Why hyperscaler earnings are strongly supportive of AI momentum

Let's just return to what we've seen on Wall Street, particularly with the AI trade. Back in the driver's seat essentially after another strong round of hyperscale earnings. Let's get across what is ahead, though Sam Ruiz joining us from T Rowe Price. Sam, welcome back.

Hey, Andrew.

So we have seen those hyperscalers have been under pressure. This week, some real positivity to the point where it obviously pushed the the major indices higher. What have you made of those hyperscale earnings?

Yeah, well, we had a very volatile July. It isn't that long ago, but it was on the back of one of the best positive months of momentum in June and most of that sold off to be one of the worst momentum months in July. Why was that? Last time we spoke, we were flagging that we had this big moment coming for earnings where later in July, the hyperscalers would all be putting up their prints for CapEx and for monetization and margins within AI. So, with this enormous rally we've seen, it is really a big question mark in the market of can this momentum be sustained and also we're spending so much money on this AI cycle. Is it worth it? Are they going to actually make money from it? It was an unequivocal beat to expectation and really positively reaffirmed to the market. AI is making money. There are so many examples, from Open AI to Anthropic, even to SpaceX. SpaceX was down a lot last night. Let's put this into context. They have said that they think by the end of this year, their run rate revenue for AI compute is going to be $100 billion. You kind of have to just let that settle. These numbers are getting so large now the context of it is lost. So, all of this is reaffirmed for us that the AI cycle is alive and well. It's just this game of expectations now. Are the beats big enough? And then it could just be, let's roll through each quarter and make sure that we're continually reaffirming the positivity.

I mean, you mentioned what you see, particularly open AI, Anthropic, they've yet to come to the market. SpaceX obviously shares have come under pressure despite those are those beats that you've just mentioned of there. And on top of that, Sam, since we last spoke, of course you've had the emergence of another low cost model coming out of China raises that question of those those so-called open weight models and the challenge that lays down to the American business models.

Yeah. So those those American sort of frontier labs call them open AI and Anthropic still have the leadership. And it's not just this Kimi out of China that's doing open weight. Even NVIDIA already has an open weight model. There is a real tuggle sort of tug of war here happening. What it is, is we know that a lot of the the world would like lower cost models. That's open weight. We know a lot of enterprise, big large software companies want open weight. Why? Because they can put all their IP into ecosystem that they don't have to share it. If they give it to Anthropic or if they give it to Open AI, they're effectively sharing their IP and maybe allowing them to compete in the future. The problem is that Kimi is about 94%-95% as good, which was quite amazing actually for an open weight model with a six month lag. But that 94 percent is still not good enough for a lot of sort of the best cases to trust that the 94%.

It's going to get better though, isn't it?

It’s like I said, it's a three to six month lag right now. So as long as the frontier labs can continue to get better and better, it's called scaling laws, how much their models improve. That means that the open weights are going to be at this lag where they're potentially just not good enough or the value the better models are adding is better and worth paying for. Really important thing here, we're debating a lot right now is, what does that mean for how quickly this accelerates the need for those Frontier labs to accelerate to the next offering? We think that they're going to have the need to actually vertically integrate their products and maybe do this more quickly into an ecosystem. Let's call it the iPhone. They want to be the iPhone of AI where all the use cases have to sit on their system and they can effectively collect the rent and the tax. And if they don't do that, then potentially they get commoditized by open weight models, which is what you're inferring.

Where then does, where do those AI hardware stocks sit? You talk about potentially being in a sweet spot. How so?

So that is actually a great follow up question because there are hard problems in the market to solve and there are easy problems in the market to solve. The hard problem is what we just said Frontier Labs versus open weight timeline, how quick does that move? What we know though is that open weight models are still very, very hardware intensive. And if they start to gain traction, if they start to take market share, we know that the combination of open and close weight models together is just even further accelerating and further growing the demand for the AI compute capacity and infrastructure that sits behind that. So, coming out of earning season and even as a result of this Kimi release we saw in July, we're actually more bullish now on what that means for, call it, the next 12 months of demand for AI hardware.

Are you then bullish overall for market growth, particularly what you're seeing on Wall Street at the moment, given these results, given your forecast there, do you think that momentum can continue?

More optimistic. So, July was quite bad, particularly. Let's put this into context. AI has been one of the single things driving the market. It's responsible for almost all of the earnings upgrades and the majority of earnings growth that we're seeing. Are we thinking that we're still going to see that going forward? Yes, earnings are an upgrade cycle. We're also seeing in the US something that could derail this inflation, is actually relatively contained. And we're not seeing the yield curve, the long end move away war, Iran sort of energies is a bit of a hotspot when we're not seeing really closely. For now, we're actually seeing economic growth, corporate earnings growth and that volatility we saw in July was not on the back of earnings compressing. Earnings are actually still accelerating, which means it was a narrative LED sentiment LED sell off. And that for us feels more like a mid cycle correction, which is really healthy in the market as opposed to warning signs that we're starting to see a deceleration of growth. And that's something where we think multiples deserve to move lower and investors should be more cautious, but we're not there yet.

What are the other risks you're looking at the moment? Not the least, of course, being a massive amount of debt been taken on to fund this AI growth and whether there's the capacity to absorb that?

Yeah, that is a key question right now. There is more than likely going to be more debt and equity issuance, particularly by the big hyperscalers. They have a really big role to prove that they can actually monetize that. And it's worth investors allocating that debt to them. Why? Because otherwise their valuations will compress. If investors don't believe the story, their cost of funding in debt markets will increase. And that makes it incrementally harder for them to actually fund their ambitions. Put it in in the context of equity markets. If they want to raise, you know, Alphabet raised $85 billion of equity, I think last quarter. If their valuation falls 20%-30%, they have to dilute more to do that. So for now, these companies can actually take on a lot more leverage than what they already have and it would still be quite responsible so long as the returns are there. So for now, I think leverage is growing and will become a problem. For now, I'd say the bigger question mark is keep track of whether they are showing proof and evidence of monetization and the payoff which they have shown last week in earnings. They are. And that means that investors for now will say we back the extra leverage because we know it's got payoff for now.