Analysis Got $18m to spare? That's the market capitalization of one of Silicon Valley's most glamorous companies this morning, after Silicon Graphics Inc. filed for Chapter 11 bankruptcy.
The size of SGI's debt - at $664m it's twice the value of its assets - is enough to deter all but the most determined bargain hunter. Apart from a ragbag of trademarks - such as OpenGL - what growth has SGI left to offer?
SGI makes a particularly painful case study for business graduates, because the company's decline was so widely heralded.
"Everyone in management had read
The Innovator's Dilemma," one former staffer said Monday, "but no one knew how to execute on it."
SGI's particular dilemma was finding new revenue to replace its high margin systems that were being commoditized by cheap PCs running Windows NT or Linux, relentless increases in the of power of commodity 3D graphics cards, and ever faster interconnects. At its peak SGI turned over $4bn a year, and employed around 10,000 people.
The company was additionally blessed by having customers across the economy. Two thirds of SGI's business was split evenly across two cyclical sectors - defense and manufacturing - but as manufacturing spending along came another splurge in defense-related spending. (Contrary to popular belief, SGI's Hollywood business never accounted for more than a fifth of its revenue - but it got most of the headlines).
When in their prime, companies have several options to prepare for the day when their cash cow dries up - including putting the company up for sale. SGI tried them all, some several times over, but was unable to find a revenue stream that kept pace with its expenses.
SGI bought into new businesses, such as RISC pioneer MIPS - SGI at the time was one of its biggest customers - and Cray Research, in a bid to widen its product range. Neither was successful. MIPS customer base fled as the vendors doubted they could gain access to the SGI was getting preferential treatment. MIPS was later spun out, finding a successful application of its technology in low power embedded devices such as phones. Cray's core business was also in decline, which SGI's mounting problems didn't help. SGI decided to spin off Cray after only three years, and it now trades as a public company. Notwithstanding recent financial troubles, the size of the supercomputer company's revenue last year indicates that Cray was never likely to be SGI's savior: $201m, up from $148.9m in 2004, but down from $237m in 2003.
After years of being dismissive about a rival computer platform with no serious graphics, SGI eventually tried to meet the challenge of the PC head-on. At first it produced a Visual Workstation PC, with custom ASICs providing a new proprietary interconnect and graphics chip, and running a tailored version of Windows NT. But SGI found it hard to market the costly new kit, which also ran into manufacturing delays. The Visual Workstations failed to win over SGI's core customers, who were using 64bit IRIX, and were too expensive to find lure customers. The effort was canned almost as soon as the workstations - and a complementary Windows server - were launched.
Thirdly, SGI's rich patent portfolio may have allowed it to create a lucrative licensing business. But patents need to be aggressively defended. In spring 1998 SGI announced it was suing NVidia for infringement of its texture mapping patent. However, the case never reached trial, and SGI's lawyers were left fuming as the company caved in only 15 months later, agreeing to license Nvidia's patents.
All three decisions - to spin out Cray, axe the Windows workstations, and abandon the NVidia lawsuit - were taken by former HP executive Rick Belluzzo. Who then promptly jumped ship to join Microsoft. The move raised eyebrows across the industry, as Belluzzo was leaving the chief executive's role at a prestigious Silicon Valley name to head up what we
described at the time as "pretty much a suicide mission" - MSN, the small and ailing internet division at Microsoft. It isn't much bigger, and continues to make a loss today.
After the Nvidia climbdown, it was never likely that SGI could enter the IP licensing business. The plot thickens, however. In early 2002 we broke the news that SGI was transferring many valuable graphics patents
to Microsoft, for the sum of $62.5m. SGI denied the deal at the time, but it wasn't long before it was
proved to be correct.
Other factors didn't help SGI, such as its decision to end of life its own MIPS processor in favor of Itanium, but in large part the woes of recent years were settled by these decisions taken in summer 1999. After that, with each fresh quarter bringing news of losses and layoffs, SGI was on the back foot.
Belluzzo's spell at Redmond was brief: in spring 2002 he announced he was leaving to start his own company. Today, he's CEO of storage vendor Quantum.
The REGister
Linear Mode
