Since a lot of people have already asked or are likely to ask: yes, it’s true. As of today, I’m “enjoying” a status I haven’t had for twenty years: unemployed. SiCortex has suspended operations, let go almost all of its employees, and initiated an asset sale. None of this could be considered much of a secret at this point, since it’s already on insideHPC, the flurry of LinkedIn updates has begun, multiple people have posted to Facebook, and the asset-sale letter went to 300+ parties. A skeleton crew is being kept on to provide support and to facilitate the asset sale, but everyone else including me turned in our keys and received our final checks etc. this afternoon.
I don’t plan to write all that much about this now. I’ll save the detailed post mortem until after the asset sale completes, both to avoid having (or being accused of having) an effect on that sale and because it’s a good amount of time to organize my currently rather disordered thoughts. Maybe I’ll even wait until I have another job, if that takes longer. Right now I’ll just say this: in many of the ways that are important to me as an engineer, SiCortex succeeded. Yeah, that’s right. We set out to do something very difficult and risky – to place a bet on computing that’s characterized by energy efficiency, high density, fast communication and high processor counts instead of raw single-thread performance – and we made it work. We made it work technically, and we made it work in the market. Our 1Q09 sales numbers represented a breakthrough, our 2Q09 numbers were going to smash that record, we were just about to make a big deal that would crack an important new market, and we were on track to become profitable some time next year. As if that’s not enough validation, all of the characteristics I just mentioned are now recognized trends in the industry, and other companies are sure to follow in our footsteps meeting the same challenges that we already did. It’s the old “second mover advantage” come back to haunt me, and my question about its applicability to “dense cluster computing” seems to have been answered.
The only failure that mattered was not technical, nor in any area of customer-oriented execution: it was purely a matter of finance and timing. There is every reason to believe that our next system based on our next chip was going to be awesome, pushing our flagship system well into the Top 500 even before we talk about linking them together, and development was well along. Unfortunately, such development is not cheap and that put us in a high-burn-rate phase right when the economy turned sour and capital became very scarce. That’s like a “perfect storm” combination of circumstances. Maybe there are some things we could have done differently, but “coulda woulda shoulda” is a dangerous game. For example, some might say we should have cut costs sooner, even if it meant delaying the arrival of that next system, and tried to “weather the storm” until a better funding environment came around. Maybe. Or maybe, under other equally-likely conditions, that would have led to a far more ignominious failure as our competitors took the same extra time to adopt our approaches and use superior resources to leave us behind technologically. No thanks. I’m not in any mood to second-guess the people who made those decisions. I’d rather believe, and I think I’m justified in believing, that we played a good game and can walk off the field with our heads held high even though the score wasn’t in our favor.
So, what does all this mean for me? Not much right away. Tomorrow we go on an already-planned vacation, so I’ll simply have my mind elsewhere. Some time after that, I’ll have to start looking for another job. I’d be a fool not to feel a bit anxious about that, but I’d be almost as much of a fool if I worried about it too much. When there’s only one course that makes sense you follow it, and worry or doubt only slow you down. For the blog, this probably means I’ll be posting less, because I don’t want to fall into the easy trap of wasting my new-found spare time online, and more of what I do post is likely to be technical. Partly that’s because I know prospective employers will be checking out the site. Partly it’s because I plan to use some of my time catching up on ideas and maybe even some projects that I haven’t had time for, and in the process I’m likely to find interesting things that I want to share.
To all of my SiCortex colleagues and contacts reading this: it was an honor working with you, and being part of something so significant Thank you, and good luck.
Update: it looks like Matt Reilly (one of SiCortex’s founders) has reanimated his blog so he can offer his own perspective. Like me, he seems to have reached the conclusion that the “SiCortex experiment” was a success in some important ways. That’s good, because I found out a while ago that when Matt disagrees with you it usually means you’re a fool.