00:00We have an absolutely momentum-driven market where everything that is hitching its ride on this particular train has done
00:06extremely well.
00:06Everything else has fallen by the wayside, so we have very high dispersion.
00:11We have a K-shaped equity market, but everything in the tech space looks very stretched.
00:16And ultimately, the question is, who's going to pay for all this capital expenditure, for all this capex that is
00:21basically fueling the micron demand?
00:24And the answer is, it's going to be the end user who uses AI.
00:28And the question is, what's the pricing power there?
00:30What's the willingness to pay?
00:31And I think the profitability projections that you have currently look very optimistic.
00:36You're expecting peak global margins.
00:38You expect margins in tech that you've never seen before.
00:41And the question is, is AI a high-margin business that can fuel all this capex boom that is going
00:48on at the moment?
00:48And the projections that we're seeing look very optimistic.
00:51Okay, I mean, Guy Johnson has views on this.
00:53And to be fair to Guy, he brings us up almost on a daily basis, this question, the gap between
00:57the ROI and the spend that's coming through on capex.
00:59What is the window then, Sebastian, in terms of when that really hits and when that becomes a problem?
01:06Because so far, it has not become a problem.
01:08That is true.
01:09So I think we're seeing three data points, I would say, that suggest we're getting closer to when it's a
01:13problem.
01:13And then we get into very dodgy territory where it looks like these boom and bust cycles that you've seen
01:19in the past.
01:19And I think we all have to open our history books to see what happened to the railroads in the
01:231880s, what happened to the radio boom in the 1920s.
01:26And the three data points that we're seeing is the frontier model labs have started to increase their prices.
01:32Everybody loves AI if it's for free.
01:34But now they've increased their prices.
01:35And what we've seen, we've gone from token maxing to customers pulling back, rationing.
01:40Secondly, if you want a high margin business, what you need is scarcity.
01:44You need to have a product that nobody else has.
01:46But what we see is a proliferation of model providers, cheap, open source Chinese models.
01:51If you're trying to raise your prices on this AI product that everybody wants, but there's a cheap, low cost
01:57option that people can route 80 to 90% of their traffic to, then you've got a problem.
02:01You've got basically a ceiling of how high prices can go.
02:05If you do 800 billion of CapEx over the next 12 months, you better have pricing power.
02:10You better have a high margin business.
02:11But this is not the cloud.
02:13This is not an oligopoly with a few providers.
02:15You've got 20, 25 very strong providers that are within five IQ points of the frontier models.
02:22This is not a high margin business as far as I can see.
02:25Isn't that tomorrow's problem?
02:26Today we're dealing with the boom.
02:28Yeah.
02:28So what you're seeing already, the hyperscalers have, for now, they've, of course, jumped a bit on Micron.
02:32The hyperscalers, the people who pay this 800 billion, they've underperformed the market by 15%.
02:37They're at a one-year low.
02:39Map the semis against the hyperscalers, the guys who get the spending versus the people.
02:44There's a big gap now.
02:45So what the market seems to be saying at the moment is the hyperscalers potentially won't see the returns on
02:50that investment.
02:51So what are you going to do?
02:52Are you going to keep spending or are you destroying value?
02:55Will you have to pull back?
02:56But the people who receive that spending, they're basically still priced for all of this to come through.
03:00So what are you going to do?
Comments