For those that know me, I’m kind of a “life is too short” kind of guy. I liked to fail fast, move on, and not look back.
However, in catching up with the VP of Sales of Ardent last night, I was reminded one of the few times I did return for closure.
National Supercomputer Centers
For a decade starting in 1985, the National Science Foundation (NSF) established and spent a pile of money (~$50 million/year) on four supercomputing centers in the U.S. – Cornell University; University of Illinois, Urbana-Champaign; the Pittsburgh Supercomputing Center at Carnegie Mellon University; and the San Diego Supercomputer Center at the University of California at San Diego. The ostensible goal of these centers was to allow scientists and researchers access to supercomputers to simulate commercial phenomena that were too expensive, too dangerous or too time consuming to physically build.
The reality was that the U.S. Nuclear Weapons Laboratories used supercomputers to run their hydrodynamics codes for nuclear weapon design and the National Security Agency used them to decrypt codes. But with the cold-war winding down these agencies could no longer be counted on to provide Cray Research with enough business to sustain the company. Commercial applications needed to be found that could take advantage of this class of computers.
The search for commercial supercomputer applications was good news for Ardent, as this was our business as well. But bad news was that the supercomputing centers had concluded that they could justify their existence (and budget) only by buying the biggest and most expensive supercomputers Cray Research made.
We Lost the Deal
At Ardent we were building a personal supercomputer powerful enough to run and display numerical simulations just about the time the National Science Foundation was funding these centers. I remember that the Pittsburgh Supercomputing Center had put out a request for a proposal for a supercomputer to replace the Cray X-MP they installed in 1986. In reading it, there was no doubt that it was written only in a way that Cray could respond.
I realized that given the amount of money the Supercomputing Center wanted to spend on buying the new Cray Y-MP (list price $35 million,) we could put an Ardent personal supercomputer next to every scientist and researcher connected to the university. I responded to their RFP by proposing that Ardent build the Pittsburgh Supercomputing Center a distributed supercomputing environment with hundreds of Ardent personal supercomputers rather than a monolithic Cray supercomputer.
As one could imagine this was the last thing the supercomputer center management wanted to hear. All their peers were buying Cray’s, and they wanted one as well. We had support from the scientists and researchers who had bought one of our machines and were beginning to see that distributed computing would ultimately triumph, but bureaucracy marched on, and we lost the bid.
In my career I’ve been involved with lots of sales deals, and for some reason losing this was the one deal I never forgot. Maybe because a win here would have meant success rather than failure for the company, perhaps because I really believed we could make the impossible happen and win. For whatever reason, I hated that particular Cray that got installed in Pittsburg.
Fast forward 15 years. Retired for a year, I ran across an article that said, “$35 Million Dollar Supercomputer For Sale for Scrap.” It was the Pittsburgh Supercomputing Center Cray Y-MP that had beaten me at Ardent. It was for sale on Ebay.
It took two semi-trailers to deliver it.
It sat in my barn next to the tractors and manure for five years. I had the only farm capable of nuclear weapons design.
Cray called two years ago and bought it back for parts for an unnamed customer still running one.