Idea Man

by Paul Allen

  To the side of the keyboard, Charles rolled a rectangular box with three buttons in a horizontal row: a mechanical mouse. When he moved aside, I sat down for a turn. It took a few moments to coordinate the movement of my hand with the position of a cursor on a screen. But soon the mouse felt like an extension of my arm, and it was then that I realized how a GUI interface could make people so much more productive.

  Charles took over again and dragged an icon representing a document file to a printer icon. A laser printer known as the Dover hummed into action. At the time, dot matrix and daisywheel printers were the bane of personal computer users: slow, loud, and prone to jamming, with just one standard font. But in the magical world of Xerox PARC, I had within seconds a perfect copy of the memo that summarized our meeting with Charles in Seattle: “What is our business? Produce and sell software for micro-mini systems on the mass market. …” The printer was the one PARC brainchild that Xerox, a copier company at heart, would successfully bring to market. A single Dover cost about $200,000, Charles said, but we both knew the price would soon plunge.

  Charles drew my attention to a yellow cable running out the back of the machine, an umbilical cord to a local area network that connected several Altos using Ethernet, yet another patented PARC technology. This nexus of personal computers had all the benefits of old-style time-sharing (a common printer and a file server for additional storage), but none of the drawbacks (slow connections, crippling networkwide crashes).

  Wow! I thought. This is going to change everything. PARC’s achievement seemed both startling and commonsensical. Surely people would want what I’d seen and touched at Charles’s lab. Of course you should be able to interact with a computer with a pointing device, or drag a file to a different folder, or push a button to print what looked like a page from a book. That afternoon in Palo Alto was a thunderclap. Once GUIs went commercial, computers would become so natural and organic that anyone’s mother would learn how to use them. At that point, it seemed to me, nothing could stop their universal adoption. They’d be like television sets; they’d be irresistible.

  PROJECT CHESS WAS so hush-hush that even industry insiders had no clue what we were doing. When I mentioned it to Charles after he started work for us around the New Year, he thought it was one of Vern Raburn’s consumer game products and wasn’t impressed. “We’ve got to focus,” he said. Then I told him about IBM’s move into the personal computer market.

  That was a different story, Charles agreed.

  We were so tight-lipped that we wouldn’t even mention the famous corporate acronym by name. We referred to our customer as HAL, after the computer in 2001: A Space Odyssey. (As movie buffs know, HAL is IBM moved one letter down the alphabet.) New levels of security were needed. When the PC prototype arrived at our office around Thanksgiving, it was closeted in a small, windowless backroom, under strict lock and key. Access was limited to a handful of people.

  A third of our staff worked on some aspect of Project Chess, but the central developments unfolded in this airless ten-by-fifteen-foot space. My direct responsibility was BASIC-86. To conserve the PC’s limited memory, IBM had directed us to embed BASIC in the machine’s ROM. Though we’d done this earlier with Applesoft BASIC, I was nervous as hell about it. The only way to fix a faulty ROM chip was by recalling the machine, and what if our new BASIC had bugs? I knew that you didn’t release complex, first-generation software and have it banged on by hundreds of thousands of users without something bad turning up.

  I hit upon a novel solution: to insert a hundred or more coding “hooks,” so that any part of the BASIC code could be patched or updated from a floppy disk. (For an analogy, think of open envelopes taped to key sections of a book, allowing new material or corrections to be inserted without reprinting the entire volume.) Those hooks turned out to be lifesavers.

  To make sure that DOS would pass IBM’s tests, I chose the steady Bob O’Rear. He would make sure that Tim Patterson’s DOS was compatible with IBM’s BIOS (basic input/output system), the built-in software that controlled the computer’s keyboard and display. In addition to IBM’s prototype, our high-security room contained a blue box called the ICE-88 (for Intel circuit emulator), a diagnostic device to expedite the debugging. The machines generated tremendous heat, and Bob and I sweated like pigs in our Bermuda shorts and T-shirts. In an unused space across the hall sat Bill’s friend Andy Evans, a volatile securities trader in need of a desk and business telephone. Whenever the markets took a turn against him, Andy screamed and hurled his phone against the wall, which could be disconcerting.

  We were working on a crash schedule with a client who was famously intolerant of slipped deadlines. We’d overnight floppy disks with each day’s progress to IBM in Boca Raton, where the software was tested. If some setback made a day unproductive, one of our programmers (unbeknownst to me) would “accidentally” reformat the disk before shipping. When IBM called to complain that they’d gotten a blank disk, he’d apologize for the error and correct it in the next shipment, buying time.

  One big problem was the flaky IBM prototype. Bob kept resoldering loose connections, but it could be hard to trace the source of the defect: hardware or software? Precious days were lost. More delays came from IBM’s fastidious testing protocol, and Bob spent a discouraging amount of time filling out corporate forms. (We joked that IBM’s slogan should have been “Better products through better paperwork.”)

  The original mid-January deadline to have DOS and BASIC working came and went, and we began to worry that IBM might pull the plug if they couldn’t make their scheduled August rollout. Rumor had it that a parallel workgroup in Japan stood ready to replace us if we faltered. On January 19, 1981, Bob expressed our concerns in writing to an IBM manager named Pat Harrington:

  Microsoft is continuing its efforts to bring up 86-DOS and BASIC on the prototype hardware but, due to problems with both hardware and software provided by IBM we have yet to be successful. … These problems have left us several weeks behind schedule.

  Six nights later, Bob got the software up and running. He broke the news to me the following morning, and I’d never felt so relieved. We still had bugs to address, and the IBM printer they sent us didn’t work, but we knew that we were on our way. On May 1, Tim Patterson left SCP to come to work for us. He was a critical reinforcement because he knew 86-DOS inside and out.

  Late that spring, Bill and Kay Nishi made another trip to Japan. For a solid week they were besieged by Japanese computer makers clamoring for a 16-bit operating system. Despite a leak in one of the trade magazines, Project Chess remained top secret, and Bill couldn’t say a word about the DOS we were developing in Seattle. Even so, that trip was telling. Microsoft’s biggest plum, it became apparent, wasn’t the version we’d made specifically for IBM PCs. The real bonanza was the compatible system that we’d call MS-DOS—the product that could be sold over and over again worldwide, under our own name, to companies that would follow IBM’s flying wedge into the 16-bit market. Between the interest in Japan and IBM’s domestic ripple effect, we began to realize that MS-DOS would be the international centerpiece of personal computer technology.

  So it was crucial for us to gain as much control over DOS as we could. In June, I returned to Seattle Computer Products to try to modify our deal for 86-DOS, offering a flat fee of $30,000 for any future licensing. Rod Brock countered by asking for $150,000 for an exclusive license. I upped our offer to $50,000 for exclusive rights and then proposed an outright purchase, throwing in favorable terms on subsequent upgrades of our 16-bit languages. That was the sales agreement that Brock and I signed on July 27, 1981, a contract that laid the foundation for what Microsoft would become. I knew it was a coup, and that a free-and-clear DOS would be a valuable asset. But I cannot say that we knew just how valuable it would be.

  As I stated in a deposition sometime later:

  Bill was very adamant that we should make the contract an agreement of sale. … Bill thought we should have complete owners
hip and control of the product. … [He] felt it was always better, if you wanted to control and benefit from the evolution of a product, to own it as compared to license it. … [He] said it would just make everything cleaner. …

  If Brock had known about IBM, he undoubtedly would have held out for more, and we certainly would have upped our offer. But he was happy to sell. He’d been hit hard by the recession and needed cash, and no one twisted his arm. Brock’s priority was to increase his hardware sales by bundling a reliable operating system with his new computers. SCP wasn’t equipped to partner with IBM as the industry moved into the next era of personal computers—that was never in the cards.* (Five years after the PC’s rollout, Brock fell on hard times and sued Microsoft in an attempt to regain control over the operating system that he’d sold us. Given the uncertainties of a jury trial, Microsoft settled.)

  IBM’s personal computer was announced on August 12, 1981, and shipped ahead of schedule in November. Everyone expected the PC to do well, but no one had anticipated that it would rule the PC market so quickly. Within four years, with Apple’s machines the sole exceptions, any microcomputer that was incompatible with the PC and MS-DOS standards would be irrelevant.

  I was proud of our team, of course. MS-DOS quickly became the cornerstone of the company’s success, and it was deeply satisfying to have played a central role in its delivery to market. But I was equally proud of 86-BASIC, our old warhorse now running on a new processor chip. While I’d tweaked and improved the vintage code that we’d originally handcrafted for the Altair, it survived mostly intact in the PC. It still had some gas left in the tank, after all.

  AS MICROSOFT GREW to nearly a hundred employees, we knew we had to follow the Silicon Valley model and share some equity to keep our top people. We weren’t yet ready to take on the complications of a public offering, as Apple had in 1980, but it was time to incorporate. We strengthened our board of directors by bringing in David Marquardt, a young venture capitalist who would ease our entrée into the financial markets. In June 1981, we filed our papers with the State of Washington.

  With the pie cut into more pieces, our stakes were slightly diluted. Under the new ownership split, Bill kept 51 percent of the equity and I retained 30 percent. The other stakeholders were Steve Ballmer, 7.8 percent; Marquardt’s Technology Venture Investors, 5.1 percent (for an investment of $1 million); Vern Raburn, 3.5 percent; Gordon Letwin, 1.3 percent; and Charles Simonyi, 1.3 percent. I kept my title as vice president, later amended to executive vice president According to my formal employment agreement, I would receive a base salary of $100,000 as a corporate officer on top of my $60,000 manager’s pay. Bill got $25,000 more as president.

  The incorporation didn’t immediately change anything, but it made our business feel more serious. That fall we moved into a larger space near Lake Washington and Burgermaster, a fast-food favorite. Bill and I took adjacent offices with a shared secretary and a short passageway between us. I could hear his every shouting match, including the battles royal with Steve Ballmer. Steve complemented Bill as a sounding board on business strategy, as I did on technical strategy. Bill remained the big-picture tactician, but Steve made us more disciplined and systematic. The two of them could get adversarial at times, with Steve’s arm-waving histrionics feeding into Bill’s pitiless dissections of what he thought everyone else was doing wrong. They were both ultracompetitive, super-high-IQ, maniacally relentless people with a tendency toward melodrama.

  Over time, their disagreements seemed to get more frequent and intense, like face-offs between bull elephants—especially when Steve tried to push Bill to ramp up hiring, the only way to keep our customers and sustain Microsoft’s growth. Not long after coming on board, he told Bill that Microsoft needed another thirty people right away, doubling our staff. I was all for it, but Bill considered it heretical. He liked to take on overhead slowly and incrementally, which could miss the boat in the tech industry. He started yelling at Steve: “Do you know what you’re doing when you ask for thirty people? Are you trying to bankrupt this company?”

  Steve bellowed back, “We don’t have a choice! We’ve got commitments and delivery dates! If we don’t make these hires, we’ll blow the contracts!”

  And Bill said, “What if the business slows down while we’re paying all these people? We’ll be wiped out! Are you crazy? We could destroy this company! Do you want to destroy us?”

  I gave Steve credit for not backing down; he kept working Bill over until he got what we needed. After an hour of back-and-forth, he said, “That’s OK, Bill, it’s on me, damn it, but we’ve got to get those people in here or we’re screwed.”

  Steve was sincere and straightforward—theatrical, maybe, but not manipulative. We didn’t always agree about the business, but we generally stayed out of each other’s way. Sometimes we’d go on recruitment trips and share a twin room in the frugal Microsoft tradition. One morning I awoke to a series of grunts. I cracked my eyes open to find Steve doing push-ups by the dozen at seven in the morning. I thought, This guy is really hardcore.

  Typically we’d tour the top computer science schools at the best universities: MIT, Cal Tech, Harvard, Yale, Stanford. (I can recall a packed lounge at MIT where students were chanting the actors’ lines in unison during a Star Trek rerun.) Bill thought it was better to get programmers when they were young and enthusiastic, before they were ruined by working somewhere else. After my stint at Honeywell, I couldn’t disagree. We wanted freshly minted bachelor’s degrees, occasionally a master’s, rarely a PhD. Above all, we were after the brightest lights. A great programmer can outproduce an average one by ten to one; with a genius, the ratio might be fifty to one.

  Fortunately, with its water and nearby mountains and urbanized core, Seattle was an easier sell than Albuquerque. And once IBM announced the PC, anyone could see that we were offering unparalleled opportunities. We might pay slightly under the norm, but our pitch was persuasive: Would you rather work on some process-control project for Dow Chemical or a state-of-the-art word processor for the IBM PC? An ambitious young software engineer wouldn’t think twice about joining us.

  Steve’s efforts finally got Microsoft growing in line with its revenues, from 40 employees in 1980 to 128 in 1981 and 220 in 1982. When a company doubles or triples in size each year, it can’t possibly stay tight-knit. But I still hung out with people from work, going for dinner at Casa Lupita or downtime at a pub called the Nowhere, where I built on my Wazzu skill set at foosball. When our six-day workweeks allowed it, there were volleyball games and barbecues at Bob O’Rear’s, with Marc McDonald serving batches of his homemade daiquiris.

  My life was more rounded than before. I bought a small sailboat. I invited musicians to my house on Lake Sammamish, where we’d make a single blues jam last an hour. And I hosted a memorable Halloween party, where I dressed as a wizard and Bill did chest slides on the balustrade from my upper floor down toward the kitchen. He’d run as fast as he could, throw himself on the banister, and glide to the parquet below. He was still edge-walking. One day he borrowed Andy Evans’s Porsche 928 and spun and bottomed the car out, nearly totaling it; the repairs took more than a year. Bill got so many speeding tickets that he had to hire the best traffic attorney in the state to defend him. He finally switched to a sluggish Mercedes turbo diesel just to stay out of trouble.

  In our farewell company photograph back in Albuquerque, nine of the eleven people were programmers, a bunch of young hackers having fun together. That changed in Seattle as we brought in MBAs to support an increasingly lucrative set of product lines. Many were hired for sales and advertising; others handled end-user testing on new features. This was basic business practice, but it inevitably funneled resources away from development. As a technology company grows, it must balance the need for innovation with the imperative to bolster existing products and keep the profits flowing. As Microsoft expanded far, far beyond the thirty-five programmers that once seemed like a pipe dream, it would get more and more diffic
ult to keep that balance.

  CHAPTER 11

  BORROWED TIME

  A week after we completed work on the PC, a form letter came from IBM: “Dear Vendor, you’ve done a fine job. …” Gary Kildall gave us a less favorable review. Some time before the August 1981 rollout, Bill and I met with him at Boeing Field, south of downtown Seattle. He’d heard rumors about our involvement with PCDOS and wanted to feel us out—and to make an appeal for the old spheres of influence. “In a perfect world,” Gary said, “you guys should do languages and we should do operating systems.”

  Digital Research had recently acquired CBASIC, which ran under CP/M-80. As Gary talked, CBASIC sounded like a shot across the bow; if Microsoft came up with a DOS, Kildall would open fire on us in the language business. But we were past the point where a competitor could knock us back on our heels, and certainly not in languages, where we had supreme confidence. At the time, CP/M-80 was running on hundreds of computer models. Its sales amounted to more than $5 million a year. But with 8-bit technology soon to be eclipsed and IBM secured as our 16-bit platform, Digital Research was destined to be an also-ran.

  Kildall did not go gently into the night. Shortly after the PC’s announcement, he claimed that we’d ripped off his software. The charge wouldn’t hold up, however. The commands that Tim Patterson had copied from CP/M-80 were both nonproprietary and common knowledge—the keyword to open a file on a disk, for example, or to send it to the printer. Tim had never seen CP/M’s source code, and he’d built QDOS (and then 86-DOS) in 8086 assembly language, from scratch. His program was a different animal, as anyone who’d used both the 8-bit CP/M-80 and the 16-bit 86-DOS could attest. For starters, Tim’s system could read or write files up to six times faster. MS-DOS had borrowed no more from CP/M-80 than Kildall had taken from DEC’s DOS-11 operating system when he created CP/M years before.