EP 148: Is the AI boom cracking? Inside July’s sell-off and beyond

Speaker A: Welcome to beyond the Benchmark, the EFG podcast with Mose Afzal.

Speaker B: Hi everyone. So today we have a very, uh, I'll say deeply requested or highly requested podcast on the current situation in uh, semiconductors technology and everything that is AI. So, um, to help me, uh, or us to navigate this uh, journey, uh, I have Henry Walters. Henry, welcome.

Speaker C: Hi Mose. Good to be back.

Speaker B: And uh, Henry is a, uh, equity analyst here at uh, efg, um, to talk about, uh, and covers the technology sector. So he will be uh, helping us to navigate everything. There's AI and semiconductors and hyperscalers and so and so forth. Um, so, um, Henry, should we just start straight away, just a quick summary of exactly what's been happening in 2026?

Speaker C: Yeah, uh, it's been a busy year, uh, and an eventful year. It's gone through kind of multip, uh, stages of kind of euphoria and brief moments of panic. Um, but I'd probably summarize 2026 overall as a year where uh, enterprise AI, uh, in terms of agentic AI adoption really hit an inflection at the start of the year with the release of uh, Anthropic's Opus 4.5 model and Claude code and the coding applications of AI really inflected at the start of the year. And you could see that in the numbers reported from Anthropic in terms of the revenue generation, uh, they ended December at about 9 billion revenues and hit in May hit about 45 billion annualized. And then, you know, they haven't. These are all kind of private companies that aren't kind of formally disclosed financials, but reportedly hitting kind of 60 billion in June. Um, and that really drove. That was one of the key drivers for um. You know, AI is monetizing really, really well the kind of AI capex trade that's fueled markets for the last few years really powered on once people saw okay, it's generating real revenues. Um, but at the same time, the other kind of effect this year was all the kind of economics of AI was really kind of being concentrated in the frontier labs. So really kind of anthropic, uh, at the charge and then OpenAI once they pivoted. And so that's caused some interesting developments more recently in terms of who's capturing the value. But um, that kind of takes you up to the end of June and then July, if I look at uh, the performance of semi index year to date to the end of June was up kind of 82%. Uh, July hit a bit of A wobble to put it mildly with a few stocks down 40, 60% that kind of range within one month. Um, so really kind of a big momentum unwind, uh in terms of it's very crowded positioning within July. But what I'd say is the fundamental data points that uh, we look at and everyone's tracking in terms of how well uh, AI demand is doing, uh, you look at the cost of compute, GPU availability, uh, tokens generated. Is there a real uh, concern that AI has hit a stall point? None uh, of that was really happening. I think it was more just around you know, market positioning concentration, you know, a lot of leverage, uh, particularly you know, concentrated in some Korean, uh, uh, some Korean traders made the press in terms of single stock levered ETFs. And so there was certain things happening under the hood um, that caused this momentum unwind. The only fundamental kind of negative data point that uh, you could point to in July during the sell off was rising rates, uh, credit spreads widening for a lot of the hyperscalers, people investing in infrastructure, uh, and that is a real potential kind of concern, something worth monitoring. I think it's kind of moderated since as we speak now. Um, but other than that there wasn't much else to kind of point to in terms of negative science for July.

Speaker B: Yeah, let's double click on those negative aspects because I think sometimes um, you know it is quite confusing and you know, let's try and put it into very simple terms. Um, so overall, and in fact I've just um, updated the numbers, uh but overall global tech earnings for 2026 are in the region of around 70% up on last year. And this is already up sort of 30 plus from the previous year. They are very weighted towards the hardware side and semiconductors and anything that uh, powers uh, AI infrastructure. So the earnings side looks know, robust, looks strong, all the points you make are very good. But double clicking into you know the air pocket of July, one can say two things. First real yield started to climb, um, 30 year US treasuries, um, you know, hit a, you know hit a, a, a range high as well as buns and guilt and so on so forth. So that certainly had an impact. And the second impact was huge amount of issuance that then fed into spread widening. So very uh, large issuance from Meta is probably the one that was um, uh, the most high profile uh, of the names. But you had other smaller Neo clouds and so and so forth. Also issuing a lot of debt over relatively quite some period as well, which we shouldn't forget. So we've seen this sort of spread widening and maybe to summarize all of this meant that we are spending or companies are spending huge amount on it, on AI. You know, ultimately over the next few years, trillions of dollars part will be coming from operational cash flow, a lot of it hopefully particularly from the hyperscalers. But there's also a lot that is debt fueled. And so um, in an environment where interest rates go up, that debt fueled slows the, the spending. So I think it's important to kind of put it in very simple terms exactly what you just said. Um, that is, and I think that is the more realistic um, catalyst for the sell off. And plus we had highly leveraged investors who got margin called and, and, and, and yeah, you know, cut off that led to that exceptional volatility and unfortunately that volatility probably will stay for a little bit I suspect, um, for a few more months um, once people have kind of calmed down and kind of moved on.

Speaker C: So I think that yeah, you've obviously got for many of the big spenders, free cash flow going to zero or negative, uh, for Meta and Google. And then you also at the start of July had the reported news that Meta was looking to rent out some of their compute, which surprised people. Yeah, when you're raising capex and you've been raising capex and all of a sudden you're saying oh, I'm going to rent some of it out to someone else and you know, the market's been believing, uh, you know, is that, is that a sign of demand is weak? Are you overspending? Are we in an oversupply? Um, I think, I think the point

Speaker B: that you've made is something you've made to me. Um, is that the Meta situation. Let's cover that.

Speaker C: Sure.

Speaker A: Situation.

Speaker B: Uh, I think it's quite interesting because um, you're right, people say oh, you've overspent, you've got too much infrastructure and actually you need to rent it out. But there's a reason for that. Right. Because spot price for a token or um, uh, um for that infrastructure was way higher than the long term contract. Explain that dynamic.

Speaker C: Yeah, of course. So um, when you build these AI clusters, you can either use it as kind of a cloud and rent it out or you can use it for internal workloads. So Meta is obviously one of the largest digital advertising businesses in the world. Um, however this year as the enterprise AI adoption has taken off, that's kind of left Meta without that growth engine. So if we talk through how if you're a cloud, if you're a NEO cloud or a hyperscaler or colocation data center, how do you monetize? You typically rent out your compute with some kind of large marquee customer or, or a group of them on usually kind of a five year contract. Why is it a five year contract? Because otherwise no one's going to lend to you if you're uh, how are you going to fund it if you don't have kind of backing and you know, some, some real credible customer who's going to rent it off you, so you rent it on this longer term contract. They've been priced around, call it you know, $15 billion per year per gigawatt. What's happened since is the value of that uh, uh, cluster or that AI infrastructure has gone up. So the spot markets are um, much, much higher than that $15 billion per year. Um, so SpaceX came in with uh, some short term contracts that typically they announced one with anthropic, one with Google and these might be 90 day contracts. Well they're talking about an earnings call this week doing six month contracts and instead of at ah, $15 billion per gigawatt, it can be 30 or 50, uh, at kind of peak pricing. And so you know, if you're meta, you're looking at that, you're thinking, oh, I can monetize this at a much higher rate just by renting some of it out and I can use that to fuel my capex. It's a very attractive option. Um, and you know, Musk said on the call this week, you know, if you can do $50 billion, uh, revenues, which we'll see if that sustains, you know, it's a, it's a one year payback on your capex, you know, whereas if you're doing the 15 billion, it's still not a bad return. Right. Angie Jassy, CEO of Amazon, was talking about, you know, less than three year payback uh, for these AM structure. You still get a good ROIC on that investment. But you know, at the moment, because demand's been so strong, there's a real premium uh, on what people are willing to pay for that computer.

Speaker B: And I think that's kind of critical because the underlying strength here is the demand. The demand is literally off the charts and supply is not coming on fast enough to be able to satisfy that demand. And hence you're creating these very distorted markets in the short term. So let's quickly go through the results, um, uh, that we've had so far.

Speaker C: Yeah, I'd say um, meta, you know, there's a lack of clarity around how they're going to monetize. They talk about, you know, selling enterprise services, becoming more of a cloud company. Uh, but that requires a very different kind of go to market distribution function. Uh, and yet you know, they're still investing heavily. I think the other, the other narrative that the market always gets uh, fixated on is who's got, who's at the frontier in terms of model development. So the market's always a bit skittish around, you know, model leadership but in aggregate, you know, in terms of the cloud, uh, cloud results. So across Google, uh, Amazon, Microsoft, you're uh, seeing very, very strong acceleration I think as a group growing over 40% year over year and you know, doing you know, well over kind of 300 billion of kind of annualized revenues across the three of them. These are massive businesses that are accelerating and the operating cash flow also supporting that and inflecting upwards, um, which is a positive sign in terms of how much debt they need to raise if operating cash flow can keep up with their capex ambitions.

Speaker B: And I think that's the critical function here is as long as your operating cash flows are moving higher and, and are strong, market will forgive you for spending. Basically, um, if it's, if your main operating business is not doing as well and um, and you're, and you're spending and you have to go to the debt markets or even equity markets to raise capital, you know, that's probably uh, a more negative signal and uh, that sort of inflection point is going to be the key thing I guess to watch.

Speaker C: Yeah, I agree.

Speaker B: Okay, so let's talk about the semiconductor companies. Um, how they look so far.

Speaker C: Yeah, uh, they move very differently. There's a lot of different sub themes I'd say within AI infrastructure as a whole. Uh, and different different flavors of the month appear uh, and disappear and what you've had is, I'd say, you know, in aggregate during the year. We've obviously had you know, the memory shortages and memory pricing. Memory's been a massive theme this year. Uh, it's, you know, it's a much more commoditized market but pricing is very susceptible to any kind of supply demand imbalance and you know, foreseeable memory shortages has driven up prices of dram and nand, you know, well, well above kind of, you know, 3.5x uh versus what it was prior. So memory's seen a huge resurgence but tends to be, you know, much more, more volatile. Um and then you have, and that's

Speaker B: why you have these sort of crazy situations in memory where the um, the price earnings multiples of these companies are low single digits.

Speaker C: Yeah.

Speaker B: Um, but um, the reality is they're not necessarily priced off price earnings multiples. They usually price to book or other um, measures uh, that you'd use to, to, to value those types of companies. But I think there's a, it's, it's a tricky one because you know, you just don't know when the peak's going to be.

Speaker C: Yeah, that's why everyone's watching. You know, price of compute. Uh, that's kind of the, the main signal or you know uh, how well is AI monetizing is really a signal for, for demand and yeah, so far it's remained very, very strong. And you uh, know the infrastructure budgets, you know they kind of grow in confidence for continually rising.

Speaker B: So overall not much concern in terms of semiconductor um, uh and the food chain of semiconductor um, results looked actually I guess okay. But it's the frothy part right of the market and that's when you've seen these share prices move up a couple of hundreds of percent in some cases a thousand percent in some cases in a relatively short space of time. That's kind of where the froth is sitting uh, in there and hence the volatility is just also is off the charts.

Speaker C: Yeah. And it's just been a very narrow market. You know, uh, I very much focus on tech but you know the other parts of the market far better than me. And you know a lot of managers are chasing what's uh, momentum and what's been working just to try to keep up. So that kind of whole positioning effects and crowding into some of these names, you know amplifies everything on the upside but similarly on the downside.

Speaker B: So let's move on to talk about um, the language models, how Claude is doing open source versus closed source.

Speaker C: Sure. So I think the again kind of repeating myself but the challenge is these companies aren't public so you kind of rely on analyst estimates and company you know, kind of sporadic disclosures on what the financials are of an anthropic or an OpenAI. Um, but if I go off kind of what's reported for Anthropic, as you mentioned, the premium kind of frontier models. So that being anthropic suite of models and OpenAI as well, they've captured the lion's share of the economics uh from kind of AI software tokens, the applications built on top of that and if you look at the kind of the reported financials OpenAI anthropic might be at 60 billion uh, revenues and the, the kind of part people debate the most is what margin do they earn on that. Ah, which I've seen you know, from, from 60% gross margins up to 80% plus, um, which you know, kind of aligns with what we were saying earlier about if people are willing to, if they're willing to spend not just $15 billion a gigawatt but 30 or maybe higher. There must, they must be monetizing it very, very well at 15 billion. Um, and what you've had is you know, collectively a lot of the tech industry or the major players are getting or have been quite nervous that a lot of the power uh, and the value capture is concentrated in these frontier labs. And so open weight models and what they are, are um, you know, models where you can freely download them, they're released, you know, for free, anyone can download them. They're still massive models so you still need to run them on servers but you can host it yourself and you don't need to pay a uh, big chunk of margin just for the privilege of using that model which you'd pay to anthropic OpenAI and typically they lag in capabilities. They're not quite as powerful as the best models. What was last year's frontier is this year's commodity has been the case for the last few years. Usually uh, there's kind of a six month lag, give or take, um, between what is at the very frontier and what's at the edge. But with the massive growth we saw in how frontier models and the value they generated at the start of this year, it's not surprising that some of the open models are now kind of hitting the point we were at the start of the year and they're really useful. And so as you mentioned there were some big releases out of China, um, from Alibaba with the Kwen models, uh, Kimi K3zai's GLM 5.2 which are all very, very capable models and can be used um, as an alternative to using kind of these expensive closed end models. And that benefits that. Who does that benefit? Who does that hurt? If the world switches towards closed models, is it bearish kind of AI infrastructure as a whole? It's clearly kind of a negative. If you're a frontier model maker and you want to charge that premium margin, it's more competition which net of everything, all else equal, that's going to hurt your business model. But if you're a consumer of tokens or if you're an inferencing uh, kind of host, you host inference, you're an inference provider as the hyperscalers are, um, that can benefit you a lot more because you don't need to share as much of the profits with an anthropic or an OpenAI. So there was this big kind of open weight uh, letter that went out with most of the, you know, VC industry, Nvidia backing a lot of hyperscalers in kind of support of open weights models. And I think it's mostly around you know, who's capturing the value. Uh really. But it's not necessarily negative for the AI, uh, CapEx and the infrastructure. You know a token's a token. It doesn't necessarily matter which one you serve. It just kind of if you're dividing the pie, who's capturing that value?

Speaker B: Yeah, I think that's going to be the debate. The reality is for most, and the cost per token for a uh, Frontier Lab token is much more expensive than a open weight one. And obviously budget, when it comes to budgets, you know, CFOs are certainly putting their foot down on uh, uh, individuals, uh, and, and departments even to say hang on a second, don't use this one, use this one because it's 70% cheaper.

Speaker C: Yeah.

Speaker B: Right. Which is, which is uh, what's kind of driving through. It'll still use the same AI compute, uh, um, uh, if you like infrastructure. But it will be just a lot cheaper.

Speaker C: Yeah. And I think that they'll be used in tandem. I don't think it's a, uh, either or. I think they'll continue to grow together.

Speaker B: So I think that feels like the direction is going. I suspect the reason why investors are focused so much on the AI infrastructure, semis and so and so forth. It's an easier decision to make than it is who's going to be the winner in AI. That's a really hard, really difficult decision for investors to make.

Speaker C: Yeah. In terms of who's going to build the applications to. Yeah. To capitalize on this there's a lot of competition. You know you've got uh, the incumbent kind of SaaS, vendors who've all pivoted to this and they definitely benefit from kind of uh, cheaper uh, LLMs and open weight models. But then you've got you know, uh, the labs don't just want to sell tokens, they want to sell kind of products and applications on top of that. So they're kind of, you know, meeting in the middle. And then, you know, additionally you have, you know, every, you know, hyperscaler selling additional services and custom, you know, custom models for every enterprise.

Speaker B: I think what's interesting is, again, trying to draw analogies here for, for our listeners is the, I call it the Netscape and Yahoo argument back in the early. No, mid-90s, and that's a really hard call.

Speaker C: Yes.

Speaker B: Sitting in late 90s or mid-90s net, no one had an idea who would be winning. And at the time everyone thought that Netscape would be the winner.

Speaker C: Yeah. And things shift very, very quickly so that, you know, if you're in the lead now, that's. It's not a given that you're going to retain that.

Speaker B: Exactly, exactly. Yeah. So, and I think that's also m. Let's, let's maybe sort of finish off in terms of, um, um, uh, the risks that this all brings about. Right. Because it's, the risks are actually quite a lot, first of all, just technology risk. We just don't know. And anyone who pretends to know is probably lying.

Speaker C: Yeah. And even you don't know, you know, you don't know what the next model is going to be capable of. You know, just the technology, it's still very nascent. Right. If you, you know, the, you talk to anyone at the AI labs and you know, they believe that the, you know, digital deity and, you know, super intelligence is coming and, you know, it's, it's a bit, you know, hard to get your head around. What does that mean in terms of products?

Speaker B: Yeah, exactly. Um, so, um, so I think demand is there, but let's talk about some of the negatives that are developing, uh, other than the technology risk. We talked about credit, uh, spreads and the, and, and the rising cost of capital that obviously was. It won't kill AI, AI demand, because it's, it's there and it will stay forever, probably, but it could slow it down. Right. Which is the, the, the challenge or, or certain companies are not able to keep up with their cost of capital and just fold. Right. That's the, that's the, uh, you know, that's. I call the macro risk. Again, we need to watch out for, uh, you know, very, um, you know, very carefully. Um, the other risks, ah, are geopolitical. Um, we obviously have a, um, China, us, uh, AI race. Uh, and that's clear.

Speaker C: Um, but yeah, you know, there are two different kind of technological spheres. Uh, you know, we talked to them, we talked about China in terms of the context of some of the open models. Um, but you know, they are clearly building their own self dependency in terms of chips and infrastructure and are you know, aggressively trying to catch up. Um, so yeah, and then the world, you know, US and China still relay, still remain kind of heavily uh, interdependent upon one another. So you know, any, any kind of rising tensions there can easily put a, put a spanner in the works in terms of the supply chains required to build all this. Um, the other kind of policy risk, or I'd say regulatory risk you see around AI is just the uh, the mass broad public negative sentiment uh, towards AI. There's been a lot of you know, surveys done and it's, it's wildly, wildly unpopular amongst the.

Speaker B: Well I think the analogy in that and I think unfortunately the, the AI leaders have, haven't been particularly good at communication. In fact they've been awful at communication. And I think um, uh, you know the way to put it is if you tell a consumer that uh, by the way, I'm building a data center to increase your electricity costs, your water cost bills are all going to go up, your electricity bill is going to go up and I'm going to take your job away. It's not exactly, you know, something people are going to vote for. And I think that sort of, that risk is clearly increasing with moratoriums now already in New York, Texas even um, starting to see some negatives.

Speaker C: Yeah, no, clearly, um, whether it's electricity prices or just as you mentioned, the heads of these companies are telling you it's going to end work as we know it. That's fair enough. It causes a bit of uncertainty across the labor market, um, and the final risk. I'd say we talked about technological risk, um, but it's just you know, adoption typically isn't smooth. You know, there, there is a chance we see you know, a slowdown in terms of enterprise AI adoption or um, or yeah, you know, willingness, willingness to spend in ever increasing AI budgets. You know, so far this year it's only surprise to the upside but you can, you can never rule that out. You know, as with any kind of technological adoption it's, it's rarely, rarely such a smooth straight line upwards.

Speaker B: Yeah, uh, I guess the last risk would uh, be cyber risk. And um, uh we've seen a lot of challenges where um, uh, OpenAI software has sort of gone amok and um, uh, cyber attacked, um, uh, different platforms hugging face in this case or the one that's gained most of the attention. What are your thoughts around that? And um, does that then force governments to really slow the adoption down because they just don't know what they can control.

Speaker C: Yeah, it's a great question. And the cyber risk is clearly very, very real. And, you know, the leading AI labs is again, you know, the closed source. You know, they. They have. I, uh, think Anthropic has Project Glasswing, where, you know, their frontier models, they'll only test with kind of trusted partners, you know, including the government. Um, and yeah, they. They clearly are very, very worried about cyber risk. You know, you had to open AI incidents with, uh, with hugging face. And then you had. Anthropic had a. I think Anthropic had another incident, uh, relatively recently, um, where, yes, your models can just kind of go rogue and, you know, uh, very, very diligently perform their task, but maybe in ways you didn't really intend to. And yeah, it's an open question about how you solve that, because it may not be a problem if it's a closed model that you can just turn off at one provider, but if it's an open model, uh, which there's zero reason to believe they won't be capable of these kind of threats, uh, then it's much more easily used by kind of bad actors around the world. So the best defense people have come up with is you use more AI to defend against AI.

Speaker B: Henry, thanks for. Thanks for taking us through all of that. I think we covered a lot of ground. I suspect we'll probably find many, many of us are going to have to reread.

Speaker C: Re.

Speaker B: Listen, should I say to the. Or reread even, uh, the, uh, uh, this podcast. So thank you very much and given how fast moving this is, uh, no doubt, um, we'll have you again again on very soon, uh, as well. So thank you very much for taking.

Speaker C: Appreciate it. Thanks for having me.

Speaker B: Thanks. Uh, so with that, we'll, uh, wrap up. Um, thank you very much, uh, for listening. Uh, as ever, if you've got any questions or thoughts, you know, please reach out to us. Uh, in the meantime, have a great day.

Speaker A: This podcast is provided for general information only and assumes a certain level of knowledge of financial markets. It is provided for informational purposes only and should not be considered as an offer, investment recommendation or solicitation to deal in any of the investments or products mentioned herein and does not constitute investment research. The views in this podcast are, uh, those of the contributors at the time of publication and do not necessarily reflect those of EFG International or New Capital. The companies discussed in this podcast have been selected for illustrative purposes. Only or to demonstrate our investment management style and not as an investment recommendation or indication of their future performance. The value of investments and the income from them can go down as well as up, and investors may get back less than the amount invested. Past performance is not a guide to future returns, return projections, or estimates and provides no guarantee of future results.