My book The Flight of a Wild Duck, will be published on September 1st. Pre Orders can be placed now on Amazon. There were several sections of the book which did not make it through the final editing primarily because they broke the flow of the book. One of them was about the evolution of the IBM compatible PC business. I thought that since it is the 40th anniversary of the announcement of the PC by IBM, I would publish that section of the book here on my blog. I will be publishing this and other essays on the books website .
The IBM PC turns the computer industry on its side.
Although I contributed in only a limited way to the early development of the Personal Computer (PC) industry, I do have a unique perspective resulting from my work at Digital Equipment Corporation, my time as President of Franklin Computer, and, most importantly, my role as Vice President of Business Development for Intel. The story of the IBM PC offers so many valuable lessons. The most of important of these is this: listen to the market.
In 1964, when IBM introduced the System/360 line of mainframe computers, it was primarily a tabulation and typewriter company. This “bet-the-company decision” sure paid off. In just a few years, IBM became the leading computer company in the world, capturing more than 70% of computer-industry revenues. Strangely, a company that made its fortune by developing a line of compatible computers at different price points, all running the same software, would end up damaged by that very same phenomenon of software compatibility.
The story of the IBM PC is not primarily about hardware. It is a story about software, as well as unintended consequences, hubris, and greed. Nobody at IBM imagined that their own actions would transform the computer industry in such a way as to significantly diminish their leadership role and even put the company’s existence in peril. Nobody at IBM could imagine that many small companies would form, each taking a bite out of IBM, let alone that among these thousands of piranhas would be the likes of Bill Gates and Andy Grove.
Sometime in 1980, the top management of IBM, watching Apple Computer’s growth, concluded that IBM should enter the nascent PC market. Apple had transitioned from serving a hobby market into building a productivity tool, driven by business applications like VisiCalc. If the PC was to create a new market, IBM management wanted to dominate it. Atari, then the leader in computer gaming and a subsidiary of Warner Communications, had contacted John Opel, IBM’s president, with an offer to develop a PC for IBM to serve the business market. Frank Cary, IBM’s CEO, asked Bill Lowe, the General Manager of its Entry-Level Systems Division in Boca Raton, to evaluate this possibility. A few months later, Lowe presented his review to the IBM management committee. He recommended that IBM acquire Atari from Warner and use a repackaged version of the Atari 800 computer as a starting point for IBM’s entry to the PC market. He even demonstrated to the committee a modified Atari 800, which had been brought to market a year earlier and was doing well. My good friend, Steve Mayer, one of the founders of Atari, had been the lead developer of the 800. This computer was designed for entertainment and games, so why Lowe thought it could serve IBM’s needs is a mystery to me. Perhaps he thought IBM could never develop a competitive PC in time to enter the market and so at least wanted to take the available opportunity. Or perhaps he cleverly manipulated the committee to give him the authority to move forward outside of the normal IBM processes.
The IBM executives wisely rejected Lowe’s Atari proposal. However, they understood the challenges of creating new products within the IBM culture. Developing a new computer at IBM could take five years or more, as IBM’s cultural focus on quality impacted both design and manufacturing. Indeed, Lowe’s team had already spent almost four years developing the IBM DataMaster.
IBM had numerous committees and procedures. Anyone could say “no,” and no one could say “yes.” To counter this, Cary had already begun to encourage the creation of small autonomous units—“Independent Business Units (IBUs)”—to counter IBM’s sluggishness. He asked Lowe to form an IBU to develop a PC. Lowe returned to the Management Committee on August 8, 1980 with a plan to set up a small team to develop the IBM PC. Don Estridge, who had been the Assistant General Manager of Entry-Level Systems, became the head of the newly formed PC team, which also included marketing and sales. Lowe committed to bringing the IBM PC to market in an astounding one year. Therefore, he explained that his team would have to go outside of IBM for most of the critical components, including the microprocessor and the operating system. The Management Committee approved this plan.
One year later, on August 12, 1981, the IBM model 5150—the IBM PC—was launched during a splashy press conference at the Waldorf Astoria in New York City. Its price tag was $1,565 (about $4,400 today). No one at IBM or anywhere else understood that the computer industry had just been turned on its side. IBM had launched a near-fatal blow to its own business, creating a virus that would almost kill IBM and that would prove a mortal blow to most of the other existing computer companies. Yet neither the losers nor the winners knew at this point what was about to happen.
Coincidently, around the same time as Bill Lowe presented to the IBM Management Committee in August 1980, I was presenting my plans for the Professional Series PC to the Management Committee of Digital Equipment Corporation (Operations Committee), the second-largest computer company in the world. The Professional represented Digital’s initial attempt to address the Personal Computer market by offering a proprietary system that was compatible with Digital’s line of computers ecosystem. While the IBM PC Group primarily bought off-the-shelf components, under my leadership, Digital designed almost everything from scratch. The Professional would not ship until mid-1983, by which time IBM and manufacturers of IBM-compatible computers would be well-entrenched in the market. Interestingly, ultimately both Digital Equipment and IBM would not succeed in the transition to personal computers. The leadership would pass to Microsoft and Intel in the most remarkable way.
In order to meet the one-year product shipping deadline, the IBM PC team used existing hardware components from other companies. Importantly, this included the microprocessor: the Intel 8088. IBM also licensed the computer’s operating system from Microsoft. The decision to go outside IBM for most of the PC’s hardware and software was probably the most consequential decision taken to date concerning the development of the computer industry. But the decision’s impact was not understood by anyone at the time, certainly not by IBM, and certainly not by Apple, at the time the leading PC company. On the day of the IBM announcement, Apple ran a full-page ad in the Wall Street Journal headlined, “Welcome, IBM. Seriously,” with the touch of arrogance only Steve Jobs could supply. At this time, Apple had about 20% of the PC market. Their share would drop below 4% just a few years later. They mistakenly thought they would be competing only with IBM.
The PC unit’s code name was “Project Chess,” and the first computer was called “Acorn.” Strangely, Bill Lowe was promoted right in the middle of the PC’s development, and Don Estridge, his number two, was now placed entirely in charge. Don was a wonderful man who died at 48, far too early, in a 1985 plane crash. I met him at a few industry events, including the PC Forum in 1982, and I liked him. We spoke in April 1983, just after I hired two key members of the original IBM PC development team to join me at Franklin Computer: Bill Sydnes, the development leader, and Lew Eggebrecht, the architect. I called Don to reassure him that I would make sure Franklin got no proprietary IBM information from either Bill or Lew. Don shared with me that when IBM was developing the PC, they knew of my work on the Professional Series at Digital. He said our project was the only one IBM considered competitive. Digital, the second-largest computer company in the world, dominated the minicomputer market in the same way as IBM dominated mainframes. It’s easy to imagine that IBM thought the competition would be other major computer companies like Digital, not startups like Compaq building IBM-compatible PCs.
The power of small teams
Even after all these years and with the opportunity to have worked closely with Bill Sydnes, I still can’t figure out how IBM got their PC done in just one year. The team was just 13 individuals, and that included marketing, licensing, and software people. But the power of small teams should not be underestimated. Yes, they used off-the-shelf components and subsystems, but these were many, and they required integrated.
On the other hand, I had hundreds of people working on the Professional at Digital and spent three years developing an unsuccessful product. The IBM team traded off functionality and quality in favor of time; the one-year deadline was held absolutely constant. At Digital, we instead favored functionality and quality. We failed to understand what Adam Osborne, the founder of Osborne Computers, meant when he spoke of being “adequate.”
Many years after the introduction of the IBM PC, Andy Grove described the change from a vertical to horizontal computer industry in his 1995 book, Only the Paranoid Survive. He reflected on this change in the context of what he termed “a strategic inflection point”:
“[A] strategic inflection point is a time in the life of a business when its fundamentals are about to change. That change can mean an opportunity to rise to new heights. But it may just as likely signal the beginning of the end.”
The structural change in the computer industry from vertical to horizontal integration would take more than a decade; throughout the 1980s, the computer industry remained dominated by mainframes and minicomputers. The transition would only become clear to everyone much later. I think Andy thought the change would be permanent, but it did not hold through the ascendance of mobile computing devices, such as the iPhone.
Can’t put the genie back into the bottle
Compaq was the first to throw a sword into the body of IBM by designing a fully compatible computer. It was followed by tens of other companies. IBM never imagined that other companies could develop a fully compatible PC without infringing on their BIOS copyright, nor could they believe that such companies could succeed even with a non-infringing BIOS. In 1984, Lance Hansche, a senior executive at Phoenix Computers, discovered a failed clone company in Texas that had successfully developed a non-infringing BIOS. Phoenix acquired the BIOS and also made changes to MS-DOS, like Compaq did, offering full IBM PC compatibility to the many companies that wanted to offer clones without going through the expense and time needed to develop a legal BIOS. Phoenix’s many customers included Hewlett-Packard, AT&T, Tandy, Gateway, and Dell. Now that any company could develop a fully IBM PC compatible computer, many did. By 1986, clone manufacturers had more sales of IBM-compatible PCs than IBM.
In 1984, IBM launched the IBM AT. This was perhaps the most consequential event in creating the PC industry. The AT utilized the Intel 286 processor which was backward compatible with the 8088 that IBM used in its original PC. This solidified the importance of software compatibility and unknowingly, IBM handed the keys to the PC Kingdom to Intel and Microsoft.
IBM, soon enough realizing that they could not compete with the clone manufacturers who had a much lower cost structure, decided to develop a proprietary system, called the PS/2, with a proprietary operating system, called OS/2, and a new bus structure, called Micro Channel. They also used their proprietary networking technology, Token Ring. The PS/2 finally launched in 1987, and they paid Microsoft to help them develop it. As silly as it may sound now, IBM paid Microsoft per line of code. So, of course, OS/2 was bloated, although it was probably technically rather strong.
At the same time, Microsoft was developing Windows. Gates tried to convince IBM to use Windows and abandon the development of OS/2. IBM did not agree and soon found itself dependent on what would soon be a competitor in the operating system business. The Power PC microprocessor, which was being developed together with Motorola and Apple.IBM based the PS/2 on the Intel 286 chip. Intel was counting on the 386 to be a big success and tried to convince IBM to use it.
IBM thought they could put the genie back in the bottle. PS/2 would be a success, and clone manufacturers would have to license technology from IBM. The resulting licensing fees would increase prices and provide IBM with the margins they needed to be profitable in the PC business.
IBM strongly resisted using the Intel 386, the first 32-bit microprocessor that was compatible with the earlier x86 processors. They had several reasons for their resistance, including their concerns that a more powerful, 32-bit PC could compete with their minicomputers. IBM was also probably exploring the development of their own 32-bit microprocessor, since they eventually developed the PowerPC in cooperation with Motorola and Apple. In an interview with PC Magazine from March 25, 1997, Bill Gates stated that IBM was concerned that Intel was not capable of getting the 386 done. I’m not sure if that was really IBM’s view or just an excuse they gave Bill.
I’m still amazed that IBM thought they could succeed with this strategy. Perhaps if the PS/2 was vastly superior to the PC clone products, they might have had a chance. But it wasn’t. It would have been much better for IBM to exit the PC business sooner—as they eventually did—to specialize on integrating PCs into the corporate world. After Asian PC companies entered the market, they drove prices down to the point that IBM had no choice to exited the PC market altogether in 2004, selling their PC business to Lenovo, a Chinese company.
Not having IBMs support for the 386 put Intel in a quandary. Intel decided to do everything they could to get Compaq to bring out a desktop PC using the 386, leapfrogging IBM. At one meeting at Intel I attended, someone said Compaq would be the rabbit that the dog (IBM) would chase. Once Compaq came out with a 386-based PC, the rest of the clone industry followed. Around this time, Dick Boucher, an Intel Vice President and my officemate, discussed the IBM situation with me. IBM had a license to manufacture a certain percentage of the 286 chips they were using. As Dick was the senior executive dealing with IBM, IBM had asked Dick if Intel would agree to let IBM manufacture more units, since they had unused capacity at their factory. IBM had negotiated rights to manufacture a portion of all future processors in the x86 family, including the 386. I suggested to Dick that Intel agree to let IBM build more 286 chips if they gave up the rights to the 386 and future x86 processors. IBM agreed,
IBM found themselves in a difficult spot. The clone manufacturers did not move to IBM’s proprietary technologies. IBM could not compete with the clone manufacturers, nor could they get the industry to move to any of their proprietary technology. IBM never brought the PowerPC to the desktop market successfully, though Apple did use the microprocessor for its Macintosh computers starting in 1994. The PC market that looked so good to IBM in the early 1980s was now a rock tied around their necks in a sea of clones and they were drowning.
The American clone companies were not immune to the pressures of Asian competitors, either. Intel was very active in developing low-cost PC manufacturers in Asia who were willing to accept low margins. It served both Intel and Microsoft well to have many PC manufacturers, none with any market power. Effectively, the PC manufacturers became distributors for Intel and Microsoft products. Microsoft, in fact, could sell their products directly to end users, bypassing the PC manufacturers entirely, and even began to sell hardware peripherals like mice and keyboards, as well as application software like Microsoft Office. Access to end users was a very important advantage for Microsoft, one which Intel did not have. Intel’s lack of such access is probably one of the reasons that Microsoft is worth so much more than Intel today.
Bring in the Clones
Compaq Figures It Out
Many think that Microsoft and Intel created the PC revolution, but Compaq actually initially set the course. When the IBM PC was introduced, Intel and Microsoft were both concerned about very different things besides the PC. Intel was worried that its x86 architecture was uncompetitive and would soon lose out to the various RISC processors being developed by competitors like Motorola and National Semiconductor. And so Intel embarked on its ambitious plan to develop the 32-bit i432. Microsoft, meanwhile, was working on a operating system, Xenix, which was based on Unix.
Compaq was founded by three engineers working at Texas Instruments: Rod Canion, Jim Harris, and Bill Murto. Canion documented the story of Compaq in a 2013 book, Open: How Compaq Ended IBM’s PC Domination and Helped Invent Modern Computing. Wanting to leave Texas Instruments and start their own company, Canion, Harris, and Murto explored several ideas, including developing an add-on hard disk for the IBM PC. Eventually, on January 8, 1982, they decided to create a portable version of the IBM PC. A nascent venture capital fund, Sevin Rosen, put together Compaq’s first financing of $1.5 million, alongside the well-known venture firm, Kleiner Perkins. Sevin Rosen had also, importantly, backed the startup Lotus Software, which developed Lotus 1-2-3, the most critical application (a spreadsheet) for the IBM PC. A little over a year later, in 1983, Compaq began shipping their portable computer. They ended that year with an astonishing $111 million in sales.
In 1982, Ben Rosen, who became the Chairman of Compaq’s Board, still owned the PC Forum, the most important industry conference when I attended with the Digital Equipment team to announce its three personal computers (the Professional, Rainbow, and DecMate II). I flew on Digital’s Learjet to get to the conference in time. A private limousine met me at the airport and took me to the speakers’ dinner, where I met Steve Jobs for the first time. My life had changed.or, at least, so I thought. Just three years earlier, I was living in Israel and developing medical computer systems for a small company. Now I was having dinner with the people who were in the process of restructuring the computer industry.
At the conference, Ben took me aside to show me the prototype Compaq Portable, which would be released in March 1983. It made little impression on me. I found it difficult to imagine people carrying around this “luggable,” which weighed 28 pounds, looked like a sewing machine, and cost more than $3,000 (more than $7,000 in today’s dollars). I did not understand the coming revolution in the industry, and it would have been incomprehensible had you told me that, less than 20 years later, Compaq would acquire Digital.
Two initial strategic decisions set Compaq on a course for success. First, and most importantly, their computer was 100% compatible with the IBM PC. That meant that all software developed to run on the IBM PC would run on their system, too. The second, was to offer a portable computer which was a product that IBM did not have.
The Apple v. Franklin ruling in August 1983, as I described in Chapter X, made clear that a BIOS could be copyrighted. The BIOS, a small sliver of software, acted as the interface between the operating system and the computer hardware. It was often called firmware, because it did not change and was stored in Read-Only Memory (ROM). Application software was supposed to access hardware through the operating system, not the BIOS directly. Still, application software developers would sometimes reach down into the BIOS to get higher performance from their applications. Once IBM released its PC, many companies began to offer IBM-compatible PCs. Those companies just copied IBM’s BIOS in the same way as Franklin had copied the Apple II BIOS. They therefore infringed on the IBM copyright in the same way as Franklin Computer had infringed on Apple’s. Starting in late 1983, then, after the Apple v. Franklin ruling, IBM used the courts to put them out of business. It was not pretty.
Compaq, by contrast, understood that they had to develop a fully compatible BIOS that did not infringe on IBM’s. Compaq did so successfully, no easy task. Compaq also had to reengineer Microsoft’s operating system, MS-DOS, so that it was compatible with IBM’s PC DOS. Eventually, Compaq licensed its changes to MS-DOS back to Microsoft, as Microsoft had incorrectly assumed that application software for other personal computers would use its application programming interfaces (APIs) without reaching into the underlying hardware. The BIOS, Microsoft thought, would be unique for every computer manufacturer. As it turned out, that was not what happened. Clone manufacturers would routinely bypass the operating system to improve performance or reduce programs’ size.
Compaq’s second critical strategic decision was to offer a portable as their initial product. Since IBM had no portable computer, Compaq’s portable provided customers permission and an opportunity, especially at larger corporations, to buy a non-IBM computer that was nevertheless software-compatible with the IBM PC. By the time Compaq was formed, the market had some portable (luggable) computers, like the Osborne I and the Kaypro II, and these were doing well. Compaq believed that the combination of portability and compatibility would be a killer proposition on the market. It was!
Three Phases of the PC Market
The PC business evolved over three major phases. Computer hobbyists drove the first phase. The second phase was driven by personal productivity applications, such as word processing and spreadsheets. Widespread adoption of email drove the third phase, in which the PC became a communications device.
Compaq successfully transitioned from the second to the third phase. By 1995, most of its profits came from servers, not the desktop or notebook computers used by individuals. Intel played a big role in making this happen. Sun Microsystems, which primarily sold workstations, had developed its own proprietary RISC microprocessor, SPARC, which it first brought to market in 1987. It considerably hardened Unix, releasing the result in 1993 as Solaris, which earned considerable share in the server market. Intel had initially wanted to capture this market with the 486 but realized that Microsoft’s version of Unix would not be adequate to accomplish this objective. Linux, an open-source Unix kernel initially developed by a Finnish student, Linus Torvalds, was gaining prominence throughout the 1990s, and the open-source versions of Unix built on top of the Linux kernel were gaining in sophistication. Intel launched a development group focused on making contributions to the Linux source in order to make Linux a viable commercial offering for servers. Along with other companies that took an interest in competing directly with Sun, Intel established a non-profit industry consortium, the Open Source Development Labs (OSDL), to consolidate efforts to contribute to the Linux ecosystem and compete with Sun. Intel eventually developed the Xeon microprocessor with features aimed at servers, particularly servers running Linux. By 2007, Sun had lost 80% of its value and was sold to Oracle in 2010. Sun should have realized they were really a software company and put Solaris on top of an Intel box. In fact, Barry James Folsom, who ran Digital’s Rainbow group, joined Sun in 1985 to develop their Intel-based business. By 1990, though, he was gone. With the exception of Apple, no successful computer company in more than 30 years has developed their own hardware and software.
By 1997, Compaq believed that they could compete with IBM in the enterprise market with their combined client (PCs) and server offering. IBM was reluctant to offer Intel-based servers, which would directly compete with their mainframes and mini-computers. However, Compaq lacked the direct sales or service channels this strategy would require. Compaq instead depended on a sizable third-party distribution network, and the potential negative reaction of their dealers kept them from building out direct sales (even though Dell had proven the model). Compaq hired Morgan Stanley to help them find a company they could acquire that had strong enterprise sales and service capabilities. A former senior employee of Morgan Stanley, Robert Greenhill, had recently left to form Greenhill & Co. He convinced the Compaq board to terminate their arrangement with Morgan Stanley and let him advise on potential acquisitions.
One candidate considered was my former employer, Digital Equipment. Digital had fallen on hard times, and Ken Olsen, the founder, was forced out in 1992. The company still had many customers, significant technology, and a lot of service revenue. Alta Vista, which for a short time became the leading Internet search engine, was also part of the deal. Digital also had a compelling next-generation microprocessor architecture, the Alpha, which could have been very powerful in servers. In fact, Digital had all the critical ingredients needed for the next phase of computing, but the company did not know how to organize to take advantage. Nor, for various reasons, did Compaq have any interest in these aspects, missing its chance to lead the next phase of the industry: the Internet.
Interested only in Digital’s sales and service organization, Compaq found, after acquiring Digital for an industry-record $9.6 billion in 1998, that these capabilities had been grossly overstated. Compaq had not realized how difficult it would be to integrate the two companies. With two very different corporate cultures, the acquisition was a disaster. The right leadership at Compaq might have managed to find a different path, but Eckhard Pfeiffer, CEO from 1991, had little insight into how to manage a technology company. He, himself, did not even know how to use a PC.
The combined organization lacked the design and manufacturing capabilities—and interest—needed to be successful with the innovative Alpha microprocessor. Just a few years later, in 2002, Compaq itself was acquired by Hewlett-Packard in a very controversial move. Many members of the Hewlett-Packard board were against the acquisition, as they saw correctly that they would be doubling down on a bad business. Carly Fiorina, Hewlett-Packard’s CEO, pushed the acquisition through over their objections, but the board members in opposition turned out to be right. The board fired Fiorina in 2005, just a few years after she bought Compaq. She later ran unsuccessfully for the 2016 Republican nomination for President of the United States.
One of the Greatest Transfer of Wealth the World Has Ever Seen
Meanwhile, Intel took on an increasing amount of the hardware system engineering. Intel not only worked diligently to establish industry standards but also created designs they provided directly to PC manufacturers. This reduced the research and development these companies needed to do, which they passed onto their customers in the form of lower prices. (However, this also created a price barrier for PC manufacturers that wanted to do their own engineering.) Intel developed chipsets to simplify PC designs and supplied complete motherboards. They even developed power supply improvements, which they passed onto power supply manufacturers in Asia. Intel also had a very robust software program in which they worked with third-party software developers to take advantage of the new features of their microprocessors. Intel also supported the development of operating systems besides Microsoft’s, such as Linux.
All of this created tension with the leading PC manufacturers, IBM and Compaq. When Compaq was reluctant to embrace the Pentium in the same way as IBM had been reluctant to embrace the 386, Intel used companies like Dell, Gateway, and Packard Bell as the “rabbits” that would force Compaq to move.
The result throughout the 1990s was a very fractured PC market in which all the power resided with Intel and Microsoft. Accordingly, Intel and Microsoft continued to grow and increase their profits. Microsoft’s Gates became a billionaire in 1987 and was the wealthiest man in the world by 1995. Microsoft and Intel were both consistently among the largest five companies in the world by market capitalization towards the end of the 1990s. A common but true joke held that Intel and Microsoft made more than 100% of computer-industry profits at this time because so many others were losing money.
Intel and IBM
Intel had become a memory supplier to IBM for several years. IBM decided to use Intel’s 8088 microprocessor for its PC for many reasons: IBM was comfortable with Intel, IBM could build a cheaper system using an eight-bit processor which ran 16-bit code, and, most importantly, Intel had a software development system that ran on the same processor. Writing software for Motorola’s or National Semiconductor’s processors, for instance, required a mainframe computer. IBM hoped to get many third-party developers for their PC, so this was a very important consideration. Lew Eggebrecht, the architect of the IBM PC, made the decision to use the 8088.
No one at Intel knew about the PC until about halfway through its development cycle. Perhaps if they had, they would have also offered IBM the possibility of licensing the operating system that Intel had developed for the 8088 development systems. There had already been discussions within Intel about entering the software business. Terry Opdendyk, who later became the CEO of VisiCorp, the company that distributed VisiCalc, the first spreadsheet, was running the Development Systems Group at the time. He argued against Intel licensing the operating system they had developed, because it would allow other companies to compete with Intel in the business for microprocessor development systems. The decision at that time is understandable. No one understood the opportunity that the PC represented but it was an unfortunate decision. It would have been very powerful to couple an operating system to a microprocessor so that they could evolve together. Strangely enough, Intel’s operating system was called ISIS (Intel System Implementation Supervisor). Gary Kildall, one of the real legends in the PC industry who went on to found Digital Research and create CP/M, had been consulting for the Intel Development Systems Group. They stopped his contract and brought all the work inside, Kildall was forced to start his own company to sell his operating system. That is how CP/M was born.
After IBM launched the PC, they worried about their dependence on Intel which was going through challenging economic times due to heavy competition in their core business of memory chips. IBM took several actions from 1982 to 1984. First, they invested $643 million for 20% of Intel stock. Second, they required that Intel license a second source for microprocessors (Advanced Micro Devices, or AMD). Third, they got certain limited manufacturing rights, as I described above. Five years later, IBM had divested themselves of all their Intel stock for a nice profit—but it was nothing compared to the future value of those shares. They also gave up their board seat. They had no idea how valuable and powerful Intel was about to become.
IBM and Microsoft
Bill Gates’s mother, Mary Gates, served on the board of the charity, United Way, along with John Opel, the President of IBM at the time of the IBM PC program. While is possible that she discussed her son’s company with Opel, and he then suggested to the PC team that they meet with Bill Gates. Frankly, I doubt this. By this time, Microsoft was the leading supplier of programming languages for personal computers. It was only natural that the IBM team would meet with them.
Some of the IBM team, led by Jack Sams, went to Redmond, Washington, to meet with Gates, then just 23 years old. Microsoft had fewer than 40 employees and revenues of $7.5 million. The IBM team wanted to license various programming languages, including BASIC. But they also wanted to license CP/M, which was the most popular operating system for personal computers. IBM mistakenly thought that Microsoft could do that. Gates explained that CP/M belonged to Digital Research in Monterey, California, founded by Gary Kildall. Microsoft and Digital Research had a good relationship; they were not yet competitors.
Gates called Kildall and said that he should meet with some people from IBM the next day, but he did not and could not tell him more, since he had signed one of IBM’s very restrictive non-disclosure agreements. The meeting between IBM and Digital Research did not go well. Kindall was not even there, leaving his wife, who was Digital Research’s main business person, to handle the IBM team. So, the IBM team went back to Redmond and asked Gates if he could do anything to help them get an operating system. Bill was familiar with a small company in the Seattle area called Seattle Computer Systems that had developed a CP/M look-alike called QDOS (for “quick-and-dirty operating system”). Microsoft paid about $75,000 for the exclusive license to QDOS. They then signed a license agreement with IBM, which was pretty much the beginning of the end for IBM’s PC aspirations. IBM did not want to pay a royalty. Instead, they negotiated a nonexclusive license for a fixed, one-time fee of $430,000. Microsoft was free to license others; of course, they had every incentive to do so since they had no upside in their deal with IBM. This, as it turned out, was one of the dumbest transactions in industry history.
The 386 Almost Did Not Happen
At first, neither Microsoft nor Intel realized the PC’s potential. Like IBM, they merely saw PCs as a new market segment. The IBM PC AT came out in 1984, around the time I joined Intel. Intel still thought that the computer industry would be dominated by vertical, proprietary computer systems. It hoped to develop the microprocessors that would be used for such systems and knew that the 8080 architecture was inadequate for that task.
In 1982, Intel launched the 286 as a successor to the 8088/8086 used in the IBM PC and compatibles. IBM used the 286 in its PC AT. (There was a 186, but since it was not fully compatible, PC manufacturers did not use it.) The architecture of the 286 had many design issues related to memory segmentation. Even as late as 1982, Intel did not view the 8080 architecture as its long-term strategy for the microprocessor marketplace. In fact, personal computing did not even make the list of the top 50 applications Intel’s 286 marketing group considered. Certainly, no one at Intel understood the need for full backwards compatibility. Compaq had to teach that lesson.
The 286, then, was merely a stopgap while the 32-bit i432 completed development at Intel’s Oregon site. The company had decided not to develop a 32-bit version of the 286, a product Bill Gates described as being “brain dead.” Intel was focused on competing with Motorola’s 68000, for which the i432 was Intel’s chosen vehicle. So, in 1982, the 286 development group was disbanded. However, there was a “skunkworks” project under an engineer named Gene Hill to design a 32-bit successor to the 286, which they called the 386. At first, they did not even think about having backwards compatibility with the 16-bit 8080 family.
The i432 was Intel’s attempt to build a “mainframe on a chip.” To use Grove’s description, the computer industry was still vertical. With the i432, Intel was targeting mainframe and minicomputer companies. As it became increasingly apparent that the i432 would be not only late but also extremely slow performing, various engineers and marketing people began to think about alternatives. An architect named Glen Meyers began work on a follow-on to the i432, which became eventually became the i960. Within the 386 design group, led by John Crawford, discussion centered on targeting the workstation market. Some designers advocated giving up backwards compatibility in favor of increased performance.
Compatibility eventually won out, for several reasons. The growth of the PC market was one reason. Another was lessons learned in seeing how Digital Equipment had successfully beaten Data General in the 32-bit market. Mike Richmond, who had recently joined Intel from Data General, was a big advocate for doing the 386 and making it fully backwards-compatible. In the late 1970s, Mike was on a team at Data General that was developing a very powerful computer that was not compatible with their 16-bit Nova offering. While Data General plowed resources into their cutting-edge design, Digital brought the less-powerful VAX to market. VAX was backwards-compatible with their 16-bit PDP-11, which allowed Digital’s customers to easily migrate their software to VAX. It also allowed Digital to utilize its existing development tools. VAX was a huge success in the marketplace, as a result of which Data General scrambled to bring out a 32-bit, backwards-compatible system, the Eclipse MV/8000. This story was covered in Tracy Kidder’s Pulitzer Prize–winning 1981 book, The Soul of a New Machine, which, according to Mike, most of the Intel decision-makers read.
On October 11, 1982, a memo sent to a number of managers and executives, including Andy Grove and Les Vadasz, stated that the 386 would be backwards-compatible with the 286. This was one of the most important decision Intel would ever take, and Intel’s own needs for compatibility played a major role. Backward compatibility meant that Intel would not have to rewrite their development environment of compilers, assemblers, and debuggers. Bill Davidow, who was responsible for Intel’s development systems, pushed this hard. When I asked Lew Eggebrecht, the architect of the IBM PC, why he selected the Intel 8088, he said he chose the 8088 because of its development environment. This was an area where Intel excelled.
To say that Intel almost walked away from the microprocessor architecture that formed the basis of the PC revolution would be understatement.
After completing deals with Intel and Microsoft, IBM began one of the most significant transfers of power and wealth in the history of the computer industry—maybe any industry. By 1992, the combined market capitalization of Intel and Microsoft exceeded that of IBM. One year later, Microsoft alone would surpass IBM in value. By 1999, Microsoft was worth three times IBM. In the process, Bill Gates, who had was just 23 years old when he made his deal with IBM, became a billionaire in 1987 and the world’s richest man in 1995.
Unlike Intel’s ambivalent approach, Microsoft wanted to sell operating systems for the PC market. Specifically, they wanted to bring Unix to the PC market. Unix, a very successful operating system developed at Bell Labs, was pretty much freely licensed, especially to universities. Microsoft got a license but were not allowed to use the Unix name. They called their product Xenix. (Eventually, Apple would also base their operating system on Unix.) Microsoft only acquired QDOS, which they renamed PC DOS, because they knew IBM needed an operating system, and they wanted to make sure that the IBM PC happened so they could sell tools for programming languages, like BASIC. It is not clear to me when Bill Gates realized that controlling the operating system would be key to Microsoft’s success. It certainly was not right away.
As the PC industry began to take off, what was going on dawned on both companies. Besides seeing the potential impact the PC could have on both of their businesses, Intel and Microsoft were odd bedfellows. Within Intel, the relationship was characterized as being “fellow travelers.” The two companies soon found themselves dependent on each other. Many in the industry eventually used the term “Wintel” to describe Microsoft and Intel. Neither Intel nor Microsoft ever used that term internally, which greatly exaggerated the closeness of the two companies. While cooperation was in both companies’ interest, particularly in product development, each company also disdained the other. This played out in many specific conflicts over the years, fundamentally resulting from differences in their respective business models. David Yoffie and Ramon Casadesus-Masanell of Harvard Business School published an excellent, very detailed analysis of this in 2005. David Yoffie was on the board of Intel at the time.
Conflicting Business Models
As Yoffie and Casadesus-Masanell explained, Intel only made money when customers bought new computers. Intel also had to push customers to buy PCs with high-performance processors, where it had less competition from companies like AMD. Intel was always looking for a way to get existing customers to purchase new computers. Intel’s marketing group, led by Dennis Carter, would research what caused customers to buy anything at the high end.
Microsoft, by contrast, could make money by selling upgrades to customers for their existing machines, thus allowing them to delay purchasing a newer computer. In the beginning, this was not such a large problem, because most PC buyers were new customers. Over time, however, the installed base became a larger part of the total market. Intel would push Microsoft to add features to Windows that required its latest microprocessors, and Microsoft would often resist. Microsoft supported the development of competing microprocessor architectures, and Intel supported the development of alternative operating system. In the PC market, neither of these efforts had much effect.
Andy Grove and Bill Gates really did not like each other. Each felt that they were more critical to the development of the PC industry than the other. Once, during a meeting between Bill and Andy at Andy’s home, Andy got so angry he threw Bill out. The two companies were stuck with each other, though. Wintel was very much a marriage of convenience, certainly not love.
Andy was disgusted by Microsoft’s product development, which he considered mediocre. He felt they were doing too much. Inside Intel, Microsoft was often tagged as “a mile wide and an inch deep.” That was pretty much true, but one could have equally labeled Intel as “an inch wide and a mile deep.”
The two companies had joint, quarterly executive management meetings in alternating locations. After I was tasked in 1992 with leading our interactions with Microsoft concerning consumer products and services, I attended those meetings for many years. That gave me a lot of insight into the dynamics of the relationship. I met most Microsoft executives, including Steve Balmer, who would eventually take over as CEO of Microsoft. These meetings, suffice to say, were characterized by a lot of shouting.
Intel’s Strategy for Exploiting the PC Market
Once Andy Grove realized the computer industry was restructuring, he wisely decided to put the full weight of Intel behind seizing the opportunity that IBM had given the company. It was a remarkable demonstration of leadership and courage.
His strategy had 10 key ingredients.
1. Exit the Memory Business
Intel was formed as a semiconductor memory company and became the leading memory supplier before Japanese firms entered the market. In 1985, Intel saw the price of memory components fall by as much as 75% due to Japanese competition, some of which was illegal dumping. In a brilliant act of courage, Andy Grove and Gordon Moore decided to exit the memory business and focus on microprocessors. The fascinating story of how Intel got into the microprocessor business has been written about a lot. I discuss it in some depth in Chapter X.
2. Focus the Company 100% on Job One
Once Andy realized the opportunity that Intel had been given, he decided to focus the company on it. I used to joke that when I joined Intel in 1984, it sold silicon by the ton (the memory business), but then it found a vein of gold (the microprocessor). I would ask: “after you extract all the gold, where are you? In a big hole.” Andy went around and said that Job One was maintaining and expanding Intel’s dominance in microprocessors. Job Two was other business opportunities. Over time, Job Two would be reduced to business that helped make Job One happen.
3. Out-Invest All Other Semiconductor Manufacturers
Intel started to make billion-dollar bets in manufacturing capabilities. While Intel did not have the most competitive microprocessor architecture, they could make up for this by keeping at least a generation ahead of their competitors in terms of semiconductor process technology. Intel could only afford these investments if they could maintain their sale volumes and margins.
4. Eliminate Second Sourcing
Significant customers of semiconductor companies did not want to depend on sole suppliers for several reasons, including supply and pricing. They therefore required that companies like Intel establish second-source suppliers. At IBM’s insistence, Intel had licensed the x86 family to AMD. Intel’s relationship with AMD became very contentious, with much legal action on both sides over time. Intel was eventually able to reduce AMD’s ability to build compatible microprocessors. Intel also grew to understand the power of branding. When a court ruled that a number like “386” could not be copyrighted (not so surprisingly), Intel decided that future products would have copyrightable names. The change in naming was first implemented with the Pentium, which would have been called the 586. AMD’s equivalent product was called the 586. Intel had to put significant effort into branding, because customers assumed that the 586 would naturally follow the 486.
5. Expand the Offering
Since Intel wanted to capture as much of the semiconductor content of PCs as possible, it developed many of the supporting semiconductor chips, which were marketed as chipsets. Making chipsets also ensured support for Intel’s newest processors, which was critical to their rapid adoption in the market. Around 1993, Intel began selling motherboards to computer manufacturers like Gateway 2000, Packard Bell, and dozens of others. Intel provided technical support, reference designs, and development systems. Besides accelerating the adoption of its newest processors, Intel could also use these capabilities to reward or punish their customers. Getting early access to design information would allow a customer to bring a product to market earlier than their competition, and time to market was critical. Some of the chips for microprocessor-supporting standards, such as an I/O bus called PCI, also turned into major businesses for Intel in their own right.
6. Transition to 386 Compatibility
The 286 was a very successful product, and many in the industry—particularly industry leader IBM—were reluctant to move to a 32-bit processor. Unless software took advantage of the additional memory address space the 386 offered, there was no reason to buy a 386 PC. Software developers, meanwhile, were reluctant to create such software without an established market: a classic “chicken and egg” problem, and a significant problem for Intel.
IBM would not move to the 386. In fact, as I described above, IBM was so convinced they did not need the 386 that they were willing to give up their second-source rights to all follow-on products if we allowed them to manufacture more 286 chips in their own factories. Intel decided that Compaq would have to go first and worked out some sales incentives.
Then Intel embarked on a very bold strategy. I remember the meeting in 1989 where Dennis Carter presented the brilliant idea to go after the 286 microprocessor via advertising: the Red X plan. At that time, PC manufacturers were reluctant to move to the 386. There was no real compelling reason for them to do so, because no software really utilized the capabilities of the 386, the first compatible 32-bit processor. Yet there would continue to be no software unless there were customers for 386-based PCs. It was a classic chicken-and-egg dilemma. Dennis proposed putting up billboards with the 286 crossed out. Intel would advertise, in other words, against its own products. Andy deserves a lot of credit for supporting Dennis’s bold idea.
7. Create an Ingredient Brand
In 1992, in a meeting with Andy Grove in his conference room, Dennis Carter placed a Japanese newspaper down on the conference table. Dennis had been Andy’s first Technical Assistant and was now running marketing for the component group. The newspaper had lots of ads for PCs, almost entirely written in Japanese. But one ad had some English, reading “Intel in It,” meaning the PC contained an Intel microprocessor. It was actually Bill Howe who ran Intel Japan that had come up with the idea in the first place. We were stunned that this company would make such an ad and realized that some value must be associated with our name. But it may also have been the result of Howe giving them a discount if they ran such ads. It took awhile for Dennis to convince Andy. Once Andy agreed, Carter went on to develop the very successful Intel Inside marketing program which involved givin marketting rebates to OEMs that supported the Intel inside program.
This was not the world’s first ingredient branding program for consumer electronics. Dolby, for example, launched one in the 1970s for audio. Under the program, Intel rebated 3% of the price PC manufacturers paid to Intel to support their advertising as long as it followed various Intel Inside protocols.
Intel almost destroyed its brand in 1994 when Professor Thomas R. Nicely discovered a flaw with the some of the math calculations of the early Pentium Processor. When the press began to report the problem, Intel took the position that it was a minor problem. The media fueled what became a public outrage. IBM got on the bandwagon against Intel, likely because they did not want the Pentium to succeed (by this time they had introduced the PowerPC alongside Apple and Motorola). I remember Andy stonewalling on the issue, rationally insisting that the problem was tiny for almost all customers. He failed to understand the public relations implications, but he learned quickly. To quell the media firestorm, Intel offered to replace any Pentium Processor upon customer request. In the end, few did so. For most customers, replacement was more hassle than it was worth, and Intel was right in saying the “Pentium bug” was not a big problem. Andy and all of Intel learned the importance of understanding customers’ emotions and the power of the press.
8. Focus on the Relationship with Microsoft
After the introduction of the IBM PC and clones, it became evident that Microsoft would supply the platform’s operating system. Fortunately for Intel, MS-DOS only ran on the Intel architecture. Microsoft could have developed a version for other microprocessors. Since none of these were successful in the PC market, however, there was no real benefit to Microsoft to commit to the product development and technical support required to sell MS-DOS on other microprocessor architectures.
Once it became clear that the PC market would comprise many different suppliers of PCs, Intel management realized that this situation was perfect for Intel. None of the suppliers had enough market share to dictate the industry’s evolution, not even IBM. But Intel also realized that the PC manufacturers’ narrow profits would keep them from doing much in the way of R&D. So, Intel set up an independent systems development group in its Oregon facilities, led by Craig Kinnie and reporting to Ron Whittier. The group included a number of very talented people, like Steve McGeady, and recorded an extensive list of accomplishments, including USB, various multimedia technologies, and Wi-Fi.
9. Leverage Intel Capital to Grow the Market
Les Vadasz and I started what would become Intel Capital in 1991 as Corporate Business Development. I had been making venture investments on behalf of Intel since 1988. One of our missions was to use venture investment to expand the market for PCs. I discuss Intel Capital in greater detail in Chapter X.
10. The PC is IT
Though it became clear that Intel could be the dominant supplier of microprocessors for the PC, the PC also had a competitor in the home. A significant effort, which I describe later, aimed to make the television the home’s central interactive device. For example, Larry Ellison, the founder and CEO of Oracle, and also Sun advocated that “The Network Is the Computer.” They evangelized the idea of a “Thin Client” that would connect to powerful, intelligent networks of computers. Intel began a campaign to convince the industry and consumers that the PC was the best possible interactive device in the home. I was very active in this campaign and discuss it in much more detail in Chapter X. In this respect, Andy came up with the powerful rallying cry, “The PC is IT,” a concept I totally embraced.
Bill Gates, too, soon understood the PC’s potential. Gates wanted to dominate desktop software and indeed all of the software industry. While Bill is very smart, he frankly is not the smartest person I have ever met. He is undoubtedly the most competitive, however. I also don’t think Bill was particularly good at mapping out the future. But he was great at recognizing others’ early successes and then stepping out in front of their parade.
Make Sure Every PC Came Bundled with a Microsoft Operating System
The key to Microsoft’s strategy was to get its operating system bundled into every PC shipped, and they were hugely successful at this. MS-DOS became the platform for application software developers. Digital Research and other operating system companies stood no chance. Though IBM had given them the opportunity, Microsoft knew how to take advantage. Microsoft also realized that they could sell upgrades directly to end-users later at a great price. (As I already mentioned, that led to conflicts with Intel, as we could not sell customers upgrades but instead relied on their purchases of replacement systems.) Microsoft sold operating system upgrades like Windows in a boxed set. When Windows was announced, customers lined up to buy it.
Leverage the Operating System to Expand into Application Software
Microsoft’s customers were the PC manufacturers and, ultimately, their end-users. Microsoft did not have to worry about the reaction of application software developers, such as Lotus, because those companies had no choice but to offer their products to run on top of the Microsoft operating system. Microsoft therefore had free rein to develop their own application software, competing with Lotus by building a spreadsheet program called Multiplan. Microsoft first offered a word processor in 1983. PowerPoint resulted from Microsoft’s first significant acquisition and came to market in 1990. Over time, Microsoft added more and more application packages, which eventually grew into Microsoft Office.
Eventually, Microsoft grew to dominate the operating system, development languages, and critical key applications (show graph to illustrate).
Embrace, Extend, and Extinguish
Embrace, extend, and extinguish was a phrase internal to Microsoft that the U.S. Department of Justice found during its antitrust investigation. Microsoft would sometimes embrace an industry standard and then extend it with proprietary functionality. Occasionally, Microsoft would acquire a startup just to put them out of business. Microsoft would also change things in the operating system, especially Windows, that would break software from other companies. The saying went, “Windows is not done until Lotus doesn’t work.”
Microsoft could establish their brand with end-users since they controlled the splash screen. They put the Microsoft and Windows logos right where the customer could see them. They also offered end-user hardware products, such as a Windows keyboard and mouse.
Intel, Microsoft, and the Next Wave
Even as both Intel and Microsoft experienced amazing growth both in terms of revenue and in terms of power throughout the 1990s, changes afoot would substantially negatively impact both companies’ futures.
For years, Intel focused on increasing their processors’ performance. Many at Intel believed that there was a price point that consumers would pay at which they wanted as much computer performance as possible. It started to become difficult, however, to find applications that could use the increased performance and would drive customers to buy newer computers with faster processors.
I vividly remember a conversation with Andy during which I told him that I had detected a change in the market. I said that before, when I went to parties, I heard people boasting about how fast their computers were. Lately, though, I was hearing people instead brag about how cheap their PCs were. I also pointed out that our competition was not other chip manufacturers but the monitor companies. Customers with only so much money to spend would have to decide between a faster computer or a bigger screen. Andy eventually came to realize this was true. As portable computers gained market share, Intel also struggled with power management. In engineering terms, higher performance and longer battery life stood in direct conflict.
After Netscape debuted, Gates came to realize that the Internet would be critical to the future of the PC industry. Microsoft developed a browser, Internet Explorer, and bundled it with Windows. This launched the U.S. antitrust investigation of Microsoft, but Microsoft won the browser wars until Google came along. Like many others, Microsoft wanted to use the browser to create a web portal, like Yahoo, Lycos, or Excite. They failed to understand that search would be the killer application and so yielded industry leadership to Google. Once they finally did, they tried (unsuccessfully) to buy Yahoo. Microsoft did develop Bing, an “also-ran” search engine.
My story of the impact of the IBM PC finishes around the end of the twentieth century. By then, the Internet was making itself felt, and later came the growth of mobile computers. Of all the dominant players during the PC era, only two thrived: Microsoft and Intel. IBM, while still around, is today a very different company. By 1993, IBM was in a nosedive, with losses that year of $8 billion driven primarily by the very changes in the computer industry that they themselves had unleashed. Most clone companies did not survive. Sadly, Intel got itself stuck in the PC era and has had trouble going much beyond that. After some time similarly floundering, Microsoft proved to be more agile. And though Apple never regained their leadership position in personal computers, they became the leader in mobile computing with the iPhone and iPad.