Chat with Andy Keane

So first of all, what are going to be your main revenue sources in the short-term?

Through this year, the vast majority of the revenue will be coming from the 1U product. Primarily, there's a lot of demand from quite a lot of people. That's mostly three markets: oil & gas, development boards, and then third is what I would call education and government. Especially universities.

Universities mostly take 1Us too? No deskside computing systems at all?

The majority's 1U. You know, what we found in the type of workstation that people have… through time, as they've got smaller and smaller, rather than computing devices they've really become more focused on visualization with a lot of memory and so on. They can't really have a Fastra kind of thing next to them. So they don't have the kind of machine they want for computing, but they've got space in the data center [so that's where the Teslas go].

One thing that's interesting with Fastra is that it's based on GeForce, not Tesla. So the amount of revenue for you is much smaller.

Yeah, that's one of the decisions you have to make. If we want a very large developer base, or do you want to [make more money out of a given user base]? We chose the latter. So if someone develops something great on a GeForce, [that'll earn me money down the road and everyone comes out ahead].

So the way you look at it is that you'll have maybe 20% of your potential customers using GeForce, but the vast majority will still use Teslas because they can't afford anything to go wrong for themselves or especially their customers?

Right, exactly.

Okay, the next question I have is on the oil & gas industry. Since Rys & Tim talked to you one year ago, the price of oil doubled, do you expect that to help you there? Since I would assume the budgets for R&D to go up.

Despite the change in prices, […] the cost was never really a consideration because if you don't invest there, you won't find the oil [so everyone invests as much as it makes sense to anyway].

But now the price that they can pay for developing an oil well has gone up.

For a new well, the economics didn't change [for us]. For an old well,[they might want to pump more of it] and that's where the change in oil prices helps us. Because, you know, you have to inject carbon dioxide and so on, it's a lot of expensive things [that need to be simulated].

So it's not really discovery, just maintenance? Do you think that business will be as big as discovery?

Exactly, and I don't really know.

Yeah, obviously Tesla is very hard to forecast, and when it comes to Oil & Gas I don't know what revenue magnitude you're expecting. All of the examples that have been published for GPU Computing had a really small number of GPUs, like with Exxon demonstrating a cluster with 40 Teslas. For a company like NVIDIA, that's nothing. So how does it scale?

Generally, it's a pilot/research cluster, it can be anywhere from 20 to 100 nodes, and then generally they'll take that and they prove whether the code will scale up. Then they'll do a few hundred nodes. And then they'll do a thousand nodes.

How much time does that take? A few years?

No, no. Probably about six months. Once it's deployable, [it scales fast]. So that's kind of where we are today.