Historically, what item(s) below did NOT contribute to the dramatically falling prices in the microprocessor industry in the 1980s and 1990s?
A) Cheaper inputs
B) Learning by doing
C) Continuous improvements in microprocessor power
D) All of the above
Answers
1. E – All of the above. You may not want to speak with Aunt Rhoda, but the fact that you both can do so on the same network (Apple’s FaceTime) makes it a positive direct network externality. And you may hate that video, but the person who posted it is enjoying the benefits of the positive network complementarity.
2. B– The likely exit of the industry by the smaller firm(s). The likely exit acts as an uncertain timeline. The closer the small firm is to exiting, the more certain the end date of the game, which causes both small and large firms to decide it is better to cheat than to collude. See Baye on Games With an Uncertain Final Period, pg. 370.
3. E – A and C. According to Goettler and Gordon’s theory, Intel’s primary competition was Intel because the microprocessors and the computers that used them were durable goods. According to their theory, in durable goods, even a monopoly firm has competition—it becomes its own competitor.
4. D – Lower prices. Baye cites potential monopoly profits, learning curve effects that drive down costs, and first-mover advantage due to consumer lock-in in Chapter 13.
5. A –Cheaper inputs. Historically, cheaper inputs may have lowered prices at some point, but they did not contribute to the dramatic drop in prices. For each new chip design, learning by doing improved manufacturing techniques, which resulted in an increased percentage of usable chips and a lower overall manufacturing cost. As new breakthroughs in upcoming chips were anticipated, consumers would either wait for the latest model before purchasing a PC, or would demand the older PCs with older microprocessors at a lower price.
The Future of the Industry
As for future competition between Intel and AMD, consumer tastes have changed. It has shifted demand, as described by Baye in Chapter Two of our text. Consumers are moving in overwhelming numbers from personal computers to more mobile devices, which has affected the x86 microprocessor producers. The new growth industry is in developing chips for smartphones, electronic tablets and other mobile devices, and Intel is still trying to get a foothold in the new market.
According to eWeek in “A Chip for All Reasons”, when addressing Intel’s investors earlier this year, CEO Paul Otellini said: “There’s been so much written about tablets that I don’t know where to start, except to say we’re on track. We’re tracking 35 designs on multiple operating systems. Some are shipping now with Windows. We’re demonstrating some Android devices now. The tablet race is nowhere near finished.” He later added “You’ll see the first Intel-based phones [using new Medfield chips] in the market the first part of next year”. ...
Within the data center or PC server market, the microprocessor battle is moving to support multiple cores to handle the large capacity needed for cloud services. The same article quotes an analyst who follows Intel. “ . . . When you look at the cloud services world—meaning Google, Facebook, Amazon, Microsoft with Azure—they’re all running x86. There’s very little of non-x86 in all those operations.” At the time the article was published, in 2010, AMD appeared to be primarily focused on cloud services. eWeek quoted Rick Bergman, who headed up the company’s processor and computing platforms initiatives at the time, .“ … despite excitement among its competitors about serving handheld devices, [AMD] doesn’t have “any immediate clients to serve the mobile form factor. But it will ‘move into that going forward.’ “AMD’s core processor and GPU technology, he said, “play great in handheld devices as well, but we’ve got a lot of opportunity we’re going to focus on elsewhere.”
Conclusion
For the moment there is still a duopoly market for x86 personal computer processors with relatively high barriers to entry. That market is less relevant to the future growth of Intel and AMD. There is little in the way of exciting new development for desktop microprocessors.Now roughly 40 years since the creation of the first portable computer microprocessor, it is a mature industry in which most purchases come from replacement.
In the Visa and Mastercard case presented by Baye, there was speculation that cooperation stifled competition between the two organizations, resulting in less innovation for consumers. The same member banks were represented on the board of each organization, and they refused to approve programs that would give either organization a competitive edge. In comparison, although the founders for Intel and AMD came from the same firm, Fairchild Semiconductor; although one of Intel's founders invested in AMD and enabled their startup; and although initially the two firms were forced to share knowledge and designs through the practice of second sourcing, they developed a bitter rivalry. From the material at hand, it would appear as though Intel deliberately pursued a program of anticompetitive actions against AMD that was anything but cooperative. The two firms spent many years in legal battles. The recent settlement may change the nature of their relationship--that remains to be seen.
According to Goettler and Gordon (2011), if Intel had operated as a pure monopoly consumers would have benefited from approximately 4.2% more innovation. On the other hand, their paper also stated that consumers received roughly the same amount of additional consumer surplus as a result of the duopoly--in this case meaning that prices were lower for microprocessors. From a consumer standpoint, given the tremendous amount of innovation already in the industry, lower prices might have been the better outcome. We'll never know what would have happened if Intel had remained the only x86 microprocessor manufacturer.
Other firms are now attempting to enter the x86 market, which doesn't bode well for either Intel or AMD. According to The Wall Street Journal in "AMD to Cut 1,400 Jobs as New Chief Revamps", " . . . chips based on designs from ARM Holdings PLC architecture not only dominate mobile devices but are working to expand into PCs and servers, the traditional market of Intel and AMD." The entry of other manufacturers will most likely end the duopoly in that portion of the microprocessor market.
The greater opportunities for growth now are in server microprocessors for the infrastructure required for cloud-based services, in which both AMD and Intel are competing, and in microprocessors for mobile technology. Neither will be duopolies. It remains to be seen how the market will evolve.
In a change from last year’s position, AMD is being taken to task by its board of directors for not getting into mobile technology more quickly. As it restructures in the coming months and lays off roughly 10% of its workforce, it promises it will take the savings and invest some of it in new development areas, such as tablet technology and cloud computing. However, both AMD and Intel lack the first mover advantage in mobile technology, and will have to play catch-up. While that role is familiar to AMD, in large part it will be new one to Intel.
Quiz Questions
1. Which of the items below is a possible outcome of a positive network externality or network complementarity?
A) 1,079,002 available applications for the salesforce.com platform
B) Talking to your Aunt Rhoda by phone every Saturday via Apple’s FaceTime program.
C) Multiple Repair Shops for Hondas in every large to mid-size U.S. city.
D) One million views of that embarrassing video of you on YouTube.
E) All of the above
2. According to Fershtman and Pakes, what would prevent collusion in an oligopoly?
A) High barriers to entry
B) The likely exit of the industry by the smaller firm(s)
C) Strong rivalry between firms
D) The sale of durable goods.
3. According to Goettler and Gordon’s paper “Does AMD Spur Intel to Innovate More which firms were Intel’s competition for x86 microprocessors during the period from 1993-2004?
A) AMD
B) IBM
C) Intel
D) ARM Holdings
E) A and C
4. Which benefits below are not potential outcomes of first mover advantage?
A) Monopoly profits due to patents
B) Consumer Lock-In
C) Reduced production cost through greater “know-how”
D) Lower prices
5. Historically, what item(s) below did NOT contribute to the dramatically falling prices in the microprocessor industry in the 1980s and 1990s?
A) Cheaper inputs
B) Learning by doing
C) Continuous improvements in microprocessor power
D) All of the above
Answers
1. E – All of the above. You may not want to speak with Aunt Rhoda, but the fact that you both can do so on the same network (Apple’s FaceTime) makes it a positive direct network externality. And you may hate that video, but the person who posted it is enjoying the benefits of the positive network complementarity.
2. B– The likely exit of the industry by the smaller firm(s). The likely exit acts as an uncertain timeline. The closer the small firm is to exiting, the more certain the end date of the game, which causes both small and large firms to decide it is better to cheat than to collude. See Baye on Games With an Uncertain Final Period, pg. 370.
3. E – A and C. According to Goettler and Gordon’s theory, Intel’s primary competition was Intel because the microprocessors and the computers that used them were durable goods. According to their theory, in durable goods, even a monopoly firm has competition—it becomes its own competitor.
4. D – Lower prices. Baye cites potential monopoly profits, learning curve effects that drive down costs, and first-mover advantage due to consumer lock-in in Chapter 13.
5. A –Cheaper inputs. Historically, cheaper inputs may have lowered prices at some point, but they did not contribute to the dramatic drop in prices. For each new chip design, learning by doing improved manufacturing techniques, which resulted in an increased percentage of usable chips and a lower overall manufacturing cost. As new breakthroughs in upcoming chips were anticipated, consumers would either wait for the latest model before purchasing a PC, or would demand the older PCs with older microprocessors at a lower price.
Barriers to Entry
Barriers to entry were and are high in the x86 industry. It required significant investment in fabrication plants and equipment, and in the human capital necessary to effectively design and manufacture the chips successfully. Human capital and learning effects have been extremely important in the industry because of the difficulty in manufacturing a high percentage of usable chips. With each new design, cramming many thousands more transistors onto a single silicon chip, the manufacturing challenge would re-emerge. Rates of saleable chips could be very low during the initial manufacturing runs for a new design.
Gordon Moore is famous in the computer industry for Moore’s Law. In a 1965 article in Electronics Magazine, he famously predicted that the number of components on a single chip would double each year. In 1975, he updated his prediction to once every two years. He also predicted that the cost to manufacture the chips would drop significantly. His prediction was dubbed Moore’s Law in 1970, and the industry results adhered to Moore’s Law through the 1990s and into the 2000s. The implications meant that Intel could not bring in high margins at the lower end of their product line. They had to constantly innovate in order to offer higher margin products, as other vendors would compete on price for the less powerful chips. The rapid, constant innovation also increased the need for learning.
Studies have been performed on the importance of “learning by doing” in the semiconductor industry (Irwin and Klenow, 1994), (Hatch and Reichelstein, 1994), where considerable knowledge is built up and shared on how to resolve manufacturing problems with the sensitive chips. Hatch (1996), as cited in Salgado (1997) describes the challenge: “It is not uncommon for yields for new semiconductor processes to start as low as 10%. In such cases, the cost of scrapped output is initially very high but falls to low levels as yield rises to 90% or higher over time”. Also, according to Creating the Digital Future: The Secrets of Consistent Innovation at Intel by Albert Yu, wafer size and yields (the percentage of useable chips) dramatically impacts profitability in the semiconductor industry. (pp.85-86).
In addition to the investment in plant, equipment, and human resources, barriers to entry also included establishing relationships with the manufacturers who would buy the chips to use in their finished products.
December 4, 2011
Overview of the x86 Microprocessor Industry
Computers rely on what are called central processing units or CPU. The CPUs are the “brains” behind today’s desktop computers and laptops, performing the instructions provided by computer programs such as operating systems and applications. The CPUs built for desktop computers and laptops are called microprocessors because they are built onto a single silicon chip, or integrated circuit. Most of the microprocessors made for desktop computers and laptops are from a family named x86 after an early Intel product, the 8086. Intel and Advanced Micro Devices (AMD) are the two primary competitors within this industry, between them earning 95% of the x86 microprocessors revenue worldwide.
Early History of the Microprocessor Industry
Long before desktop computers were even considered possible, engineers were building large, bulky computers. Vacuum tubes were used in the earliest computers, as well as in other devices. The tubes were fragile, took up a lot of space, and had heat requirements that required significant power consumption. They also needed to be changed out periodically as they wore out. One of the critical steps enabling the move from these room-sized computers to the desktop, laptop, and mobile computers we use today was discovering how to utilize transistors to perform the work the vacuum tubes performed in the early days of the industry. Transistors were much smaller, less fragile, and did not require heat. Also, they did not have to be changed periodically. The transistor still had a limitation however—they had to be connected to the electrical circuits of the computer, which was difficult work requiring the soldering of thousands of tiny wires. To make the transistor practical, they needed to be “pre-connected”.
Two men are credited with the creation of the integrated circuit that solved the pre-connection problem. Jack Kilby at Texas Instruments holds the patent for the integrated circuit. His first model was built with germanium. Robert Noyce developed a circuit with silicon which resolved some of the remaining technical shortcomings of Kilby’s chip, thus removing the need for vacuum tubes.
The history of Intel and AMD is inextricably tied to a firm named Fairchild Semiconductor. Robert Noyce was general manager at Fairchild, and among his direct reports was an engineer named Gordon Moore, who headed up research and development. As Fairchild became more bureaucratic and less innovative, many of the original founders moved on to other companies. In 1967, National Semiconductor moved from Connecticut to Santa Clara and began hiring many of Fairchild’s talented engineers. Believing that Fairchild was irreparably harmed by the loss of human capital, Noyce and Moore founded Intel in 1968. Their intent was to build integrated memory circuits and compete with Fairchild Semiconductor and National Semiconductor.
AMD’s founders also came from Fairchild Semiconductor, roughly nine months after Intel’s founding. By the time they decided to move, investors felt that the field was already full of viable competitors in the semiconductor industry. While Noyce was able to secure funding for Intel in one afternoon, AMD’s funding was much more difficult to secure, and up until the last moment the founders were unsure they would raise enough funds to move ahead. In fact, Noyce invested in AMD, even though he knew they could turn into a competitor of Intel’s in the future. There is no information to imply that Noyce’s connection led to any collusive activity, or that he had any influence whatsoever on AMD’s plans.
In addition to Fairchild Semiconductor and National Semiconductor, other competitors in the industry in those early days were made up of a mix of existing and newly founded companies. They included Motorola, Mostek, founded in 1969 by ex-employees of Texas Instruments, and Zilog, founded in 1974 by two ex-Intel employees, one of whom had invented the microprocessor.
Intel began initially to build memory chips, but in 1970, the Nippon Calculating Machine Corporation contracted with Intel for a set of chips. The chips were for a desktop calculator named the Busicom. The chips needed to carry out basic arithmetical calculations. However, an Intel engineer, Ted Hoff, thought of a better plan. He wrote up a plan for a central processing unit (CPU), a memory chip, a read-only memory chip, and another device to handle the input and output. However, he didn’t build it as he was pulled off onto other projects. A new engineer from Fairchild Semiconductor, Federico Faggin, was assigned to the project instead, and designed and built the chips. Intel’s history would have been dramatically different had the Japanese firm stuck to the original terms of the contract. But instead, the calculator market had become much more competitive, and they came back demanding a price cut. Robert Noyce, after speaking with the engineers on the project, knew Intel had created something they would want to keep, and arranged to refund some of the money to the client for the rights to the design of the chips, including the microprocessor. Intel launched its first microprocessor, the 4004, in November, 1971, gaining the first mover advantage.
Network Effects
The microprocessor industry makes an excellent case study for what economists like to refer to as positive indirect network externalities or network complementarities. According to Katz and Shapiro (1994), consumers benefit when they utilize a network such as a telecommunications network, and others are also using the technology. Translated, it means being the sole user of a telephone wouldn’t provide much value. The largest part of the value is in the number of users on the network. These are typically referred to as network effects, or direct network externalities as the users are on the same physical network. Katz and Shapiro also bring up a similar effect with durable goods, which they describe as the “hardware/software paradigm”. The consumer must not only determine which hardware they prefer, but also must determine how much software may be available for it. They liken the paradigm to computers, and among other examples, credit cards networks, in which the card is the hardware, and the merchant acceptance acts as the software. With computers, the benefit is a network complementarity, or indirect network externality because of the size and availability of complementary products and/or services.
These positive network complementarities make a product more attractive to a consumer. A classic example is the attractiveness of the iPhone. When is it considered as just a phone, it is not nearly as attractive as when it is considered as a device with over 500,000 available applications, including games. That positive complementarity can have a dark side, however, in that it can tie consumers to a particular good or service because of the benefits of the network, in a phenomenon known as “consumer lock-in”. The partnership between Microsoft, with its Windows operating system, and Intel as the de facto standard for personal computer microprocessors, is a type of network that locked consumers in. Called the “Wintel” platform, consumers remained with these types of computers because of the greater number of applications, the ability to share documents and other files with other computer users, and the ease of finding qualified support.
Like many products, microprocessors are not valuable in and of themselves; for them to provide value, they must be an input to a finished product. In the early 1970s Intel continued to innovate in microprocessors, but considered the memory chip their main line of business. Microprocessors weren’t bringing in that much money, as there weren’t that many products in which they could be used. Intel turned down opportunities to build the computers.Gordon Moore was not interested, as he couldn’t see the reason why someone would want one. He couldn’t envision the positive network complementarities.
Intel might not have seen the potential, but by 1977, external firms began to offer smaller computers using the Intel chips. MITS released the Altair computer kit, using Intel’s 8080 chip. Radio Shack also had a kit. Steve Wozniak and Steve Jobs were also innovating, with a portable computer that could actually be picked up by a person. A program called Visicalc was written for the Apple II. It was the first commercial spreadsheet, and was designed for use by accountants and bookkeepers. It helped to answer Gordon Moore’s question of why someone would want a portable computer. It demonstrated a use for these computers in the business world, and demonstrated the potential market for creating software programs for the devices. Others began writing programs, including word processing programs which freed up secretaries from the tyranny of the typewriter. Network complementarities began to take off for the portable computer.
An almost accidental sales call on IBM secured Intel an order to provide the 8088 microprocessor to IBM for its launch of a personal computer. IBM insisted on a second source, and Intel turned again to AMD. A contract was negotiated in 1982 that would last for 10 years, in which AMD would act as a second source for Intel but also for Intel to second source for AMD as well. Ultimately they ended up in a court battle over the contract.
Motorola had supplied its 68000 processor for the Apple computer, but IBM’s PC, using Microsoft’s MS-DOS as the operating system and Intel’s microprocessor became the de facto standard for much of the personal computer industry by allowing “clones” to be made by other manufacturers. Because of the contract with IBM, Intel’s chips became the standard microprocessor for the clone industry. The 8088 was based upon the 8086 chip, hence the x86 name for the line of microprocessors. Intel swept AMD along in its wake as the second source.
Intel Potentially Restrains Competition
In “A Dynamic Oligopoly With Collusion and Price Wars”, Fershtman and Pakes (2000) posit an industry with at most a handful of competitors and use game theory to determine the likelihood of collusion. If there are two firms, their conclusion was that “collusion breaks down when one of the two is near an exit state, but again just how near depends on where the larger competitor is.” (pg. 221). They point out two reasons for collusion being hard to sustain in such an industry. In large part, it has to do with the potential for the smaller firms to exit the industry in the foreseeable future. In theory, the near term potential for exiting takes the game from an infinitely repeatable one to one of a shorter, potentially definable period, as described by Baye in our text on page 363. The closer a small firm is to exiting, the more definable the period, and the greater the incentive for the small firm to cheat. However, there is also an incentive on the side of the larger firm, which is not covered in the game theory chapter in Baye. According to Fershtman and Pakes, there is incentive to predatory behavior on the part of the larger firm. “The large firm realizes that the future will be better if it can ‘force’ the small firm(s) to exit and then monopolize the market until the next entrant arrives. Since the smaller firm is more likely to exit if there is no collusion, the larger firm has an incentive to deviate from the collusion pricing in order to increase the likelihood of the smaller firm exiting.” (pp. 221-222).
According to Fershtman and Pakes, the potential for collusion is more likely when there are two firms who can be described as strong incumbents. While Intel can be characterized as a strong incumbent, AMD cannot. AMD more closely matches the description of a firm likely to exit the industry, especially if Intel could find a way to starve it of revenue. Intel’s predatory behavior didn’t result in lower prices for newer microprocessors in order to drive AMD from the industry. Instead, as Intel attained a significant portion of the market and built its reputation, it appears that it took steps designed to reduce AMD’s revenue to an untenable level.
The relationship between Intel and AMD began to deteriorate from the point of their 1982 licensing agreement, which was mentioned in a previous posting. Jerry Sanders, the CEO of AMD, could not resist describing the great deal AMD had achieved to the press, and characterizing Intel as having gotten the worst end of the bargain. Some say Andy Grove took offense at his language. Whatever the reason, Intel began a campaign to try to undercut the agreement and force its early termination. The agreement called for the possibility of Intel accepting product designs from AMD and opting to act as a second source for them. Second sourcing was a standard practice in the semiconductor industry. It was followed by firms who would not risk a high profile project by relying on only one supplier of a critical input. Second sourcing required the leading firm to share its design with the second source, who would then manufacture a similar part, often for less money. Through second sourcing, there was a significant knowledge spillover in the industry.
If Intel accepted any of AMD’s product designs and agreed to second source them, it would reduce or remove the need for AMD to pay royalties for second sourcing Intel’s product designs. AMD wanted to second source the 80286 chips. AMD proposed two product designs Intel could second source. According to Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company, an internal memo within Intel has since come to light details Intel’s strategy at that point:
"1/ Assure AMD they are our primary source through regular management contact and formal meetings
2/ Take no more AMD products under the current agreement"
Intel had determined that it had enough of a hold on the market to sell the upcoming 80386 chip without a second source. AMD would no longer be able to undercut Intel’s margins. Intel’s management believed by not second sourcing the new chip, it could keep the price of the 80386 higher for a longer period of time. However, during this period, it needed to convince AMD it would give them the 80386 design for second sourcing while still using AMD as a second source for the 80286 chips. Another memo outlining Intel’s policy states, “Maintain a second-source, business as usual posture in the market place. . . . We do not want them [AMD] to go to Hitachi or NEC, and should not stimulate them to do so.”
As the situation dragged out, AMD determined to invoke the arbitration clause of their contract. Originally they thought it would be a short process, estimating roughly 12 weeks. The arbitration, which ultimately took three years, began a series of bitter legal battles which weren’t settled until 2009.
Halting second sourcing meant AMD was forced to then create its own designs, and for the most part, it delivered competitive, value-priced microprocessors. In fact, for a period of time it became a stronger, more credible competitor. Intel was accused of taking additional steps to prevent AMD’s successful competition at about that time: they began paying rebates to PC manufacturers such as Dell and HP. The rebates required the manufacturers to buy less of AMD’s products. According to the European Union’s competition commissioner, during the period of October 2002 – December 2007, the rebates helped Intel maintain a share of at least 70 percent of microprocessor sales. In May, 2009 the European Union fined Intel $1.45 billion dollars and ordered them to stop paying the rebates.
Aizcorbe, Ana. (September 1, 2005). Why Are Semiconductor Price Indexes Falling So Fast? Industry Estimates and Implications for Productivity Measurement. Bureau of Economic Analysis.
Baye, Michael R. Managerial Economics and Business Strategy. New York, NY: McGraw-Hill/Irwin.
Fershtman, C., Pakes, A. (2000). A Dynamic Oligopoly with Collusion and Price Wars. The RAND Journal of Economics, volume 13, No. 2 (Summer, 2000), pp. 207-236.
Jackson, Tim. (1997). Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company. New York, NY: the Penguin Group.
Kanter, James. Europe Fines Intel $1.45 Billion in Antitrust Case. (2009, May 13).
Retrieved from http://www.nytimes.com/2009/05/14/business/global/14compete.html?pagewanted=all,
Katz, Michael. L. Shapiro, Carl. Systems Competition and Network Effects . Journal of Economic Perspectives. Volume 8, Number 2 (Spring, 1994), pp. 93-115.
November 20, 2011
Since my last posting I’ve scanned a variety of scholarly articles on oligopolies, but the most pertinent one I’ve found is titled Does AMD Spur Intel to Innovate More? It is a working paper dated June 20, 2011, by Ronald Goettler of the University of Chicago and Brett Gordon of Columbia University. It has been conditionally accepted at the Journal of Political Economy.
In the Visa and Mastercard case, the question was whether cooperation between the two stifled competition. In the case of AMD and Intel, since Intel retains such a large share of the market, the question becomes whether innovation would be greater or consumer surplus would be higher if Intel were a monopoly. Does the existence of AMD, with such a small share of the market, cause Intel to innovate more or less?
While a large part of the mathematics in the paper was completely over my head, the authors made a series of points which can help us all better understand how duopolies and monopolies work in certain environments. I will summarize the main points here.
Based upon the sample data spanning from 1993 to 2004, the authors concluded that consumer surplus was 4.2% higher with AMD in the market, but innovation would have been 4.2% higher had AMD not been a competitor.
Does AMD spur Intel to innovate more? June, 2011, conditionally accepted at Journal of Political Economy, accessed November 4, 2011. http://www.columbia.edu/~brg2114/files/dynduo.pdf
November 5, 2011
Up until now I have been exploring questions surrounding the duopoly of Mastercard and Visa. From this point on, I'll compare and contrast their situation with that of another duopoly, Intel and AMD.
Here are descriptions of both Intel and AMD from the New York Times:
The Intel Corporation (INTC) is the leading maker of semiconductor chips and is known for its quest to make computer chips ever smaller, faster and cheaper. It also makes circuit boards and other semiconductor products, which are the building blocks of computers, servers, consumer electronics and communications devices. Intel supplies about 80 percent of the PC microprocessor chips worldwide.
In an effort to expand beyond its core chip-making business, Intel said in August 2010 that it had agreed to buy McAfee, the computer antivirus software maker, for about $7.7 billion in cash. With its McAfee purchase, Intel is gaining an entrance into the security tech sector, one that is expected to continue growing quickly.
Intel is also pushing hard in the highly competitive race to put chips in smartphones and televisions and other consumer devices that are gaining computer power. In October 2010 it reported a flurry of sales of the company’s chips in TV and set-top boxes, figures that seemed to bode well for Intel’s consumer aspirations. The company sold one million chips aimed at smart TVs last quarter, and its only prominent loss was getting shut out of Apple TV.
Advanced Micro Devices, Inc. (AMD) is a global semiconductor company. The Company primarily offers x86 microprocessors, for the commercial and consumer markets, embedded microprocessors for commercial, commercial client and consumer markets and chipsets for desktop and notebook personal computers (PCs), professional workstations and servers. It also offers graphics, video and multimedia products for desktop and notebook computers, including home media PCs, professional workstations and servers and technology for game consoles. As of December 25, 2010, the Company had two segments: Computing Solutions and Graphics. Computing Solutions segment includes microprocessors, chipsets and embedded processors and related revenue. The Graphics segment includes graphics, video and multimedia products and related revenue, as well as revenue received in connection with the development and sale of game console systems that incorporate its graphics technology.
Unlike Visa and Mastercard, who ended up in court because they appeared to be collusive, intending to prevent other credit card companies from competing with them, there is no doubt that in this duopoly, Intel and AMD are bitter rivals. Intel has faced a number of legal battles over anti-competitive behavior against AMD. In April, 2009, Intel paid AMD 1.25 billion dollars to settle allegations of antitrust violations. Initially, per an article in Business Week, AMD claimed Intel was behaving in an anti-competitive manner, stating in court documents that it "used multibillion-dollar payments to coerce PC makers such as Dell (DELL) and Hewlett-Packard (HPQ) into using Intel products to the exclusion of AMD's chips."
At the same time, Intel had a long-standing patent dispute with AMD: it had an agreement with a subsidiary to allow for patent-sharing. AMD moved to spin off the subsidiary as a separate company, GlobalFoundries. When GlobalFoundries tried to take the patent agreement with it, Intel cried foul, stating that the agreement did not allow AMD to extend it to a third-party organization. Without the patents, AMD and GlobalFoundries would not be able to manufacture chips of any market value.
The agreement called for quarterly meetings between the companies and methods to mediate disagreements in the future. It also enabled the firms to cross-license each other’s patents for five years. The bigger question is whether the agreement will change the relationship between the two companies and the nature of the duopoly. AMD is known for lower priced chips, and up until the past few years, provided good "value for money". However, they have often lagged behind Intel in delivering transformative technology.
The settlement with AMD helped take some of the pressure off Intel, who had been ordered by EU regulators in May of 2009 to pay a fine of 1.45 billion for antitrust violations. As part of the determination, according to the New York Times, Neelie Kroes, the competition commissioner for the EU, "ordered Intel to stop offering rebates that were conditioned on buying less of a rival’s product, or not buying them at all..." In July of 2009, Intel appealed the verdict, but the court takes approximately two years to reach a verdict.
In the meantime, the FTC was also threatening to investigate Intel, and there was speculation that the agreement with AMD was intended to ward off a lawsuit. However, Intel's anti-competitive woes were not yet over; the FTC moved forward with its case and in December, 2010 Intel settled with them.
http://topics.nytimes.com/top/news/business/companies/intel_corporation/index.html
http://topics.nytimes.com/top/news/business/companies/advanced_micro_devices_inc/index.html?inline=nyt-org
http://www.businessweek.com/technology/content/nov2009/tc20091115_692400.htm
http://www.nytimes.com/2009/05/14/business/global/14compete.html?pagewanted=all
http://www.ftc.gov/opa/2010/08/intel.shtm
October 29, 2011
This week I'll examine whether Mastercard and Visa are poised to lose their overwhelming control of the payment industry due to current technology changes in the market.
On April 13, 2011, Senator Dick Durbin of Illinois wrote an open letter to Jamie Dimon, the CEO of JP Morgan Chase, regarding the profits the banks, Visa, and Mastercard were enjoying through their interchange fees.
"For years, card-issuing banks like Chase have agreed to let the Visa and MasterCard duopoly fix the interchange fee rates that banks receive from merchants each time a debit card is swiped. The banks get the fees but they do not set the fees...
Visa and MasterCard have incentive to constantly increase interchange rates and there is no countervailing market force to temper these fee increases. Visa and MasterCard want as many of their debit cards to be swiped as possible because they are paid a network fee by merchants each time a card is swiped. By raising interchange rates, Visa and MasterCard can entice banks to issue more of their cards. Because Visa and MasterCard have enormous market power and control around 80 percent of the debit cards in consumers' wallets, merchants cannot realistically say no to accepting Visa and MasterCard and have no leverage to negotiate fee rates with them."
Of interest is the statement that there is no countervailing market force to temper the interchange fee increases. However, the market may have already produced a new, competitive means of payment to be carried on an alternative network: smartphones. Nicknamed in some articles as "m-commerce", mobile phones are a big growth area for digital payments. Multiple organizations are in a struggle to see who will win in this new digital age. As profiled in a Bloomberg article, AT&T, Verizon to Target Visa, MasterCard With Smartphones, dated August 2, 2010, the carriers have banded together to take on Visa and Mastercard. The payments would be carried by the Discover network.
Analysts of the payment card industry felt that merchants would be willing to embrace the new payment method as they have been battling with Visa and Mastercard over fees for many years. This type of payment system is already effective in several countries outside the U.S., including Japan and the U.K. As cited in the article, an industry consulting firm, Mercatus, states "More than half of U.S. consumers, and almost 80 percent of those between the ages of 18 and 34, will use mobile financial services within five years". If Visa and Mastercard do not find a way to embrace the newer technology, they stand to lose their positions of power in the payment industry.
To ensure they'd don't lose control, Visa and Mastercard have been developing their own plans for mobile phone payments. In contrast to earlier actions cited in the case in Baye's Managerial Economics and Business Strategy, this time they don't appear to be perfectly aligned. Visa has multiple initiatives underway, including joining the network AT&T and Verizon started, now named the Isis Network, as well as developing a concept called the "digital wallet". Described as a "wallet in the cloud", it will be open to competing card vendors such as American Express. Mastercard has in turn developed a partnership with Google and acquired DataCash, which handles online payments.
The rush to develop market dominance in digital payments doesn't stop with the mobile carriers, Discover, Visa, and Mastercard. American Express is developing partnerships with Facebook and others to promote its own method of supporting digital payments. PayPal is also introducing electronic payment capability via mobile phones. Their offering doesn't require a special near field communications (NFC) chip in the phone, or a special NFC-enabled reader in the store. They are betting that merchants will drag their feet in re-equipping their stores, and therefore a non-NFC method of payment will gain the largest market share.
While the switch to mobile phone payments won't happen overnight, it seems to be the wave of the future. Only time will tell if the Visa and Mastercard duopoly will be broken by the disruptive changes coming with m-commerce. Some industry analysts feel the mobile carriers can mount an effective challenge, given their expertise in handling large scale billing and payments. In addition to the firms who will process the payments, there is uncertainty surrounding the technology to enable mobile phone payment, with multiple alternatives being backed by different players in the market. The winner will also be determined by the mobile phone vendors, who must provide the phone hardware, and the massive network of merchants who must decide which technology they'll adopt. It's too soon to tell who will win. Only one thing seems certain: at some point, carrying credit cards will seem as quaint as carrying checks or cash feels now.
http://consumerist.com/2011/04/senator-durbin-to-chase-ceo-youre-already-gouging-the-consumer-so-stop-complaining.html, accessed October 29, 2011.
http://www.bloomberg.com/news/2010-08-02/at-t-verizon-said-to-target-visa-mastercard-with-smartphones.html, accessed October 29, 2011.
http://www.computerworld.com/s/article/9216425/Isis_says_carrier_backed_mobile_payments_accelerated_not_dialing_back_?taxonomyId=77&pageNumber=2, accessed October 29, 2011
http://techcrunch.com/2011/08/07/how-visa-plans-to-dominate-mobile-payments-create-the-digital-wallet-and-more/, accessed October 29, 2011
Oct.14, 2011
Last week I examined whether the dual management structure of Visa and Mastercard inhibited competition and therefore innovation, using the newer chip technology as the representative example of innovation in the industry. This week I'll focus on how Mastercard and Visa's duality may affect the evaluation of performance by the member banks, and therefore affect overall market share and innovation, as discussed in the paper, On Nonexclusive Membership in Competing Joint Ventures, by Jerry A. Hausman, Gregory k. Leonard, and Jean Tirole, and published in spring, 2003.
As mentioned below,Visa and MasterCard have roughly 75% of the dollar volume of transactions, and 86% of the general purpose cards issued. In the paper, the authors note that "over 6000 entities issue credit cards". They state that the HHI of the issuers is quite low. However, the issuers must work within the services provided by Visa and Mastercard, and duality gives them an unusual level of insight into the performance for both. Objective measurements, such as percentage of the market, are available to everyone, including non-members. But detailed comparisons in terms of overall system performance, management of the associations, and incentives to provide one card vs. another are available only to member banks. According to Hausman et. al, that evaluation can affect which cards they offer their customers, thereby driving and possibly changing market share. According to their research, the sensitivity of the members would differ if their membership was exclusive vs. non-exclusive, as it is today. If membership were exclusive, meaning a member bank could only issue either Visa or Mastercard, the banks would be more sensitive to large quality issues, and there would be an incentive for member banks to switch all of their customers from one card to another if quality were poor enough. With the current system of duality, member banks are more sensitive to small changes in quality, as they find it less expensive to switch customers incrementally. This is most often achieved by convincing a new customer to take one card over the other.
What is the end result of this sensitivity? According to the authors, "we conclude that duality facilitates benchmarking for incremental innovations and inhibits benchmarking for large ones." While they find that the associations have opted for the economically efficient level of investment in innovation, they seem to feel that the type of innovation encouraged by duality is the slower, more incremental type rather than the larger, more disruptive changes encouraged by exclusive, and more competitive, arrangements.
Hausman, Jerry A., Leonard, Gregory K., and Tirole, Jean. On Nonexclusive Membership in Competing Joint Ventures, The RAND Journal of Economics, Vol.34, No. 1 (Spring, 2003), pp. 43-62.
Oct. 3, 2011
Visa and MasterCard are the two largest providers in the market for general purpose credit card network products and services. Together, Visa and MasterCard account for 75% of the dollar volume of transactions, and account for 86% of the number of general purpose cards issued. American Express, Discover/Novus, and Diners Club are the significant competitors in the general purpose card market. Both Visa and MasterCard are joint ventures (associations) that are owned and operated by the member banks that issue cards and provide card acceptance services. The member banks have an interest in both Visa and MasterCard (known as duality) and have representatives serving on the board of directors or important committees of both Visa and MasterCard.
Since the member banks of Visa and MasterCard have a significant interest in both Visa and MasterCard, they have little incentive or desire to compete directly against one another. The member banks have refused efforts by MasterCard management to develop its brand through marketing campaigns. The banks felt that this action would hurt the Visa brand, which the banks have a stake in. They have also not supported new product development, unless the new development was available to both Visa and MasterCard.
In addition to avoiding competition between brands, the member banks have prohibited its banks from issuing competing credit cards, such as American Express and Discover/Novus. They have also worked jointly to discourage merchant acceptance of competing cards and access to ATMs for cash advances.
Related Textbook Material and Suggested Teaching Opportunities
The primary parties:
• Visa and MasterCard are joint ventures owned by the member banks that issue cards and provide card processing services.
• Visa and MasterCard are separate entities with separate management and staff. However, the member banks provide representatives to the board of directors and important management committees. Therefore, both Visa and MasterCard have representatives of the same member banks setting their corporate policies and strategies.
The Competitors:
• American Express, Discover/Novus, and Japan Credit Bureau (JCB) are the primary competitors in the general purpose card market. Many retail companies issue their own cards for use at a single retail store, but these services do not compete directly with Visa and MasterCard.
Barriers to Entry:
• Barriers to entry are very high in the credit card market. In order to gather a large network of merchants, credit cards must establish a large consumer base. However, it is very difficult to gather a consumer base until there iswidespread merchant acceptance. This Catch-22 makes it extremelydifficult for new entry into the market.
The Primary Issues:
Dual ownership of Visa and Mastercard is believed to lessen competition between them.
Restraint of competitors prevents them from competing in the market
• The member banks that own Visa and MasterCard do not want the management of one of the associations to compete in a way that would put the other at a disadvantage. Neither Visa nor MasterCard has made efforts to distinguish its brand from the other; because of the feeling this could harm the other brand.
• Competition is also stifled in the development of new technologies. The member banks have resisted the development of new products unless the technology would be available to both Visa and MasterCard.
How have Visa and MasterCard worked to restrain its competitors from competing in the market?
• Visa adopted a rule that prohibited its member banks from issuing credit cards that it deems to be competitive, including Discover/Novus and American Express. This rule did not apply to banks issuing MasterCard credit cards. MasterCard adopted a similar rule to prevent its member banks from contracting with American Express.
• In the mid-1980's, Visa and MasterCard passed rules that prohibited American Express and Discover/Novus from processing credit cardtransactions through their terminals. This exclusion made it difficult for American Express and Discover/Novus to gain acceptance from merchants, since the merchants would have to operate separate terminals. Eventually they loosened these restrictions due to merchant demands, however,processing fees were higher for American Express than for other third party processors.
• Visa and MasterCard have also tried to exclude other competitors from the automated teller machine (ATM) market. Visa and MasterCard ownworldwide ATM networks, Plus and Cirrus respectively. Visa andMasterCard allow users to use both networks, but exclude American Express and Discover/Novus from using their ATM networks. As a result, American Express and Discover/Novus have had to negotiate anddevelop ATM network agreements with smaller fragmented networks.
What evidence is there that Visa and MasterCard would compete more aggressively without duality?
• Exclusionary rules do not apply outside the United States, so the banks have credit card companies have more incentive to compete. Several foreign banks have relationships with American Express and Visa orMasterCard. Therefore, MasterCard and Visa have an incentive to provideinnovative products and services, to ensure that the banks continue to promote their products.
I have downloaded a number of articles on network effects and collusion to begin my research. This week I'll briefly explore the basis for the claim that Mastercard and Visa's relationship restrained competitive behavior, which in turn limited the adoption of new technologies and services which could have benefited the consumer.
The smart card serves as a classic technology example highlighting how consumers can be affected by stifled competition. Per the case in Baye, "a smart card differs from the cards in the U.S.in that it can store information on an integrated circuit, instead of, or in addition to, a magnetic stripe. Integrated circuits are capable of storing significantly more information than magnetic stripes." They are often used in banking, healthcare, and government.
According to the case, in the 1980s, MasterCard developed and extensively tested smart cards.They determined in 1987 that it would give them a competitive advantage over Visa. However, bank representatives on MasterCard's executive committee wouldn't move ahead unless Visa would also do so, and Visa said no. The important point to note is that Visa said no in the U.S. Smart cards are the preferred choice in Europe, including cards issued by Visa and Mastercard. The reason, according to the case, is that outside the U.S., Visa and Mastercard do compete against one another.
A September 13, 2011 posting on Credit Newsline at http://creditnewsline.com/news/visa-and-mastercard-pushing-u.s.-towards-smart-cards/, noted “cards that use chips provide much more security compared to magnetic stripe cards, according to CBSNews.com. As a result of the use of magnetic stripe cards, the United States has higher fraud rates for counterfeit cards, which costs banks, merchants and consumers billions every year.” In addition to experiencing greater security problems at home, Americans are also inconvenienced as they travel outside the U.S., because in certain countries, it is becoming more and more difficult to use our magnetic stripe cards. Newer machines cannot read the information on our cards. Our technology is becoming obsolete.
In "United States v. Visa USA, Inc., 1999-2 Trade Cas. (CCH) ¶ 72,584, the Department of Justice alleged that the Visa and MasterCard corporations agreed upon a number of practices, including an agreement not to engage in certain types of product development. The firms allegedly agreed not to develop and market, or delayed development and marketing of smart cards, commercial cards, and methods for making Internet transactions more secure." (Robert H. Lande & Howard P. Marvel, The Three Types of Collusion: Fixing Prices, Rivals, and Rule, Wisconsin Law Review, 2000:941). In 2001, the District Court determined that the government had not provided enough evidence to prove its case that Visa and Mastercard's dual governance unreasonably restrained trade. However, the judge determined their rules preventing their member banks from issuing American Express or Discover cards did unreasonably restrain trade under Section 1 of the Sherman Act.
It may be difficult for the government to make the case that Visa and Mastercard's dual management structure has inhibited competition and therefore innovation. Fortunately for the American consumer, the technology finally will be adopted: Visa has reconsidered and according to Credit Newsline, is now requiring that processors in the United States be able to support merchant acceptance of chip transactions by April 2013. Needless to say, Mastercard also recently announced it would roll out the technology.