00:00We just heard from NVIDIA's CEO, Jensen Huang, all excited about the numbers, and obviously the market's excited as well.
00:08Is there a sense that we're underpricing risk in this AI space?
00:12Well, I think NVIDIA is absolutely the center of this and a major beneficiary.
00:17I mean, they are the supplier of the key components of the AI factories.
00:21So from their perspective, they are doing extraordinarily well at maintaining their edge, going up and down the stack.
00:28So they're not just producing chips, but they're trying to get into the application layers as well.
00:33And so they're a spectacular company and beneficiary.
00:36When it comes to risk, there's lots of risks.
00:39And there's risks on the valuation front.
00:41There's risks on the leverage front, the amount of leverage that's getting pushed into the system.
00:47And there's possibly risks in terms of technology and technology change over time.
00:51And there is such a thing as too much leverage, too much risk.
00:56Do you think, do you get a sense that we're seeing it in this space?
00:59Yes, NVIDIA is doing well, but it doesn't mean that everybody else has to come along.
01:05Yeah.
01:05So I think the competition for NVIDIA is other chip manufacturers.
01:10So there's certainly competition from AMD.
01:13There'll be competition from Google producing its own chips, TPUs.
01:17There's competition from Intel and others.
01:20But NVIDIA is managing to maintain a very significant edge in terms of the quality of its products and development.
01:28I think that will continue.
01:29I think where the risks are away from NVIDIA, it's further up, it's the builders of the factories, the massive building of data centers.
01:38And data centers themselves are big boxes with energy supply, water supply, a lot of other components in there.
01:46There's a lot of leverage starting to creep into the building of those data centers and at different levels for the project level, SPV level, at different layers, the company level.
01:59And I think that's what people are keeping a keen eye on is how much of that, how much of the growth has been funded by debt, which may be fine over time.
02:07There's a lot of cash flow that's coming out of them, but over time, that's where risk can be a big problem where, you know, if there's a downturn in pricing at some point or step change in pricing, then it can, it's the debt holders that will get hurt.
02:21So what I'm hearing, what I'm hearing is that it is too soon to be talking about the bubble bursting.
02:27Yeah, I think that's right.
02:29I don't think anything will happen tomorrow.
02:31I think that over time, the question is whether the growth can be sustained in demand.
02:37I'm obviously an optimist.
02:38You know, I'm investing in application layer companies.
02:40So I'm optimistic about adoption of AI over time and that it's going to become more and more widespread.
02:47But I think there are risks.
02:49If you are entirely concentrated on building for one form, one type of architecture of AI and everything in your factory is designed and optimized just for that, it's possible there could be some, I think it's sort of 15, 20% chance, but there could be a step change in architecture that results in a reduction in the need for that brute force computes for every unit of intelligence.
03:14So that could cause a step change in pricing at some point in time.
03:18That's one risk to keep an eye on.
03:21And, you know, there's a risk that there could be just a slowdown in growth for some period of time, which then causes a reduction in demand or just huge amounts of supply that causes a reduction in demand, reduction in the ability to keep prices competitively high.
03:35So all of those things could lead into, you know, reduction in pricing the data centers, particularly if you've got speculative data centers, you haven't got contracts locked in for a long period of time.
03:45And that could then result in challenges for people who hold the equity in these data centers when, obviously, the debt holders will be, you know, further up the capital stack, more likely to have their capital preserved.
03:58But if you're in the equity, you're going to get hurt.
04:01Mark, talk to us about how you're invested in AI and what's the strategy there.
04:06Yeah, so my strategy is, my view is that over time, the profitability will go up the stack.
04:11It's happened in every other way, whether it's mobile or cloud or any other shift.
04:16And so the application layer is where we're investing.
04:19And we're investing in companies that are using AI, using these foundation models, using these data centers to solve real business problems.
04:26And that could be a problem, for example, in advertising.
04:31So we've invested in a company called StackAdapt, which is the leader for small and medium businesses in advertising, programmatic advertising in North America.
04:38It could be in health care.
04:40So we've invested in a company called Skin Analytics that is the leader for skin cancer diagnosis.
04:46It could be in mechanical design collaboration software.
04:50We've just announced an investment in a company called CoLab that is the leader in collaboration for complicated design teams to visualize components.
04:59For example, aircraft components or automotive components and predict where the problems could be in manufacturing those components.
05:06So there's a myriad of ways that you can use AI to better solve real-world business problems.
05:14And we're focused on companies that are solving these problems, solving them efficiently, better, faster, cheaper, and doing it in a way that creates real value for the end customer.
05:25We're focused on the unit economics of that value creation.
05:28Is it better?
05:29Is it increased revenue?
05:30Is it reduced cost?
05:31There's a lot of reduction of cost, but even better, increasing the revenue for the end customer.
05:36Faster, better, cheaper.
05:37It does seem you'll find those companies in China.
05:40Do you?
05:40Yes, but I'm not investing in China.
05:43Why not?
05:43So because there's a bifurcation of the world, it's difficult to invest in advanced technologies in China and in the rest of the world.
05:51Sadly, the world has changed.
05:53So it's difficult to do that.
05:54You've got to keep a keen eye on where that competition could come from China.
05:57I think, as Jensen Huang has mentioned, 50%, 60% of the world's top AI researchers are in China.
06:04They're going to keep innovating.
06:05They're going to keep producing extraordinarily interesting products.
06:09And so you've got to keep a keen eye on what might come out of China to disrupt your products.
06:15But I'm not investing in it.
06:16I'm investing in North America and Europe.
06:18We've actually got a little bit of a tilt towards, you know, away from the valley to elsewhere in North America, whether elsewhere in the U.S. or Canada or U.K. and Europe, which is a little less frenzy and competition for competition for each opportunity.
06:33Sticking with AI, we know that a debate right here at the New Economy Forum is who's going to win the AI race.
06:40Do you have a view on that?
06:41Is it going to be China or would it be the U.S.?
06:45I think, well, one thing I remind people when people ask who's going to win the race.
06:49If you remember a few years ago, the race was who's going to win the 5G race.
06:53You tell me today who won the 5G race.
06:56I mean, that was the big question.
06:58Probably if we were sitting here five, six years ago, it would be the question.
07:02And I'm not sure who won the 5G race at the moment.
07:05So I think it becomes dispersed.
07:07It becomes part of everyday life.
07:08We have 5G on our phones.
07:10We use it all the time.
07:10So is it a matter of framing the AI tech race?
07:14How would you reframe it?
07:15I think so.
07:16I mean, I think there will be winners everywhere.
07:18I think China is going to produce competitive foundation models, which they continue to do, open source models.
07:24There will be an increasing amount of adoption within China.
07:27There will be countries around the world that will use Chinese software and software products and hardware products.
07:34I think the U.S. has edge, particularly on hardware and compute.
07:39So I think it's very hard for China to catch up on the semiconductor side, particularly without advanced lithography equipment.
07:48It's very hard for them to get past 7 nanometer node.
07:51So I think that they're going to have to continue to innovate with a reduced edge of compute capability.
08:00But they will do that.
08:02I think the framing around, you know, who's going to win AGI, I think, is years off before we reach true AGI.
08:09I don't think that's going to happen tomorrow.
08:12And so I don't think the question is who's going to reach AGI first is a good framing.
08:15I don't think it's going to be a good framing.
Be the first to comment