As VP of Marketing at our new startup, the CEO literally threw me out of the building and told me not to return until I understood the market and could identify the key applications and customers for Ardent’s new personal supercomputer. (See the previous Ardent posts for context.)
Supercomputers
With the introduction in 1976 of the Cray-1, supercomputers were defined as the fastest vector-processing computer, one in which a single instruction performed operations on an array rather than a single number. Cray’s first customers were the U.S. Nuclear Weapons Laboratories which used supercompurters to run their hydrodynamics codes to simulate what went on in the first microseconds in a nuclear weapon and the National Security Agency (with Cray putting in a special population count hardware instruction) used to facilitate decryption of codes.
At first only the national laboratories and the largest companies could afford to buy supercomputers, (the Cray-1 cost ~$9 million) but over time scientists and researchers were also starting to use them. Companies wanted to run numerical simulations to model things that were too expensive, too dangerous or too time consuming to physically build. Because of the vagaries of how floating point units in computers were designed, your average IBM mainframe of the day would take forever to run a simulation application. A supercomputer could be a 100x faster.
What Markets?
At Ardent our hypothesis was that if we could build a desktop supercomputer powerful enough to run and display these numerical simulations there were enough customers to make this a big business. My job was to figure out what markets Ardent should target, who were the key customers in these markets and what applications these customers had to have.
The problem was I didn’t have a clue. And while others in our new startup came from companies like Digital Equipment Corporation (DEC) that had sold computers to automate scientific instrumentation and process control, the computers we were building at Ardent were targeted to different customers and markets.
The one thing I did know is that we were probably going to be running some of the same applications as the market leader Cray. I concluded that my first job was to understand what Cray’s markets, customers and applications. When I learned that Cray users would be giving papers at the Society of Petroleum Engineering conference in Denver the next day, I got on a plane to listen and learn.
Follow the Leader
At the conference I attended a bunch of technical sessions, and got lost when the speaker got past, “My name is xxx.” I could see that quite a few oil companies were buying or thinking of buying their own supercomputer. As I walked out of the conference hall, I ran into a small booth with salespeople from Cray. Since their computers were way too large to bring to a trade show, the Cray booth just had literature describing their machines. I grabbed one of each piece of literature, stuffed into my bag, and wandered through the exhibit hall looking at other hardware and petroleum software companies.
Later I sat down for lunch and began to leaf through the bag of data sheets and brochures I had collected. Hmm… typical booth stuff…key chains, data sheets, pens… Until I got to the material from Cray.
Intelligence
As I leafed through the Cray sales material, a glossy magazine with the headline Cray Channels jumped out of the pile. Skimming a few pages, I realized that this particular issue was all about computational fluid dynamics, one of Cray’s key markets. The articles described the applications these users depended on and featured interviews with their most important customers.
I went back and looked at the cover not quite believing what I was reading was real. Cray Channels was describing my market, applications and customers for me. Was it possible that my market research was being handed to me by the existing market incumbent?
I kept thinking, “could this be possible? Did Cray ever publish any more of these magazines?” I looked at the cover of the magazine again and almost fell off my chair. It was Volume 7 issue 2. These magazines had been published for the last seven years. Could Cray have actually been describing their markets and users for that long?
I ran back to the Cray booth and as casually as I could, asked the salesman about the magazine. He assured me that each one profiled a different market, applications and users. I could order back issues from their publications department.
I don’t remember how quickly I got to a payphone, but I’m pretty sure that every back issue of these magazines were on the way to Sunnyvale by the end of that day.
Supercomputer Porn
When the back issues of Cray Channels arrived at Ardent, I ripped open the package with the Cray return address and eagerly started to flip through them. I was excited about what I was going to learn, yet somehow felt guilty, as if I really shouldn’t be looking at them. The pictures were great but I was reading it for the articles. But I was breathing heavy and it felt like I was looking at supercomputer porn. The magazines got passed around to all the engineers until they were dog-eared and worn.
I spent the next few days building a table with three columns: markets, applications, key customers. At the same time I had found the Wall Street analyst who followed Cray (now a public company) who kept a list of where every one of Cray’s machines was installed. I could now cross-correlate the markets by company who used supercomputers.
I started sharing what I had learned about potential target markets with our engineering team and my CEO. We agreed that now we had a roadmap, it was time to hit the road, talk to Cray customers and learn about supercomputer markets and applications in detail.
Lessons learned:
- If you are in an existing market or trying to resgement an existing market you need to understand the market leader
- Market leaders tend to educate the market
- Step one for a startup is know what the market leader knows
