In this podcast episode, host SPEAKER A interviews Sid Chouraria, founder of SC Marwar Capital, about the unprecedented scale of the current AI infrastructure build‑out. Sid frames the AI rollout as a historic industrial transformation comparable to electrification or the interstate highway system, citing projected cumulative AI cap‑ex of about $6 trillion from 2025‑2030 and $750 billion in 2024 alone from hyperscalers. He contrasts today’s spending—driven by highly profitable incumbents with deep cash reserves—to the unprofitable, speculative outlays of the late‑1990s dot‑com era. The conversation then moves to macro‑level risks (rising yields, Fed policy, power availability, labor shortages, critical minerals) that could affect financing and execution. Sid outlines where value is likely to accrue across the AI stack, highlighting “pick‑and‑shovel” winners such as Nvidia, Samsung, SK Hynix, Micron, and TSMC, which together command massive operating cash flows and have seen multi‑digit stock price multiples. He also notes the emergence of power‑ and cooling‑focused firms (e.g., Bloom Energy) as data‑center heat‑dissipation becomes a bottleneck. The hosts discuss market pricing, the lag between cap‑ex and ROIC due to accounting, and the broader impact on S&P‑500 performance, with AI stocks driving 70 % of the index’s rally. Regional dynamics are touched on, emphasizing strong gains for Asian semiconductor firms. The episode concludes with a recognition of the scale, uncertainty, and potential long‑term productivity gains of the AI build‑out.
Chapters
Joining us now, Sid Chouraria, the founder at SC Marwar Capital. Sid, great to have you on the show today. Now, you are framing the AI buildout as potentially the biggest thing since electrification or the buildout of interstate highways. So I would love to know what gives you the confidence that this is going to be a success like those 2 things and not a buildout of excess more like we saw in the '90s with, say, fiber.
Well, that's a great question. I mean, it's a $6 trillion question. You know, just to recap, as I said, I think the AI right now we're in, it's one of the largest industrial buildouts. The scale exceeds the historical national-scale industrial buildouts in the past. So the Apollo Project, electrification, interstate highways, etc. As we, as we know, the largest hyperscalers are expected to spend up to $750 billion in AI CapEx in 2024. That's a staggering increase of 80% growth rate in CapEx. I mean, those numbers are huge. The S&P alone, the S&P 500 CapEx is going to increase 33% this year. I don't recall the last time we saw such big figures. So to put that in context, if you estimated CapEx by Goldman Sachs and others, 2027, we're looking at $1 trillion in CapEx and then we're between 2025 to 2030, that's a cumulative estimated of $6 trillion. Now, obviously, there's a lot of variables here. And some of the important differences between, say, the 1999 dot-com bubble and now is that a lot of the spending right now— a lot of the spending back then was young, unprofitable companies. And today, the spending is highly profitable companies that are incumbents with enormous cash flow, low cost of capital, and billions of existing users. Now, there are risks with this, as there is with any such big, large buildouts. The hyperscalers are really— have changed from ultra-high margin, capital-light businesses into massively capital-intensive infrastructure operators. Their cash flow this year is expected to be eaten up by their capital investments. So there's uncertainty around power availability, labor shortages, financing, regulation, critical minerals, and whether the AI demand And to your question, will ultimately justify the trillions of dollars of investment. I'll raise one point here before I turn it to you, is that I suppose the question really is where the value will accrue across the industrial supply chain, from the chips, data centers, power infrastructure, cooling systems. The question also is, is it gonna raise US productivity? Is growth going to be reallocated because of the resources that will expand the production frontier? And will margins increase in the S&P over the long run? And how is labor going to be reallocated? So even if parts of AI are in a bubble, history shows that excess infrastructure can still become foundational. The fiber optic overbuild, as you point out, of the late 1990s, which was excessive, laid the foundation for broadband streaming, cloud computing, and mobile internet. A lot to unpack here. I'll stop there and see if you have any questions or comments on what I've laid out.
Yes, Sid, nice to see you, my old friend. It's so nice of you to join the program. Look, you know that my brain automatically goes to the global macro, so I do want to turn to the news of the day, and that is rising yields across the world right now. I mean, this is dampening risk appetite right now. And when you're talking about sort of long-term CapEx plans and spending. I just wonder if the market is now pricing in the Fed potentially hiking rates, whether that threatens some of what we're seeing here.
Yeah, I'm a bottoms— Sam, it's great to see you again and thanks for inviting me to the show. I'm a bottoms-up fundamental stock picker, but obviously macro and the interest rates are, as Warren Buffett says, are gravity to stock prices. So when the differential between the long-term bond rates versus the dividend yield of the S&P 500 gets out of whack. And that's really when we should pay real attention. And right now, the equity risk premium does not justify or sort of given the valuations and the run-up that you've discussed with stocks does not necessarily give investors a margin of safety relative to what one can expect with sort of long-term bonds. And if the Fed does indeed raise rates, which is not the consensus, I do have a few friends who are expecting that because they do believe it is actually going to be inflationary. That would certainly cause, cause valuations because interest rates are gravity to stock prices. That is not what I spend my time thinking about because it's really hard to— it's really hard to focus on. My focus is really, are earnings going to grow and am I paying a reasonable price for it?
So as you look at the earnings questioning who will grow, who is going to see ROI on these massive CapEx investments, and if you're spending the right amount for these companies now, who do you view as capturing the most economics across the stack? Chips, cloud, power, software, of course, should be a part of this conversation. Where do you currently see the most durable margin pool?
Yeah, just before I get into that, the top 15 AI stocks have are now worth $33 trillion as of yesterday. Before OpenAI released ChatGPT in November '22, the same group was worth just $8.8 trillion. So we've seen a massive $24 trillion just in the top 15 stocks where value has accrued, to your question. And they account for roughly about 20% of global equity stock market. So that's, that's massive. 15 stocks, 20% of global equity value. Now, since late March, the tech AI stocks have driven 70% of the S&P 500 rally, and without that, the S&P 500 would have underperformed Europe and Japan. Now, to answer your question, I think there is— the AI spending is originating a lot in the US and China, but many of the winners, which some exist in the US, but there are many in areas like Asia, Taiwan, Korea, Japan. So we've seen some explosive moves this year. Korea is up 74% year to date, 200% Over 12 months, Taiwan's up 40% over the last year to date and 80-90% over the last 12 months. There was this really interesting chart that was put up where it demonstrates how the cash flow from the large hyperscalers are moving to sellers of products in shortage. And those sellers, who are those sellers? There are a few of them which are very prominent. Nvidia, of course, the leader of the pack, which is estimated to have operating cash flow $150 to $170 $70 billion this year, which is insane. Samsung, which is expected to have $120 to $140 billion in operating cash flow. So still trading at single-digit P/E multiples. Korea's Hynix, which is trading at— which is expected to have operating cash flow of $100 to $120 billion this year. Again, totally insane. These are mind-blowing numbers. I've not seen them for large caps in my career. And I doubt I will see this again happen. Micron, $54 billion, and TSMC, which is $46 billion. So these 5 companies have accounted for the significant CapEx moving from sort of the cash flow moving from the large hyperscalers. I do think that there are lots of different winners across the industrial supply chain, even in the power, power electricity area. You've seen the stock of Blue Energy, that's up significantly. So there's a lot of winners. And right now what's happening is the cash flow from large hyperscalers moving to some of these pick-and-shovel plays. And the trend is strong and I expect it to continue.
Yeah, it certainly is. I was having a discussion with some data center guys the other day who said these things, these chips, particularly Nvidia's, are just getting so powerful that they're so hot. We need to cool them down, and the previous cooling systems are just not cutting it anymore. They're actually having to run water over these chips, like actual water with obviously a chemical in it so it doesn't damage the chip, in order to keep things cool in the data center. So you can really get a picture of that broad ecosystem, as you say, Sid. And you know, one man's capex, another man's revenue. Why do you think the markets have this sort of indigestion? With some of these companies, though, and their spending plans? And I take sort of Meta and Microsoft for an example just in recent weeks.
Well, the future is inherently very hard to predict. You know, if you ask— I mean, we would be doing this interview 15 years ago on video conference. You know, we think the future is very inherently unpredictable. Right. We've had these seismic shifts, the internet, mobile, cloud, and none of it was really predictable. And You know, we've had some blowups with the Fed back in 1999. Again, as I point out, largely unprofitable. What's different right now is many of these incumbents are profitable, cash rich, but the figures are just staggering. Like, I don't think anyone would predict— would have predicted— the best analyst on Wall Street would not have predicted that Google would be spending close to $200 billion in CapEx this year. I don't think even Google would know this 4 years ago. So There is. It is. It is right to be somewhat concerned because the figures are staggering. I think investors should flow towards where the money is going rather than focus on— well, rightly so— focus on the return on invested capital. And, you know, the stock market's been rewarding the pick-and-shovel plays. And as I pointed out, some of those companies have had very, very significant runs. Nvidia is up 18x since ChatGPT was announced. You have SK Hynix, which is up 22x. You have HD Hyundai Electric, which is up 32x. You have Bloom Energy, which is up 16x. Palantir. So the money has been flowing away from the hyperscalers, at least right now, with the exception of Google, which has done well, to some of these pick-and-shovel plays. I think eventually, Sam, the question is going to come, is this a healthy ecosystem? Eventually, companies spending hundreds of billions in AI infrastructure will need to earn acceptable returns for the pace of spending. But right now, the ROIC is not yet showing up to decline on the financial statements because of the way accounting works and whatnot. So I think right now, as investors are concerned, but they don't yet see earnings declines because of the way accounting works.
And Sid, you are not alone. And as you look at some of those numbers, double and triple checking them, because Sam and I regularly are checking second sources to make sure there aren't typos as we see these numbers come across and they seem astronomical. But we're sadly going to have to leave it there. Really appreciate you joining us today. We both hope you have a wonderful weekend. Sid Jararia from SC Marwar Capital.
