What turning AI compute into a financial asset would look like

Yahoo Finance Video · NVDA news via Yahoo Finance Video · 3 min read · original

00:00 Speaker A

So the simplest way of thinking about it is that compute is processing power. And it's processing power now particularly for AI type services, basically. So

00:08 Speaker A

if you're a big cloud company, uh or a neo cloud like Core Weve, you're selling compute, basically. And if you're a big user, maybe an AI lab, maybe any number of other companies now that want AI services, you're a buyer of compute.

00:22 Speaker A

And there's an obvious potential for a sort of futures market here for some sort of financial market around this.

00:28 Speaker A

It's very, very difficult to get right. This is not a homogeneous asset.

00:34 Speaker B

Right.

00:34 Speaker A

It's going to be really, really difficult to turn into financial assets, but so much money is being spent on this now that there is an obvious drive to do it.

00:41 Speaker A

Many people are trying. I think it will come to something eventually. The question is whether the current wave of things work or whether we have to wait a lot longer to see a sort of financial market to match the real physical market underlying it.

00:54 Speaker B

I mean, to your point, the things that are commodities that there are markets for are homogenized things.

01:00 Speaker B

I think about oil, I think about ags, I think about metals, right?

01:03 Speaker B

You might have different grades, fine.

01:05 Speaker A

You do.

01:05 Speaker B

But but they're, you know, it's oil is oil at the end of the day.

01:09 Speaker A

So this is the debate I always have with people. If you talk to people, you know, startups and companies that are trying to financialize compute, they'll always say, oh, those assets aren't homogeneous, right?

01:17 Speaker A

You're buying a a barrel of uh jet fuel in Singapore and a barrel of crude in, you know, the Midwest.

01:23 Speaker B

But there are different markets for jet fuel for crude.

01:24 Speaker A

There are, there are certainly. But they say you can still have Brent, you can still have WTI, you can still have a benchmark even though you know that there's lots of products and they float around.

01:34 Speaker B

Okay.

01:34 Speaker A

Now, the reality is, I think with compute, we're talking about something that's much less homogeneous than anything like that.

01:41 Speaker A

You're talking about something where even if you're renting a chip, like a hopper chip from Nvidia, a H100 chip, even renting the same chip from different providers, not just in different parts of the world, but in the same part of the world can vary in price by two, three, four times.

01:58 Speaker A

Right?

01:58 Speaker B

What determines that price?

02:00 Speaker A

All sorts of things. The the stack, the actual physical housing of the chip, the uh the reliability of the data center, of all sorts of sort of nuts and bolts things.

02:12 Speaker B

What power is it linked up to?

02:13 Speaker A

Absolutely. Yeah. All of this matters enormously. Um it's something that a financial markets investor knows nothing about, right?

02:20 Speaker A

So you want clean, simple benchmarks. There are companies trying to do this on silicon data with their own indexes.

02:27 Speaker A

The reality is it's it is very, very difficult.