Comments on: First GM, now Silicon Graphics. Lessons learned?

The optimist in me wants to believe that even the most raging egos must know that all glory is fleeting. What with Silicon Valley's famous chronic self-absorption, that's not a sure bet.

[-fjtorres-]

SGI's problem is they failed to recognize they were a video card company and not a computer company . So when they got rid of their video hardware engineers and IP (most of which ended up at NVIDIA) and their graphics software staff and ip (most of which went to Microsoft) their fate was sealed. Picking up Cray just as clusters were starting to eat into the supercomputing arena, focusing on servers as the internet bubble peaked and started to burst, focusing on linux when every startup from chineato south africa was doing the same, all simply highlighted the fact that the company's management didn't understand what the heck they were doing.
As is, getting $25 million sounds pretty high; they must have a lot of corporate real estate in their holdings to get that much. :-)

[Dan Owen]

That was a cogent observation that SGI was a graphics card developer. I remember the diagrams of their data flow which look exactly like a graphics card pretending to be a computer.

Still, using them back in the early nineties was 'way cool'!

[dascha1]

so, you feel sorry for SGI? Working for one of its biggest consumers (customer paid millions) in the 90's I remember hearing even their janitors were driving new very expensive sports cars.

As for AG's quote, well, can only say the exception is when folks see you in the accessibility field. Then they think you belong on an isolated island in Hawaii.

[meh130]

The GM comment is interesting. Some may recall the reason SGI acquired Cray Research in the first place was because Cray was going out of business, and the U.S. had no other vector supercomputer producer. Cray going out of business would have left the Japanese with a monopoly on the vector supercomputer market. Enter then President Bill Clinton the U.S. government to encourage SGI to acquire Cray. Shades of the same today with government sponsored deals being done for troubled banks and auto companies. The truth is, SGI did not want Cray. They certainly could benefit from the vector products, but SGI thought its upcoming Origin 2000 NUMA system would be a Cray T3 MPP killer.

Added to the challenge of integrating Cray into SGI was SGI's management at the time, which makes current banking leaders look pale in comparison.

Next was the proprietary Windows Visual Workstation, which was obsolete by the time it shipped. Some would claim it failed because it was designed for Windows 2000, which was a year late.

But the bit one was the brilliant idea to embrace Itanium. First, with a plan to run their IRIX operating system on an Itanium-based Origin NUMA machine, and later the idea to dump IRIX and put Linux on the Itanium NUMA machines. SGI executed on this plan very well. However, it did not pay off as John Mashey and others at SGI had hoped. Some would claim this failure was due to Itanium 2 being years late.

SGI always seemed to mistime everything. But ultimately, when you let the government get into your business, and then decide to depend on the innovation of others for your sucess, you have problems.

[martin1212]

Interesting reference to Clayton Christensen's book. In fact that book (Innovator's Dilemma) was required reading at SGI when I worked there, mandated from the very top. For those who don't know, the book contains several case studies of companies that missed the impact of disruptive technologies on their business and so ultimately failed. What makes the SGI case so interesting is that, contrary to some speculation, the company were well aware of the dangers ahead but was still unable to adapt. It's almost as if once companies get into that situation they are predestined to fail, whether they see the dangers ahead or not. It is just not possible to adjust the business model and market direction quickly enough.

[adam_hartung]

All companies run the risk of SGI when they get too focused on their "core markets" or "core technology." They forget that markets shift, and what was once "core" becomes an anchor. Today all companies better look at SGI and Sun and realize that if they don't focus on markets and develop new solutions they'll become obsolete as well. Read more at http://www.ThePhoenixPrinciple.com

[sam99999999]

Back in the "heyday" you could drive by their main campus and see nearly new office furniture being thrown out daily. Just 'cuz it wasn't "trendy" enough.

That, and tons of half eaten food, shrimp, sandwiches, and drinks. What a profligate waste. Their HR department was a major part of the problem--they thought they ran the place, and fomented politics up the kazoo.

It was like Apple in the bad old 90% gross margin days. SGI was warned, but they learned none of the lessons.

[walrus50]

I started working on SGI graphics workstations in the late 1980s, if memory serves me correctly. We jumped from two DEC PDP-11/34 computers (RSX-11M Op Sys) with and Evans & Sutherland calligraphic image generator to a SGI IRIS 4D/80GT Graphics workstation; 16 bit to 32 bit world; and to a UNIX Op Sys and full shaded images. I was project manager and senior(only) programmer. This was a huge leap for our driving simulator. Those days and the early 1990s were the good days for SGI. After that, it seemed like one long slide downhill, although I left the workplace in 2002, just after an ONYX2 purchase and the purchase of two SGI desktop graphics workstations for support. The purchase of Cray made little sense to many in the industry. The purchase of, then sale of MIPS made little sense as well. I've ben thinking for several months that SGI would not survive this recession. But, they seemed to stop being what they were best at being some time ago: a manufacturer and supplier for very powerful graphics workstations. The best in the business IMO. SUN was absolutely shocked/dumbfounded when we chose an SGI system over the much weaker (graphics capabilities) SUN system. My driving simulator lab stayed with SGI as long as I was there, although we used an excellent Star/GRAPHICON Image Generator for a few years....SGI was the host and support computer. Some visits to SGI campuses, including Orlando, FL, and Mountain View, CA. Very nice, but anything would seem nice compared to the austere, almost prison-like atmosphere of the federal government facility I worked at for 18 years. The one thing that did bother me about SGI was that they delivered graphics systems. That was probably what most people wanted. I needed an Image Generator, something for out-the-window simulation. Their graphics systems tried to be all things to all people. By contrast, the Star/GRAPHICON Image Generator was intended for out-the-window simulations. It was remarkably easy to program because it addressed my needs, rather than everyone's needs. OTOH, perhaps that was an SGI strength.