Compaq’s Most Unexpected Decision
SHORTLY AFTER WE TOOK CARE of the immediate priorities that resulted from the PS/2 announcement, we began to look at our long-term alternatives. There were two obvious choices.
First, we could stay with the industry standard, continue to deliver performance advantages over the PS/2, and hope for a split market with two standards. When the PS/2 was announced, we had immediately and strongly stated our continued support for the existing standard, arguing that there was no need for 32-bit slots in the near term. This was the path most consistent with our public position. And our sales had continued to surge, despite IBM proclaiming its Micro Channel as the way of the future. But since technology was advancing so rapidly, we didn’t know how long it would be before there was a real need for a 32-bit slot.
Second, we could reverse engineer the Micro Channel and gradually switch to delivering products compatible with the PS/2. I had consistently taken the public position that we would deliver Micro Channel products if and when the market demanded them. We had carefully kept this door open, and most industry observers believed we would eventually take this path. We knew it would be a lot more difficult this time, and we would still probably have to buy a license from IBM and pay it royalties. Most of our competitors were doing that. But to us, taking this route looked like a miserable existence, with IBM calling all the shots and allowing us no real future—certainly not as a technology leader. Even so, we kept this option viable.
We had expected IBM to introduce a new 32-bit slot but keep the 8- and 16-bit slots compatible with the existing standard. When IBM not only introduced a new 32-bit slot, but also eliminated the compatible 8- and 16-bit slots, we were amazed. Even though IBM said publicly this was necessary to achieve additional performance, we couldn’t find any reasonable explanation, other than to reduce or eliminate competition from the clones. In doing so, IBM exposed an unexpected weakness we exploited in ongoing publicity battles. More important, IBM had opened up the possibility of Compaq doing “what IBM should have done.”
A number of our competitors were suggesting that the industry develop a specification for an advanced 32-bit bus. Such an approach would offer the clear advantage of not having to pay IBM royalties for the use of its Micro Channel. Zenith had even proposed a formal Institute of Electrical and Electronics Engineers (IEEE) committee to develop a specification, and some work had been done on it. But we strongly believed such an approach wasn’t viable. It would take far too long to get all the companies to agree, and almost certainly end in a compromise that would not be capable of winning the inevitable performance comparisons with Micro Channel.
As we looked down the road to a time when we would need a 32-bit slot on our PCs, the attractive choice was to create our own design. We believed we were the only company that knew how to achieve the advanced 32-bit performance needed to beat Micro Channel and at the same time maintain complete backward compatibility with the existing standard.
The situation wasn’t quite that simple, however. Timing was a critical issue. If we waited too long before introducing our 32-bit slot, there was a great risk the Micro Channel would become so entrenched it couldn’t be displaced. If that happened, we would be forced to adopt IBM’s 32-bit bus late in the game, placing us at a significant disadvantage to our competitors who had made the move earlier. Whatever direction we decided to go, we needed to do it sooner rather than later.
Then there was the question, “Whose technology is best?” IBM had introduced some powerful new capabilities in its Micro Channel, at least on paper, so our 32-bit bus would have to be carefully designed to beat it.
All issues, however, paled in comparison to the really big one: How could we possibly come out on top going up against IBM? Sure, we had built a great reputation and were widely viewed as a technology leader. But Compaq was no IBM—not even close. Even if our 32-bit bus was better and faster, and even if the companion 8- and 16-bit slots were backward compatible, we had almost no chance to beat IBM in a head-to-head battle of this significance. Every time we even thought about going up against them, we could already hear the analysts saying, “Compaq’s bus makes more sense, but I’ll bet on IBM.”
At this point, in the early fall of 1987, we hit a low point in our confidence. Our excitement over the likelihood that our annual sales would exceed $1 billion was severely tempered by a growing realization that we would probably have to switch to Micro Channel-based products. That prospect was a real downer.
Then one day in a product strategy meeting, someone asked the right question.
OCTOBER 28, 1987, 10:00 A.M.
Our product strategy team gathers in the small conference room next to my office on the eighth floor of CCA-4, one of the new buildings on our headquarters campus. Present are Mike Swavely, Gary Stimac, Hugh Barnes, Steve Flannigan, Jim Harris, and myself, along with several others.
I open the meeting. “Let’s get started. I’m having a real problem accepting the idea that we can’t come up with a way to avoid having to switch to Micro Channel–based products. But I also know that we can’t risk a direct confrontation with IBM over who has the best bus. So we need to get our heads clear and think out of the box.”
I pause for a moment. With frustration sounding in my voice, I ask, “Is there any possible way the Micro Channel can be beat?”
Almost without thinking Swavely blurts out, “Sure, if the whole industry supported a better alternative. But they would have to do it pretty soon. Time is running out.” He stops and thinks about what he just said, then adds, “I don’t see any other possibility. The question is how to get the industry to move fast enough.”
“OK,” I say. “Let’s focus on that for a minute. Think. Is there any possible way to get the industry to agree on a new bus quickly? What could we do to speed up the process?”
No one speaks for a while. Finally Barnes says, “One way to avoid the slowness of a committee process would be to design the bus ourselves. We’re the only company with the technology to make sure it’s completely backward compatible. And I think Paul Culley may be the best hardware architect in the industry, so he would be the right guy to do the design. Especially the new 32-bit part. If we could show our competitors a working bus with the right capabilities, it’d remove a lot of the risk.”
Then Jim says, “One thing we’ve got going for us is that none of the others want to follow the Micro Channel either. But so far no one has come up with a viable alternative. If we could convince them our bus design would work, I think they’d all want to join in.”
Swavely jumps in. “I’m sorry to play the devil’s advocate, but won’t everyone be afraid of becoming dependent on Compaq just like they are on IBM? We’re their second largest competitor.”
Stimac, starting to see some potential, responds, “Absolutely. So we’d have to set it up so they don’t become dependent on us. We’d have to give up control of the design to the whole group, or to someone independent, like Intel. After the initial design is complete, of course.”
Flannigan adds, “I think Microsoft would really get behind this. Bill [Gates] is very concerned about IBM gaining so much control. Really, all the software companies feel that way, so we could probably line up their support too.”
Barnes is thinking about the design. “To have the same kind of capabilities as the Micro Channel will require several ASICs. Do we let everyone develop their own chips, or do we give those to them as well?”
I reply, “It might work either way, but it sure is a stronger offer if we take it all the way through chip development and let everyone buy chips from an independent supplier. And it speeds up the whole process. It should make their decision easier if we remove the cost and timing risk of chip development.”
Everyone thinks about it for a moment. Suddenly my face lights up. “What if we do this behind the scenes?” I ask. “Imagine a press conference with all the leading PC companies except IBM standing together and announcing support for a 32-bit bus that is what Micro Channel should have been. That would be powerful.”
Swavely says excitedly, “Now we’ve got something that could really work. That would be amazing.” Swavely didn’t get excited very often.
The brainstorming continues to gain momentum and goes on for another hour. Everyone is getting more and more pumped up as one objection after another is addressed. No one had entered the meeting with any inkling—even any hope—that something like this could emerge. But there it is, another completely unexpected decision that is solidifying in the moment. “The Process”—the way we turn critical decisions over and over until we find a clear path—is truly one of our secret weapons.
I summarize what we’ve come up with. “Compaq will create the specification for an advanced 32-bit bus and slot. Then our engineers will design the bus controller circuits and develop the ASIC chips that do the complex processing while maintaining compatibility with the existing 8- and 16-bit buses. We’ll get Microsoft’s support for our plan and to help us sell our competitors on the idea. They’ll also develop the operating system software needed to handle the advanced bus. We’ll allow a third party, probably Intel, to build the ASIC chips and sell them to our competitors on the same terms and availability that we receive. We’ll not get any compensation or royalties on the chips that Intel sells.
“We’ll visit the leading application software and add-in board companies that’ll benefit from the capabilities of the advanced bus first and bring them in to support the plan too. I’m thinking of networking hardware and software, work group applications, file server applications, and advanced workstation displays and software.
“And we’ll do this while keeping it a secret for as long as possible. Then, when it’s all ready, we’ll organize a major industry announcement where we’ll introduce the new bus and the industry coalition that supports it. We’ll all be viewed as equals.”
As the team listens to the summary, the magnitude and significance of what we’ve come up with begins to sink in. We’re planning to do what no other industry leader has ever done before: spend tens of millions of our own money to develop what could become the most advanced and valuable technology in the industry—and then give it away. Yet there’s no doubt in my mind this is the only path that makes sense for Compaq. It’s the only one that’ll enable us to continue to be an industry leader far into the future—that is, if it works.
All revolutions start with broad discontent, but it always takes a spark to set them off. This meeting was the spark that started the revolt to save the PC open industry standard from becoming closed and controlled by IBM. The oddsmakers might not have viewed our chances very favorably at this point, but all the ingredients for success were there.
But first, we needed to be sure we could design a bus that would deliver what we were trying to achieve. The development project would require the efforts of many of our best engineers and programmers, but the first step was to design the basic architecture. We asked Culley to design an advanced 32-bit bus with features comparable to, or better than, those of the Micro Channel that would also be completely backward compatible with the industry standard. I wasn’t surprised to find out he had already been thinking about the idea and was eager to move forward. This would be the most complex design he had ever tackled, and it was the most important.
Several weeks went by without a word from Culley. Then one day he walked into Barnes’ office and said he had come up with a way to do it. After briefly discussing the design, Barnes leapt into action and put together a team to build a “proof of concept” to test the idea.
It was late December before the engineering team reached a point where they were sure. Barnes and Jim met with them to test their conclusions, and in the end they all agreed. Compaq could do it.
Next, we faced the question of whether we should continue to reverse engineer the Micro Channel as a fallback position. Should we continue until we were completely sure we could get enough of our competitors to sign up to support our advanced bus? There was a sense we were running out of time. It had already been nine months since the introduction of the PS/2, and would take at least six more to get the design ready and everything lined up for a major announcement. To move as quickly as possible, we had to shift a significant number of engineers off the Micro Channel project. It was time to put up or shut up.
JANUARY 5, 1988, 10:00 A.M.
Almost 300 people gather in a large meeting room at Compaq’s headquarters in northwest Houston. Hugh Barnes, head of Compaq’s desktop product development, has called together all the development personnel for an important announcement.
Barnes opens the meeting. “Thanks for coming. As all of you know, since last May we’ve been splitting our resources between new ISA product development and Micro Channel product development. The effect of this is that we’re not able to move forward in either area as fast as we need to. We’ve decided that we can no longer afford to keep the resources split, so we have to pick one or the other.”
Barnes pauses. “I want to make it clear that everyone in this room is critical to future product development. No one will lose their job.”
He pauses again, and the room is so quiet you can hear a pin drop. The Micro Channel development team has made great progress toward creating a product that Compaq can take to market, but everyone knows if we end up following IBM down that path, we’ll no longer remain the technology leader we’ve become.
When he speaks again, everyone is totally focused on his words. “After carefully considering the situation, Compaq’s management has decided to shut down the Micro Channel development and focus all our resources on advancing the performance of our industry-standard products, including developing an advanced 32-bit bus that’s compatible with the industry standard.”
This time, there’s a loud—almost deafening—cheer from the audience. This is the path they were hoping for. This is the path their hearts are really in.
When I heard from Barnes about the reaction of our development teams, I was really sorry I hadn’t been there to share in their jubilation. We all knew there would be unexpected problems, but now we were fully committed to do whatever it took to make an advanced 32-bit bus succeed.
As soon as the proof of concept was complete, the rest of the project team immediately went into action. Swavely made a list of the PC companies to recruit for the alliance. We decided we needed two tiers in the alliance, because we wanted a manageable number of companies working together to make decisions. At the top of the list was Hewlett-Packard (HP). The list soon grew to over twenty companies.
Then we set up a meeting with Bill Gates at Microsoft to bring him in on the idea and to get his support. We had some concern that Gates might worry about damaging his relationship with IBM, but when Stimac, Swavely, and I visited Microsoft and laid out the plan, Gates didn’t hesitate. He wanted this to work and would give it his full support, he said, including updates to Microsoft’s operating systems. He would figure out how to deal with IBM later. When Stimac told him that we wanted to recruit a number of PC companies to join us, Gates said he had relationships with all of them and would get us in touch with the right person at each company. He would do everything he could to help us sell the idea.
Next, we set up a meeting with Intel to get its support. We were moving so quickly Barnes spent most of the night before the presentation in his hotel room, creating hand-drawn slides. As usual, Intel’s executives were a little skeptical at first, but it quickly became clear to them that Compaq was going to make this happen.
We offered them the opportunity to be the supplier of ASIC chips for the advanced bus. They were interested, but a lot of contractual details had to be worked out before they were fully on board. They also had a relationship with IBM they needed to be careful not to damage. Before long, Intel was fully engaged and supportive. The company had become very close to us during the final stages of the 386’s development and had seen us orchestrate the announcement of the first 386 PC. Its executives knew that if anyone could pull this off, it was Compaq.
We believed that our PC competitors would be the hardest to sell on the idea. We expected this group would be the most skeptical of both our probability for success and our real motive behind the plan. So we got everything lined up as perfectly as possible before we began to call them. The list of things to prepare was very long.
We needed to show very clearly how the advanced bus worked and how it compared with Micro Channel. It was also important to show the marketing aspects of the plan, which was a big part of convincing the competing companies the plan would work.
Our presentations had to be extremely professional and thorough; once we started talking to our competitors, we wanted to sign them up quickly. We also wanted enough of them signed to reach critical mass before IBM had a chance to hear about what was happening and mount some kind of interference. This might have been overkill on our part, but it was indicative of how important the success of this project was.
The first company Stimac and Swavely visited was HP. HP was very important because of its market position and technical capabilities. Even though we were prepared to do all the work ourselves, we could use some help from its resources. We also believed HP joining in would help sell the idea to the other companies.
In April 1988, Bill Gates, as he had offered, called Robert Puette, general manager of HP’s PC division, and strongly suggested he listen carefully to what Compaq had to say. The meeting was set for early May, when the right people from HP could be there.
At the meeting, Stimac started the presentation with technical aspects of our proposal. Then Swavely followed with how he viewed the positioning of the advanced bus, the coalition, and the effect it would have. Many questions were asked throughout the presentation, but Stimac and Swavely were well prepared with solid answers. At the end of about two hours, the HP team was very interested and very impressed. But this was a big decision for them, as it had been for us. They needed time to consider it all.
When Stimac and Swavely got back to Houston, they met with the strategy team to discuss the results of the HP meeting and consider the timing of our next move. We concluded the odds were good HP would join us, so we decided to wait awhile to give its executives a chance to make their decision before meeting with the next company.
Over the next three weeks, there were many phone calls back and forth between Compaq and HP engineers, working out details and building a bond between the two teams. Finally, a call was set up between Robert Puette and me. Puette told me that HP was in, but he wanted to hear me commit that Compaq was going to make this a true industry coalition and did not plan on controlling the expanded bus. He had no way of knowing how strongly I was committed to doing just that. I replied, “Absolutely!”
In late June, Stimac and Swavely began to schedule meetings with other computer companies as fast as possible. With Compaq, HP, Microsoft, and Intel committed to the coalition, we were now confident the others would be relatively easy to convince that the “bus” was about to leave the station and they needed to be on it.
As it turned out, the other companies weren’t quite that easy to convince. Although the presentation was compelling, hearing the idea for the first time was somewhat shocking. All the companies needed time to absorb what they heard and then talk to their decision makers. Some of them had already made strong commitments to support the Micro Channel and had to think through how this new coalition would affect those commitments.
Dell Computer, for example, was touting the fact that it had hired the IBM engineer who had led the design of Micro Channel to lead its own efforts. Despite our assurances, some companies were concerned that Compaq would end up dominating this area, similar to the way IBM was dominating the Micro Channel. Many of the companies required more conversations to clarify various points.
Decisions to join the coalition trickled in slowly through July and early August. We had set the announcement date for September 13, 1988, and I began to worry we wouldn’t have enough support to be credible. We decided on a cut-off date for companies to be included in the press conference, as a way to nudge those hanging back to go ahead and join us. We ended up with nine companies large enough to be part of the founding group. In addition to Compaq and HP, they were: AST Research; Epson America; NEC; Olivetti; Tandy; Wyse; and Zenith. We would become known as the “Gang of Nine.”
One of the many issues that had to be addressed was the name of our bus. The name needed to convey advanced capabilities but, even more important, compatibility with the existing industry standard. The name we decided on was “Extended Industry Standard Architecture,” or EISA. The acronym was short and easy to pronounce, and the four words communicated exactly what EISA was all about.
Everything was falling into place as the day of the announcement drew near. When word of the press conference began to spread, many of the computer companies that had been on the fence, including AT&T and Dell, agreed to join us. They were too late to be included in the press conference panel, but they wanted to be counted in the alliance. A week before the announcement, there were more than 60 companies supporting the alliance; by the day of the announcement, the number had grown to 80. Every company in the industry except IBM wanted this coalition to succeed, and a strong feeling had developed that it would.
SEPTEMBER 9, 1988, 6:00 P.M.
It’s the Friday before the press conference and I’m on the West Coast finishing meetings scheduled around a convention I am attending. I’m in a discussion at a San Francisco hotel going over some of the final details for the press conference when I’m handed a note. Bill Lowe, president of IBM’s Entry Systems Division, is on the phone. I had placed a call to Lowe’s office a few days ago, but didn’t expect him to return the call. Without mentioning the caller’s name, I look around the room and say, “I need to take this call.” I ask that the call be put through to a phone in the hallway outside the meeting room.
When the phone rings, I pick it up. “Hello.”
“Hello, Rod. This is Bill Lowe.”
“Hi, Bill. Thanks for returning my call. I’m sure you’ve heard that there’s an announcement next week.” I pause briefly, trying to decide how much to tell him. “A number of PC companies are joining together to support an advanced 32-bit bus that’s compatible with the existing standard. It’s open to all companies that want to join in, so I wanted you to have the opportunity to do so.”
Lowe replies, “I understand. IBM isn’t interested in joining, but I want you to know that we’d welcome you as a licensee of the Micro Channel. We’d be willing to work with you to help you get to a Micro Channel product quickly.”
For a moment, I don’t reply. I’m dumbfounded. I’d never expected Lowe to turn the offer around. Nine months ago we couldn’t have passed this up, but now we’re on a much better road. I almost say, “Your timing is a little off.” But instead I tell him, “Well, thanks for the offer. We’ll think about it.”
Lowe responds, “OK. Let me know what you decide. And with regard to your announcement, I think you’re making a big mistake.”
I resist the temptation to argue the point.
After I hang up, I stand there for a moment, wondering if Lowe is serious or just taunting me. At this point, I wouldn’t have changed to the Micro Channel if Lowe offered the license free with no royalties. EISA is the right thing for our customers, and it will significantly outperform Micro Channel.