A year to the day after accepting a promotion to chairman and CEO of then-beleaguered Silicon Graphics in Mountain View Calif., Robert Bishop was at SGI Federal's offices, where he talked with Computerworld reporter Sami Lais about the high-end graphics hardware maker's struggle to refocus and rebuild, and what he plans for the company's future.
CW: What made you accept the CEO position last year when SGI was in trouble?
Bishop: In August of last year, we were in crisis mode. There were a lot of bad feelings surrounding the company. Morale was in tatters. And the press was happily carving us up. The financial press took us apart completely. Our customers were worried. I'd been with SGI since June 1985, so I had a lot of blood in [the company]. I couldn't tolerate the idea it wouldn't survive.
CW: What did you do first?
Bishop: My first act was to reassure our customers. What really held the fort, what gave us the luxury of making the transformation was [SGI's] customer support and professional services [divisions]. We couldn't have transformed ourselves without them.
CW: What was involved in the transformation?
Bishop: We turned down some business that wasn't right for us, and by that I mean ERP [enterprise, resource planning], EDP [electronic data-processing], electronic commerce. We said, let's get the hell out of the money-counting business, and let's stick to computing associated with the company's core business -- what it builds, what it designs.
How to take orders -- billing -- those are all contextual matters. I look at it like this: An automobile company's job is to make automobiles. A chemical pharmaceutical's business is to design new drugs. An aeronautics company's job is to make aircraft. A film company makes films.
Our job -- where we think we can bring something valuable to those companies -- is by concentrating on doing what we do best, and letting the contextual matters be outsourced.
CW: You divested some divisions?
Bishop: Yes, there were four steps in our divestiture.
First there was Cray, although we kept 800 engineers. But we just couldn't marry the Cray model with parallel computing. [Cray Research Inc. a mainframe company, was sold in the past year.] Then there was all the MIPS. [SGI had a 65 percent interest in MIPS Technologies Inc., and sold its embedded technology division.] We sold two software companies: Kasenna Inc., which had a highly successful media-streaming product, MediaBase. It needed two years to get into full development, and it depended on money from venture capitalists. Plus, it would have had to be ported to our platform.
The other company was a content creation company in Japan, Silicon Studio KK. And that just didn't fit with our mission.
CW: You also restructured management and appointed three new vice presidents in the past month.
Bishop: Executive vice presidents. They form my three-man management team. Ken Colman is sales, service and marketing. Warren Pratt is engineering and manufacturing operations and Harold Covert is CFO and chief administrative officer. Harold was involved in the turnaround at Adobe, so he knows the road map of a turn. We recruited him from his most recent position at Red Hat. I think we're the only company with a Linux-educated CFO.
CW: So where are you putting your efforts?
Bishop: Well, we only have $2.5 billion of a $250 billion market, so we have room to expand. We're close to the inner cycles of our customers. They're looking for better performance every day. That's where we can deliver.
CW: What direction are you taking in research and development?
Bishop: We're returning to a focus that somewhat parallels our origins. We had drifted into a computing space that was not very successful. Our engineers began to design for that [original] kind of customer.
To deal effectively with our customers, we have to move at the speed of thought. If we can't keep up, we can't help. We have to create a system that has the speed to let them visualize their thoughts, that lets designers ask what-if questions and see the results instantly.
CW: Who are SGI's customers?
Bishop: Well, Disney [Studios] has thousands of our systems, which they used in making [the movie] "Dinosaurs."
George Lucas's studio [Industrial Light & Magic] also has our machines, and used them to do special effects for "The Perfect Storm," which is the perfect use of SGI technology.
The new Hayden Planetarium, which opened in February in New York City, runs its Space Theater -- a 429-seat theater under a huge dome with a 3-D model of the Milky Way -- on an SGI [Onyx2 workstation with seven subsystems].
NASA is using SGI supercomputers to create a digital galaxy, a database of a billion stars. When they were telling us what they wanted to do, they said they wanted to be able to see how things worked faster than the speed of light. We worked with them to develop a way to do that. We're building them a 1,024-processor supercomputer.
CW: What are they asking for?
Bishop: Power, speed, flexibility and more of it. Response to the Origin 2000 has been enormous, but we're looking beyond NUMA [SGI's component-based supercomputer architecture].
We're continuing to develop on our MIPS/Irix platform where we can control development, where we can move at our own speed, rather than on the Linux/Intel platform where we have no control. We can't dictate to Intel, "Hey, get that A-64 chip out now."
But we'll transfer as much of our MIPS/Irix technology as is practical to the Linux/Intel platform. There are certain economics involved in that lower cost of entry. We want to capture those users so as they grow, as their mission-critical applications scale up, we'll scale with them.
When you get to a certain level, Unix dominates that space. Linux really hasn't made inroads at that level. Unix still dominates in reliability and scaleability.
The second dimension is that to solve the truly difficult problems, the best solution is shared memory, single-image systems. I believe clustering the IA-32 chip with Linux cannot produce that quality.
CW: When I spoke with SGI about 18 months ago, future development in the MIPS line was up in the air. Is it still?
Bishop: We are more than 100 percent committed to the future of MIPS/Irix and to development on that platform.
We've got a road map, first for a 400-MHz machine, then for a 1-GHz and on to, I don't know, a THz machine, all MIPS-based. The value of such a product lies in its scaleability and its capability to handle the big computing jobs. No one can match us on that. At the most, Linux/Intel may be 50 percent of our business in maybe five years.
CW: And Linux on MIPS?
We'll be putting Linux only on the Intel platform, not on MIPS.
CW: Typical SGI customers are using large amounts of data and working with large files. They've got to have huge and maybe unique storage needs. Disney Studios developed its own storage area network running over HIPPI [High-performance parallel interface] for its "Dinosaur" movie, which was entirely digital.
Bishop: Yes, I know. The "Dinosaur" movie was done entirely on SGI [machines]. We'll be marketing [the Disney SAN] solution.
CW: Will SGI be marketing the HIPPI SAN Disney developed?
Bishop: Yes, it's part of our storage strategy, which we announced this month. Storage and backup are terribly important to our customers, making it terribly important for us to provide SAN capabilities.
CW: You've talked about the past year. Where will SGI's thrust be in the coming year?
Bishop: Growth. And staying focused, staying with what we like to call the fun stuff. I'm dissatisfied with our stock price and with our market cap. I intend to bring that up. I think we can deliver the numbers Wall Street wants.