Conclusions & recommendations for intel corporation case | Operations Management homework help

Conclusions & Recommendations For Intel Corporation Case umyzeed2 Case 11 Intel Corporation: 1968–2013 Charles W.L. Hill School of Business, University of Washington Seattle, WA 981095, June 2013

inTroducTion

In 2012 Intel was the leading manufacturer of micropro- cessors for personal computers in the world, a position that it had held onto for more than two decades. Over 80% of all personal computers sold in 2012 used Intel microprocessors. The company reported revenues of $53 billion and net pro ts of $11 billion. Meanwhile, Intel’s only viable competitor, AMD, which in the early 2000s had been gaining share from Intel, lost $1.2 billion on sales of $5.4 billion. Despite its historic dominance, the future looked uncertain for Intel. The rise of mobile devices had led to a strong substitution effect, with sales of PCs fall- ing as consumers switched to smart phones and tablets for many of their computing needs. In the rst quarter of 2013, global PC sales fell 14% on a year over year basis according to the research rm IDC. This was the worst yearly decline since IDC started tracking PC sales in 1994, and the fth quarter in a row that PC sales had fallen. At the same time, sales of smart phones and tab- lets were booming. IDC predicted that sales of tablets would grow almost 60% in 2013, and that tablet ship- ments would exceed those of portable PCs.1 The crux of the problem for Intel is that most tablets and smart phones used microprocessors that are based on technology licensed from ARM Holdings PLC, a British company whose chip designs are valued for their low power consumption, which extends battery life. While Intel has a line of chips aimed at mobile devices—the Atom chips—microprocessors incorporating ARM’s technology were found on 95% of smart phones in 2012 and over 30% of all mobile computing devices, a cate- gory that includes tablets and PC notebooks.2 Moreover, in 2012 Microsoft issued a version of its Windows 8 operating system that ran on ARM chips, rather than Intel chips, creating a potential threat to Intel’s core PC business. The FoundaTion oF inTel Two executives from Fairchild Semiconductor, Robert Noyce and Gordon Moore, founded Intel in 1968. Fairchild Semiconductor was one of the leading semi- conductor companies in the world and a key enterprise in an area south of San Francisco that would come to be known as Silicon Valley. Noyce and Moore were no ordinary executives. They had been among the eight founders of Fairchild Semiconductor. Noyce was gen- eral manager at the company, while Moore was head of R&D. Three years previously, Moore had articu- lated what came to be known as Moore’s Law. He had observed that since 1958, due to process improvements the industry had doubled the number of transistors that could be put on a chip every year (in 1975 he altered this to doubling every two years). Fairchild Semiconductor had been established in 1957 with funding from Sherman Fairchild, who had backed the founders on the understanding that Fairchild Semiconductor would be a subsidiary of his Fairchild Camera and Instrument Corporation on New York. By 1968 Noyce and Moore were chaf ng at the bit under management practices imposed from New York, and both decided it was time to strike out on their own. Such were the reputations of Noyce and Moore that they were able to raise $2.3 million to fund the new venture “in an afternoon on the basis of a couple of sheets of paper

C-173

84487_case-11_ptg01_hr_C173-C185.indd 173 22/10/13

4:55 PM

C-174

Case 11 Intel Corporation: 1968–2013 containing one of the sketchiest business plans ever nanced”.3 When business reporters got wind of the new ven- ture, they asked Noyce and Moore what they were in- tending to do, only to be greeted by vague replies. The two executives, however, knew exactly what they were going to do—manufacture silicon memory chips—they just didn’t want potential competitors to know that. At the time, sales of mainframe computers were expanding. While these machines used integrated circuits to perform logic calculations, programs and data were stored on magnetic devices. Although inexpensive to produce, it was relatively slow to access information on a magnetic device. Noyce and Moore knew that if they could build a silicon based integrated circuit that could function as a memory device, they could speed up computers, making them more powerful, which would expand their applica- tions and allow them to shrink in size. These memory chips were knows as dynamic ran- dom access memories (DRAMs). While much of the theoretical work required to design an integrated cir- cuit that could function as a memory device had already been done, manufacturing DRAMs cost ef ciently had so far proved impossible. At the same time, some key research on manufacturing was being done at Fairchild. This research included a technique known as metal oxide on silicon, or MOS. Noyce and Moore wanted to mass- produce DRAMs, and after looking at other possible alternatives, they concluded that commercializing the MOS research was the way to do it. This prompted some cynics to note that Intel was established to steal the MOS process from Fairchild. andy Grove To help them, Noyce and Moore hired a number of re- searchers away from Fairchild, including, most notably, a young Hungarian Jewish émigré called Andy Grove. At Fairchild, Grove had reported directly to Moore. At Intel he became the director of operations with responsibility for getting products designed on time and built on cost. Through the force of his own personality, Grove would transmute this position into control over just about ev- erything Intel did, making him effectively the equal of Noyce and Moore, long before he was elevated to the CEO position in 1987. Grove was an interesting character. Born in 1936, he went into hiding when the Germans invaded Hungary dur- ing World War II and managed to escape the Holocaust. After WWII, the tyranny of the Germans was replaced by the tyranny of the Soviets as Hungary became a satellite state of the Soviet Union. In 1956, after the failure of an uprising against the Soviet puppet government, Grove es- caped across the border to Austria, and made his way to the United States. He put himself through college in New York by waiting on tables, and then went to UC Berkley for graduate work, where he received a Ph.D. in chemical engineering in 1963. His next stop was Fairchild, where he worked until Moore recruited him away in 1968. Over the next three decades, Grove would stamp his personality and management style on Intel. Regarded by many as one of the most effective managers of the late twentieth century, Grove was a very demanding and according to some, autocratic leader who set high ex- pectations for everyone, including himself. He was de- tail orientated, pushed hard to measure everything, and was constantly looking for ways to drive down costs and speed up development processes. He was known for a confrontational “in your face” management style, and would frequently intimidate employees, shouting at those who failed to meet his expectations. Grove him- self, who seemed to enjoy a good ght, characterized this behavior as “constructive confrontation”. He would push people to their limits to get things done. As he once noted, “there is a growth rate at which everybody fails, and the whole situation results in chaos. I feel it is my most important function. . . . to identify the maximum growth rate at which this wholesale failure begins”.4 Grove demanded discipline, insisting for example, that everybody be at their desks at 8 a.m., even if they had worked long into the night. He instituted a “late list”, requiring that people who arrived after 8 a.m. sign in. If people arrived late for meetings, he would not let them attend. Every year he sent around a memo to employees reminding them that Christmas Eve was not a holiday, and that they were expected to work a full day. Known as the “Scrooge memo”, many would be returned with nasty comments scrawled over them. May you eat yellow snow, said one. A very neat man, if people’s desks were messy, Grove would publically criticize them. Accord- ing to one observer, “Andy Grove had an approach to discipline and control that made you wonder how much he had been unwittingly in uenced by the totalitarian re- gime he had been so keen to escape”.5 Grove controlled managers through a regular budget- ing process that required them to make detailed revenue and cost projections. He also insisted that all managers establish medium term objectives, and a set of key re- sults by which success or failure would be measured. 84487_case-11_ptg01_hr_C173-C185.indd 174 22/10/13

4:55 PM

He instituted regular one-on-one meetings where perfor- mance was reviewed against objectives, holding manag- ers accountable for shortfalls. He also required monthly management reviews where managers from different parts of the company would meet to hear a presentation of its current strengths, weaknesses, opportunities and threats. The goal was to get managers to step back and look at the bigger picture, and to encourage them to help each other solve problems. Grove would also practice management by walking around, inspecting facilities and of ces, demanding that they be clean, something that earned him the nickname “Mr. Clean”. He pushed the human resource department to institute a standard system of ranking and rating that had four performance categories; “superior”, “exceeds expectations”, “meets expectations”, or “does not meet expectations”. People were compared against others of their rank. Pay raises and later, stock option awards were based on these rankings. Despite his autocratic style, Grove was grudgingly admired within the company. He was a brilliant prob- lem solver, a man with tremendous control of facts and details, someone who was determined to master the challenging technical projects that Intel was working on. Moreover, while he drove everyone hard, he drove himself harder still, thereby earning the respect of many employees. The MeMory chip coMpany Making a DRAM using MOS methods proved to be extremely challenging. One major problem—small partials of dust would contaminate the circuits during manufacturing, making them useless. So Intel had to de- velop “clean rooms” for keeping dust out of the process. Another was how to etch circuit lines on silicon wafers, without having the etched lines fracture and break as the wafer was heated and cooled repeatedly during the manufacturing process. The solution to this problem, identi ed by Moore, was to “dope” the metal oxide with impurities, making it less brittle. Intel subsequently went to some lengths to keep this aspect of the manu- facturing process secret from competitors for as long as possible. Intel, of course, was not alone in the race to develop a commercial process for manufacturing DRAMs. Among the potential competitors was another semiconductor company started in 1969 by Jerry Sanders, a former mar- keting director at Fairchild. Sanders started his company with the help several other Fairchild employees who had not been recruited by Intel. Called Advanced Micro devices, or AMD, the company found it tough to raise capital until it received an investment from non other than Robert Noyce, who saw something he liked in the amboyant Sanders. Driven by constant pressure from Andy Grove, whose “in your face” management style was bearing fruit, albeit at some human cost, by October 1970 In- tel succeeded in producing a DRAM chip, named the 1103, in relatively high yields (which implied that rela- tively few chips had to be discarded). The 1103 could store 1,024 bits of information (zeros or ones), which was 4 times as much as the highest capacity semicon- ductor memory device currently available. Since the xed costs required to establish a manufacturing facility were very high, the key to making money on the 1103 was high yields and high volume. If Intel could achieve both, unit costs would fall enabling Intel to make a lot of pro t at low price points. In turn, low prices implied that DRAMs would start to gain wide adoption among computer manufacturers. The 1103 put Intel rmly on the map. The chip soon became the memory technology of choice for computer makers, and by the end of 1971, 14 out of the world’s 18 leading mainframe computer makers were using the 1103. However, Intel did not have the market entirely to itself. Computer makers did not want to become depen- dent upon a single source of supply for critical compo- nents. To avoid this, most computer makers mandated that components had to be at least duel sourced, and for Intel, this meant that if it wanted business, it had to license its technology to other companies. Intel rst li- censed the rights to produce the 1103 to a Canadian rm, MIL, in exchange for an upfront payment and per unit royalty fee. Before long, MIL was competing against Intel in the market for the 1103, but MIL made a critical mistake in their manufacturing processes, and it wasn’t long before a stream of former MIL customers were knocking on Intel’s door. Along the way, Intel received an inquiry from two disgruntled engineers at Honeywell, asking if Intel was interested in building memory systems. The idea was to mount thousands of 1103 chips on a circuit board that could then be plugged into a mainframe computer to in- crease its memory capability. Impressed by the idea, Intel promptly hired the two engineers and set up a division to do this. Before long, the new division was selling circuit Case 11 Intel Corporation: 1968–2013

C-175

84487_case-11_ptg01_hr_C173-C185.indd 175 22/10/13

4:55 PM

C-176

Case 11 Intel Corporation: 1968–2013 boards to customers running IBM mainframes. This was something of a coup: IBM would not even consider buy- ing the 1103, and had started making its own memory chips. Now Intel had access to a formerly closed market that accounted for 70% of all memory sales. Around the same time, an accidental discovery at Intel led to a second product line—erasable program- mable read only memory (EPROM). Read only memory chips (ROM) were nding wide applications in comput- ing. ROM had desired data, a program for example, per- manently burnt into its circuits. ROM was used to store programs, such as a machine operating system, or part of that system. The troubling thing about ROM is that if an engineer made a mistake in programming the chip, he would have to burn another chip, which was a pains- taking and time consuming process. While exploring the reason for failure of 1103 chips in the manufacturing process, Dov Froham, another ex Fairchild researcher at Intel, found that the cause was that some of the “gates” inside the chips had become disconnected; they were oating. Froham realized that this aw in the 1103 had a potential use; it might enable an engineer to design a ROM chip that could be programmed with ease in a few minutes. Moreover, he found that the data on such chips could be erased and rewritten by shinning an ultra violet light on it and the EPROM was born. Engineers loved the EPROM chip, and once Intel solved the manufacturing problem and started to produce EMROM chips in large quantities, demand surged. Bet- ter still, for two years Intel had a virtual monopoly on the product. While other companies tried to produce similar chips, they were unable to solve the manufacturing prob- lems, enabling Intel to charge a relatively high price for a product whose cost was falling every day with advances in cumulative volume. The BirTh oF The Microprocessor By 1971 Intel had already created two revolutionary in- novations in the semiconductor industry, the DRAM and the EPROM chips. A third, the microprocessor, was also created that year. The microprocessor was born out of an inquiry from a Japanese company. The company asked Intel if it could build a set of eight logic chips to perform arithmetic functions in a calculator it was planning to produce. Intel took on the project. Ted Hoff, one of the inventors of the DRAM, wondered if it might not make more sense to build a miniaturized general purpose com- puter, which could then be programmed to do the arith- metic for the company’s calculator. The project was given to Federico Faggin, an Italian engineer who made some of the basic breakthroughs on MOS technology while working at Fairchild. Although the Japanese company subsequently decided not to build the calculator, Intel pushed ahead with the project. Faggin, who worked 12 to 14 hour days for weeks on end, produced several prototypes in short order. (A source of irritation for Faggin was that despite the long hours, his boss, following Grove’s lead, constantly complained that Faggin was late for work!) Due to Faggin’s efforts, by November 1971 Intel had its third product, the 4004 microprocessor. In an article in Electronic News that accompanied its introduction, and which described the 4004 as a computer on a chip, Gordon Moore heralded the 4004 as “one of the most revolutionary products in the history of mankind”. No one paid much attention. People in the computer indus- try viewed the 4004 as a fascinating novelty. Although small and cheap, it could only process 4 bits on informa- tion at a time, which made it slow and thus unsuitable for use in the computers of the time. The 4004 was followed by the 8008 microprocessor, which could process eight bits of information at a time. Although faster, it too was a product in search of a market. In an attempt to speed adoption, Intel started to sell development tools that made it easier and faster for outside engineers to develop and test programs for new microprocessors. Slowly the microprocessor began to make inroads into the computer industry, primarily in peripherals such as printers and tape drives. The personal coMpuTer revoluTion By the mid 1970s and embryonic new industry was ap- pearing, the personal computer industry. A company called MITS based in Albuquerque, New Mexico pro- duced the rst true personal computer. The MITS Altair used an Intel 8080 microprocessor, which was priced at $360. The rst program offered for sale with the Altair was a version of the BASIC programming language, written by Bill Gates and Paul Allen, and designed to run on the 8080. The two had moved to Albuquerque to 84487_case-11_ptg01_hr_C173-C185.indd 176 22/10/13

4:55 PM

be near to MITS, and they had established a company of their own, Microsoft. The Altair was sold primarily to hobbyists who wanted to write computer code at home (for which Microsoft Basic came in handy). In short order, a number of companies sprung up making personal computers. The most successful of the early companies was Apple Computer, which introduced its revolutionary Apple II in 1977. By this time, a num- ber of other companies were also producing micropro- cessors, including Motorola, whose processor Apple used in the Apple II. The Apple II was a big commercial success, in no small part because it was easy to use for, and because one of the most successful early programs, a spreadsheet called VisiCalc, was written to run on the Apple II. The commercial success of the Apple II got the world’s largest computer company, IBM, to take the nascent personal computer seriously. IBM started to de- velop its own personal computer in 1979 in a top-secret project. To speed the product to market, IBM took a mon- umental strategic decision—it decided to use “off the shelf components” to build the PC rather than develop everything itself, which had been the norm at IBM. Orig- inally the company planned to use a microprocessor from Motorola and an operating system called CP/M from a company called Digital Research. However, Motorola was late developing its product, and Digital Research’s CEO, Gary Kildall, proved to be dif cult to work with. Casting around for alternatives, IBM contacted Intel, offering to purchase it’s latest microprocessor, the 8088, which was a derivative of Intel’s 8086 chip. However, IBM did not tell Intel what the microprocessor was to be used for (originally Intel was told that it was to go in a printer). As part of the deal, IBM insisted on alternative sources for the 8088. Reluctantly Intel allowed AMD and a number of other companies to produce the 8088 under license. A 1982 cross licensing agreement with AMD, which gave AMD the right to produce the 8088 chip, would come to haunt Intel for years to come. For the operating system of its rst PC, IBM decided to use MS-DOS, a Microsoft operating system. Origi- nally developed by Seattle Computer, and called Q-DOS (which stood for quick and dirty operating system), Q-DOS was purchased by Microsoft for $50,000 when Bill Gates heard that IBM was looking for an operating system. Gates renamed the product, and quickly turned around and licensed MS-DOS to IBM. In what was to be a stroke of genius that had enormous implications for the future of all parties involved, Gates, sensing that IBM executives were desperate to get their hands on an op- erating system in order to get the IBM PC to market on time, negotiated a nonexclusive license with IBM. Executives at Intel, who by now had realized that IBM was developing a personal computer, were pro- foundly unimpressed with the choice of MS-DOS and Microsoft. After a visit to Microsoft, one Intel executive noted: “These people are akes. They’re not original, they don’t really understand what they are doing, their ambitions are very low, and it’s not really clear that they have succeeded even at that.”6 For its part, Microsoft had to produce a version of MS-DOS that would run on the Intel microprocessor. From now on, like it or not, Microsoft and Intel would be joined at the hip. Introduced in 1981, the IBM PC was an instant success. To stoke sales, IBM offered a number of ap- plications for the IBM PC that were sold separately, in- cluding a version of VisiCalc, a word processor called EasyWriter, and well-known series of business programs from Peachtree Software. Over the next two years, IBM would sell more than 500,000 PCs, seizing market lead- ership from Apple. IBM had what Apple lacked, an abil- ity to sell into corporate America. As sales of the IBM PC mounted, two things hap- pened. First, independent software developers started to write program to run on the IBM PC. These included two applications that drove adoptions of the IBM PC: word processing programs (Word Perfect) and a spread sheet (Lotus 1-2-3). Second, the success of IBM gave birth to clone manufacturers who made “IBM compat- ible” PCs that also utilized an Intel microprocessor and Microsoft’s MS-DOS operating system. The rst and most successful of the clone makers was Compaq, which in 1983 introduced its rst personal computer, a 28-pound “portable” PC. In its rst year, Compaq booked $111 million in sales, which at the time was a record for rst year sales of a company. Before long, a profusion of IBM clone makers entered the market, including Tandy, Zenith, Leading Edge, and Dell Computer. This entry led to market share fragmentation in the PC industry. By 1982, Intel had a replacement chip ready for the IBM PC, the 80286 microprocessor. The 80286 was des- perately needed since the 8088 was painfully slow run- ning some of the newer applications. IBM introduced a new PC, the AT, to use the 80286 chip, and priced it at a premium. Demand was so strong that IBM put the AT on allocation, which opened the door to clone makers, par- ticularly Compaq. By now, 70% of the microprocessors sold to PC manufacturers were made by Intel, with AMD Case 11 Intel Corporation: 1968–2013

C-177

84487_case-11_ptg01_hr_C173-C185.indd 177 22/10/13

4:55 PM

C-178

Case 11 Intel Corporation: 1968–2013 accounting for a signi cant portion of the remainder. For the 80286, Intel had cut the number of licenses down to 4. It also ran an intensive marketing and sales campaign, called Checkmate, which was successful in getting many Original Equipment Manufacturers (OEMs) to use Intel’s version of the 80286 in their machines. The draM deBacle In 1984 Intel booked revenues of $1.6 and made almost $200 million net pro t, up from $134 million in revenues and $20 million in net pro t a decade earlier. The growth had been dramatic. However, Intel’s share of the DRAM market had been sliding for years. New entrants, particu- larly from Japan, had been grabbing ever more DRAM sales. They had done this by undertaking large scale investment to build ef cient fabrication facilities (fabs) and paying meticulous attention to quality and costs, do- ing everything possible to drive up yields. One source suggested that while peak yields and U.S. DRAM plants, such as Intel’s, were around 50%, in Japan they were closer to 80%. This translated into a huge cost advantage for the Japanese producers. The American manufacturers, Intel included, had made the crucial mistake of underestimating the Japa- nese threat. Demands from computer companies for second sources had helped to facilitate diffusion of the underlying product technology and commoditized DRAMs. In such a market, advantage went to the most ef cient, and this was the Japanese. Moreover, Japanese companies seized the lead in developing more power- ful DRAM chips. While Intel had created the market for DRAMs, and dominated the market for 1K chips, in each subsequent generation it fell further and further behind. By 1983 when fth generation 256K DRAMs started to appear, Intel was a year behind in the development cycle and as a consequence, was at a distinct cost disadvantage when it introduced its product. Somehow, despite Grove’s aggressive leadership, Intel’s share had fallen to only 1% of the total DRAM mar- ket. To regain market share, management understood that Intel would have to build a new fabrication facility, at a cost of $600 million, and throw company R&D resources behind an effort to bring a next generation 1 megabyte DRAM chip to the market. To make matters worse, the DRAM market was in a big slump, bought on by over- capacity as a result of aggressive investments by Asian pro- ducers, and Intel was losing money in the DRAM business. Faced with this bleak prospect, Intel’s senior manage- ment had to decide whether to continue to compete in the DRAM business, the market they had created, or to focus resources on the more pro table microprocessor market. It was not an easy decision. Irrespective of the econom- ics, there was enormous emotional attachment within the company to the DRAM business. Many at Intel wanted to build a 1 M DRAM. There were also valid arguments for staying in the DRAM business. Some thought that DRAMs were the technology driver in semiconductor manufacturing, and without the knowledge gained from making DRAMs, Intel’s microprocessor business would suffer. In addition, there was the argument that custom- ers would prefer to buy from a company that offered a full product range, and if it exited the DRAM business Intel would not be able to do that. As Andy Grove describes it, a crucial point arrived when he and Gordon Moore were discussing what Intel’s strategy should be. Grove asked Moore, “If we got kicked out, and the board bought in a new CEO, what would he do?” Moore’s reply, “he would get us out of memories”. Grove then said, “why don’t we just walk out of the door, and come back and do it ourselves.” It was one thing to make the decision, another to imple- ment it. Grove removed the head of the DRAM division, recognizing that he was not the man to wield the ax, and replaced him with another manager, who promptly “went native” and started to argue for going ahead with the 1 megabyte DRAM chip. He too was replaced, and a year after the decision was made, Intel nally exited the DRAM business. The Microprocessor Business In 1987 Gordon Moore stepped down as CEO of Intel, passing the torch on to Andy Grove, although Moore re- mained as Chairman. Grove, who held the CEO position through until 1998, and was then chairman until 2005, had no intention of letting Intel’s dominance in micro- processors go the same way as its DRAM business. chip design By now, it was well understood at Intel that the market had an unquenchable thirst for more powerful micro- processors. Software was advancing rapidly, with new 84487_case-11_ptg01_hr_C173-C185.indd 178 22/10/13

4:55 PM

applications becoming available all the time. Running these applications quickly required more computing power, and users were willing to pay a premium for this. Intel knew that consumers would only be too happy to replace their old PCs with better, faster machines. It thus became critical to develop and introduce newer micro- processors. At the same time, the market demanded backward compatibility. The new machines had to run older software, and this implied that each new genera- tion of chip should be able to run older programs. This requirement implied that too a degree, Intel was locked into the microprocessor architecture that had started with the 8086 (from which the 8088 was derived), and con- tinued with the 80286. The next microprocessor in what was now known as the x86 architecture was the 80386, or i386 for short. First introduced in October 1985, i386 was a 32-bit microprocessor that was much faster than the i286. Intel had been trying for over a year to get IBM to intro- duce a machine based on the i386, but IBM seemed to be dragging its feet. The problem for IBM was that an i386 PC would be very close in power to minicomputers that IBM was making a lot of money on. Fearing that i386 machines would cannibalize its product line, IBM seemed to want to keep the i386 of the market as long as possible. At the same time, Apple computer had intro- duced a new machine, the rst Macintosh, which used a Motorola microprocessor. The Apple Mac was the rst computer with a graphical user interface and a mouse. As it started to gain market share, Grove feared that the market might switch to the Apple standard, making it more critical than ever to get i386 based machines on the market. Intel had an ally in Compaq Computer. In 1986, Compaq took advantage of IBM’s sloth to be the rst to introduce a PC built around the i386. Compaq seized the lead from IBM, other computer makers quickly followed, and from then on, IBM started to lose in u- ence and share in the PC business. As the high margin i386 chip gained traction, Intel’s sales exploded, hitting $2.9 billion in 1988, while pro ts surged to $450 million. Over the next two decades Intel continued to drive the industry forward with regular advances in its x86 architecture. These included the i486 (introduced in 1989), the rst Pentium chip (1993), The Pentium Pro (1995), various derivatives of the Pentium Pro architec- ture, and more recently, its 64-bit Core 2 Duo and Quad processor line, rst introduced in 2006. The latest Intel processors have pushed the limits of performance by building two or four processors into a chip. Intel prices new chips at a premium then drops prices as manufac- turing yields improve. It is not unusual to see prices drop by 30–50% in one year. By continually increasing the performance of its chips, Intel was able to vanquish several potential com- petitors, including a series of fast chips from AMD in the early 2000s, and several chips based on an architecture known as reduce instruction set computing, or RISC, that during the 1990s seemed to threaten Intel’s market domi- nance. One notable RISC chip arose out of an attempt by Apple, Motorola and IBM to seize momentum away from Intel with a RISC processor called the PowerPC. However, few companies outside of Apple adopted the processor. The limited volume meant high costs, which were further compounded by manufacturing problems at Motorola, and the PowerPC never gained wide ac- ceptance. In 2006, Apple effectively killed the PowerPC when it announced that it would henceforth use Intel mi- croprocessors in its machines. Following Moore’s law, successive generations of Intel chips have used ever-smaller micron geometries to cram ever more transistors on a chip. Intel’s 8088 chip, introduced in 1979, had 29,000 transistors, the i486 chip, introduced in 1989, had 1.2 million transistors, and by 2012, its most powerful PC chips contained 1.48 billion transistors. By 2012 Intel was working with such small sub micro geometries that more than 100 million tran- sistors could t onto the head of a pin! Compared to its original 4004 chip introduced in 2012, the chips Intel was producing in 2012 ran 4,000 times as fast and each transistor used 5,000 times less energy, while the price per transistor had dropped by a factor of 50,000. Driving forward chip design and production requires very heavy R&D spending. By 2012, Intel was spending over $10 billion a year on R&D, or 19% of sales. This was split between spending on chip design, and spending on improving manufacturing processes. Manufacturing processes Designing and manufacturing these devices requires constantly pushing against the limits of physics and tech- nology. Microprocessors are built in layers on a silicon wafer through various processes using chemicals, gas and light. It is an extremely demanding process involving more than 300 steps and, on modern chips, 20 layers are connected with micro circuitry to form a complex three- dimensional structure. Intel is pushing the frontiers of sub Case 11 Intel Corporation: 1968–2013

C-179

84487_case-11_ptg01_hr_C173-C185.indd 179 22/10/13

4:55 PM

C-180

Case 11 Intel Corporation: 1968–2013 micron geometry. The company is currently is produc- ing transistors that measure just 22 nanometers, whereas most other semiconductor manufacturers are still making 45 nm or 32 nm chips (a nanometer is one billionth of a meter). Intel newest factory in Arizona, designed to come on line in 2014, will push this frontier still further making chips that have just 14 nm geometry. To carve features this small on a silicon chip, Intel uses a technique known as extreme ultra violet lithography. This is a way of printing circuit patterns onto silicon chips that goes beyond lasers and lenses, and utilizes xenon gas and microscopic re ec- tors. If it sounds incredibly complex and esoteric, this is because it is at the leading edge of what is scienti cally possible. Indeed, each new generation of Intel chips relies upon pushing processes beyond what was attainable just a few years earlier. So complex is the manufacturing process, that the high tech fabrication plants, or foundries, required to make microprocessors cost up to $5 billion each. By 2012 Intel had 16 of these plants around the world. Too equip its plants, Intel works very closely with equipment vendors. Due to its scale, Intel enjoys considerable lever- age over equipment suppliers. In some cases, Intel will design a new machine itself, and then have equipment vendors manufacture it. In others, Intel works closely with the vendors on the design of a piece of equipment. As a result, Intel itself holds hundreds of patents relat- ing to the processes for manufacturing semiconduc- tors. Whenever equipment is developed speci cally for Intel’s requirements, vendors are generally prohibited from selling that equipment to other companies, such as AMD, for a given period. When installing new equipment, the goal is to gain manufacturing ef ciencies through increased yields, or other process improvements. For example, in the 2000s Intel switched from using 200 mm to 300 mm wafers in its manufacturing processes. The larger wafers allowed Intel to put more microprocessors on each, increasing throughput and signi cantly lowering costs. Intel is currently working to develop the commercialization of 450 mm wafers and is forecasting that it will start to make microprocessors on 450mm wafers by 2016/2017. If it can achieve this, it will be the rst in the world to do so. This may give Intel an advantage in manufacturing ef ciencies that will be very hard for other chipmakers to match. To boost yields, raising the percentage of processors that come of the line operating perfectly, Intel uses so- phisticated statistical process control procedures. Since even a microscopic piece of dust can contaminate a chip, the speci cations that Intel works to are extremely de- manding and tight. Over time, Intel has turned yield im- provement into a precise science. With each succeeding generation of microprocessor geometry, the company seems able to achieve a steeper learning curve. By con- stantly pushing out the envelop with regard to manufac- turing technology, product design, and yields, Intel has reportedly been able to reduce its unit manufacturing costs for a processor by as much as 25–30% a year. Typically, Intel will re ne new manufacturing pro- cesses in one factory, perfecting yields and reducing costs, and then transfer those processes to other facilities. To do this, it relies upon a methodology known as “Copy Exactly!” Under this methodology, engineers spend up to four years perfecting a new manufacturing technique in one of Intel’s development factories in Hillsboro Oregon. Once they are satis ed with the results, they work to meticulously import every last detail to other factories around the world. Engineers strive to duplicate even the subtlest of manufacturing variables, from the color of a worker’s gloves to the type of uorescent lights in the building. Employees from around the world spend more than a year at the development factory, learning their small piece of the new recipe so they can bring it back to their home factory. The idea is to capture the in nite number of intangibles that have allowed a pro- cess to succeed in plants that have already brought it online. According to one Intel manager: “It’s not just there’s a speci cation or a recipe or a program you put into a machine. It also is what the human being does and how they interact with the machine.”7 The extremes to which Intel engineers go to control the precise conditions in its dozen or so factories has be- come legendary. A few years ago Intel engineers were trying to gure out why one plant in Arizona wasn’t hitting the benchmarks achieved at another in Oregon, where the processes were rst developed. Then it hit them: Arizona’s desert air was so much drier than the air in Portland, and the engineers in Arizona were skipping several steps taken in Oregon to dehumidify. Intel scien- tists theorized that the dehumidifying, besides removing water, also eliminated impurities such as ammonia. So engineers began adding water vapor to the air in the Arizona foundry, essentially making Portland air, and then subjected it to the same dehumidi ers used in Oregon. It worked! According to one engineer, this “shows the level of things you’ve got to worry about when you try to make something as complex as the chips we make.”8 84487_case-11_ptg01_hr_C173-C185.indd 180 22/10/13

4:55 PM

intellectual property From the i386 chip onwards, Grove was determined to ensure that Intel was the only supplier in the world of its architecture. AMD, however, believed that under the terms of the 1982 technology sharing agreement be- tween the two companies, it had rights to Intel’s designs. Intel simply refused to hand over technical speci cations for the i386 to AMD, sparking off a lengthy court battle between the two that persisted until 1995. In the end, the two chipmakers agreed to drop all pending lawsuits against each other, settled existing lawsuits, and signed a cross-licensing agreement. Irrespective of the nal set- tlement, AMD had spent $40 million a year on legal fees alone. Senior management attention had been diverted by the ongoing legal battle. AMD had been slow to de- velop its own version of the i386, waiting instead to get speci cations from Intel, which Intel only shared after ordered to in a 1990 ruling. intel inside For years, Intel had viewed its customers as original equipment manufacturers, focusing its m

[TRUNCATED]