Today in part II of this SGI focus, our sister publication, Computer Business Review, looks at where SGI’s weak spot lies………. While the threat to SGI’s position may not lie with weaknesses in the MIPS technology, the rising performance capabilities of microprocessors which ship in greater volumes may lure third party software companies away from MIPS to Intel’s Pentium or Merced. For SGI there is a real danger of an eroding software base [as suppliers move] towards Intel and Microsoft platforms, observes Microprocessor Report’s Linley Gwennap. Andrew Feit, an analyst at Dataquest, points out the cost of owning SGI systems. There is little support for industry standard components so you have to buy from SGI which is lot more expensive. So far, SGI customers have been willing to do so but as cheaper, and more standard systems become available, that may change. Underscoring the challenge to SGI’s position, several of the mainstream workstation vendors are now putting advanced graphics capabilities directly onto their processor chip. SGI argues that a separate graphics chip is the way to increase visual computing performance; putting graphics processing on the CPU is a less expensive and less powerful approach, says SGI. McCracken dismisses suggestions that NT or Intel are becoming suitably powerful for SGI’s customers in fields such as computer aided design and data modeling. Such users are primarily drawn by the commodity products costing a fraction of a SGI solution. McCracken has no desire to try to counter-attack by offering NT, even though one of its partners, NEC, offers NT on MIPS-based computers. My feeling is that NT is an Intel market, he says. It’s going to be hard for a non-Intel company to be successful there. Once people shift from innovation [technologies] to NT, they want standard NT. They don’t want NT on something weird like an Alpha chip. So I think it’s a clone market. Sure, we could come out with an NT system, but why would you want to buy it from Silicon Graphics? I don’t think we’re a commodity company. Of all the computer companies, we’re on the opposite end of the scale [from NT workstations], the least commodity market driven company. As far as the low end is concerned, McCracken points out that Sun, rather than SGI, is in the front line in terms of competing with NT/Pentium Pro systems. He believes SGI is distant enough from the action to be caught in the crossfire. We can sit and watch the Sun/NT stuff compete and learn from it, he says. According to IDC, half a million NT/Pentium systems shipped in 1995, but This personal workstation market will continue its rapid growth, primarily in commercial markets and will make only modest inroads against Unix in the technical markets. SGI is still far enough out in front, according to Van Dorsten. PC platforms cannot yet compete with SGI systems in terms of graphics bandwidth, he says, even though it is possible to make the PCI bus more efficient, as workstation vendor Intergraph has demonstrated.
Hollywood heights
There are plenty of emerging markets for SGI which are beyond the reach of the mainstream workstation vendors. Hollywood seems to have an inexhaustible appetite for computing power and SGI kit is increasingly used in digital studio production and for special effects such as the much- publicized dinosaurs in Jurassic Park. Developing new systems for Hollywood is no vanity trip for SGI. As more movie directors turn to digital technology to enact special effects, SGI equipment has become a standard adjunct to Hollywood production. Market watcher, Frost & Sullivan sees major growth opportunities for companies such as SGI with its Alias/Wavefront software in the video and post production hardware and software markets. The firm estimates that the value of this market will rise from $1.66 billion in 1995 to $2.70 billion by 2001. Hollywood is a big money spinner for SGI. The entertainment industry is a market where there’s not much price competition and SGI can continue selling high profit margin systems, says
Van Dorsten. The enormous demands placed on creating high performance digital image production systems has become the new testing ground for SGI technology, while the role of the military in defining the leading edge of visual computing has declined. The benefits are passed on to other SGI customers, says McCracken. The software that designed the dinosaurs in Jurassic Park can be adapted to modeling the shape of cars, he points out. Another growth area for SGI is the Internet/Intranet market where it provides high performance servers which offer good price/performance for larger sites that handle more than 100,000 hits per day. Key customers include Netscape, which has the largest volume of Web site traffic in the world, and online service company America Online which has five million users and is adding new users at 300,000 per month. There is nothing booming as much as multimedia databases and the web. The Web business [at SGI] is producing about $300 million [in revenues] up from nothing a year ago, says McCracken. About two thirds of that business is for private company Intranets. SGI, itself, is a model for Intranet adoption. It has transferred most of its internal administration to its Intranet and about 300 servers host 75,000 Web pages in addition to the 14,000 Web pages available on its public Silicon Surf Web site. Logically, SGI is also trying to apply its expertise in visual computing to bring 3D computing to the Web. It is a strong backer of the Virtual Reality Markup Language (VRML) or ‘Moving Worlds’ standard that has attracted about 50 companies including Netscape and Apple. However, another of SGI’s efforts to tap new markets, its foray into interactive TV (ITV) to provide video-on-demand and home shopping, has gone suspiciously quiet. Among other ventures, SGI has been working with Time Warner since 1993 to trial interactive TV to 4,000 consumer homes in Orlando, Florida. (SGI beat Sun to this contract, which then redirected its software towards the Internet and called it Java.) Whether the lack of progress has been down to technology hitches or a realization that demand for such systems is weaker than first anticipated is still undisclosed. The problem with many of these trials is that they were too small, so that even if they were successful, it would be difficult to extrapolate those results to the general population, says Reza Jafari, vice president of Electronic Data Systems’ (EDS) entertainment and broadcasting division. He added that many large ITV ventures are on hold as the industry tries to determine what kinds of services consumers will pay for.
Acquisition strategy
While Hollywood and the Internet provide textbook new revenue streams for the high-end focused company, SGI is completing its acquisition of supercomputer manufacturer Cray Research in a cash and stock deal valued at $783 million. SGI expects the addition of Cray will increase its revenue base by 20% and give it a 43% dominant share of the $1.9 billion high-end supercomputing market. The lack of strategic planning behind the merger worries some observers. SGI says it is acting as a ‘white knight’ in the deal and that Cray approached it because it was having difficulty in persuading potential customers that it had the finances to continue in business. Cray has been shifting over to massively parallel computing architectures to achieve better price/performance but SGI is inheriting a line of computers which uses a mix of processor types – Cray-designed, Sun SPARC and DEC Alpha – at different levels. Cray’s architecture is a mess, comments Microsoft research consultant Gordon Bell, designer of DEC’s 1980s hit minicomputers PDP/11 and Vax. According to Willy Shah, director of marketing for high performance systems at SGI, Cray will now dispense with all previous processor engines and standardise on MIPS. Cray has suffered badly from the slowdown in demand for supercomputers which has followed the end of the cold war. But Shah sees things picking up in this area. There are new opportunities in the intellig
ence area, in battlefield simulation and applications like modeling nuclear weapons explosions. Others feel that Cray will be a drag on the culture of innovation. Cray may be beneficial to SGI in the long term but I feel that for at least the next couple of years it will be a boat anchor dragging the company down, says Feit at Dataquest. I think that the Cray acquisition poses some formidable challenges, agrees Van Dorsten at Hambrecht & Quist who responded to the acquisition by downgrading his recommendation on SGI stock to a hold from a buy. Cray has a very large field sales force, it’s in Wisconsin and it has been losing money. But as a longer strategic move it might make sense. SGI has limited experience in handling acquisitions. It has only made three major acquisitions in its history, MIPS Technologies in 1992 and digital film production companies Alias Research and Wavefront Technology in 1995. But all three have proved highly successful. As SGI moves into 1996, its challenges are clear. Its low end is being nibbled at by NT/Pentium processor platforms and its high end is facing competition from Sun, HP, IBM and Digital. The key question for SGI is can it continue to drive its graphics technology into new markets, says Van Dorsten. SGI maintains that its customers never say ‘we have enough processing power’ and I believe they can continue finding new markets. But he raises the question: Will those markets be able to continue to pay for the higher performance SGI provides?
The cost of innovation
As long as SGI can continue to afford the high costs of innovation, as long as it can continue to speed ahead of the competition in terms of added value and performance, and as long as fresh market segments continue to provide opportunity, the company is likely to sustain a 20% to 30% growth rate. But staying on that high-end ledge is risky. Although SGI has shown a good track record in managing that risk in the past, as a $2.2 billion company the stakes are now big and getting bigger, and managing those risks is a constant and growing challenge with little room for mistakes.
SGI background
SGI was founded in 1982 by Jim Clark, an assistant professor at Stanford University and six of his students. He spotted a need for a low cost workstation designed for 3-D graphics. After all, he reasoned, the world is three dimensional yet most of the computer systems at the time could only display two dimensions. While at Stanford, Clark and his students developed the idea for the Geometry Engine, a chip designed to handle 3D imaging. I took the concept of the Geometry Engine to all the major computer companies. I spoke with Hewlett-Packard, DEC and others (but) none of them were interested. They said that there was no demand. The founding team went ahead and developed its first workstation, the Iris, and they focused exclusively on the 3D graphics niche with workstations costing $75,000 and up. Two years in, Clark persuaded Hewlett-Packard executive Ed McCracken to join the company. McCracken says that from the beginning, SGI decided to specialize in high performance workstations since Sun, another workstation firm that sprung from Stanford, had already been producing low-priced workstations for two years. We had to develop a strategy which played against Sun’s focus on the lowest common denominator Unix computer, says McCracken. We went after niches and tried to dominate those niches with a 60% to 80% share, and with the best software, the best visualization. But staying in the lead also meant it had to cut the prices of its workstations while constantly adding value. This led SGI to search for more powerful processors and it pioneered the use of RISC microprocessors in 1987 with a line system based on microprocessor from MIPS Technologies, a company it went on to acquire for $406 million in 1992. With its high-end niche focus, SGI’s share of the total workstation market has never been large. It was not until 1990 that its share rose above 5% and it now stands only a couple of
percentage points above that, says Dataquest.
