Latest / The Bear Market Podcast with Fexingo: Surviving Downturns, Buying the Dip, and Long-Term Resilience / How AI Computing Power Is Becoming a Tradeable Commodity
Transcript
- Lucas: So there's a headline from this morning that I think is genuinely one of those 'watch this space' moments. CNBC ran a piece about turning AI computing power into a tradeable commodity — literally, the computing capacity itself, bought and sold like barrels of oil or bushels of wheat. Luna: I saw that. 'The new oil?' they called it. And it's not just a metaphor — there are actual efforts to create futures contracts or spot markets for compute. Lucas: Exactly. And the timing is interesting. Markets are ripping — the S&P 500 at 7,511, the Dow at 52,000, the Nasdaq at 26,376. The VIX has collapsed 26 percent in five days, down to 16.41. So we're in this 'risk on' euphoria, but underneath, there's this structural shift happening in how we think about AI infrastructure. Luna: Right. Because if compute becomes a commodity, the economics of AI change dramatically. The big cloud providers — they've been the ones hoarding the GPUs, setting prices opaque. A transparent market would change who gets access and at what cost. Lucas: And it would create investable assets. Think about it — if you could buy a compute futures contract, you could hedge against AI capacity constraints, or speculate on the buildout. The CNBC article mentions that startups are already trying to standardize compute units, like a 'compute hour' on a given chip type. Luna: It's reminiscent of what happened with bandwidth in the late nineties. People tried to trade it as a commodity, but the standardization was a nightmare. Compute is even harder — there are different architectures, memory configurations, latency requirements... Lucas: That's the key challenge. But the demand side is undeniable. Every earnings call now mentions AI capex. Microsoft, Amazon, Alphabet, Meta — they're spending tens of billions combined. If you could create a liquid market for that capacity, you'd unlock a whole new asset class. Luna: And it would also be a way for retail investors to get exposure to the AI buildout without picking individual stocks. An ETF that tracks compute prices? That could be interesting. Lucas: Speaking of interesting — did you see Michael Burry's comments on SpaceX yesterday? He said he's tempted to bet against SpaceX but passed because the options are too expensive. Luna: Burry betting against SpaceX — that's a headline. But he's been bearish on a lot of things. The options comment is telling though. It shows how much speculative premium is baked into high-profile names. Lucas: And SpaceX isn't even public yet — though the IPO chatter is everywhere. But Burry's point is that the valuation expectations are already so high that the risk-reward on the short side isn't there because options are priced for extreme moves. Luna: Which brings us back to the broader market. The VIX at 16.41 after a 26 percent drop in five days — that's complacency. The VVIX, the VIX of the VIX, is at 87.69, down almost 19 percent. That's a regime of low perceived risk. Lucas: And the yield curve is still positively sloped — 10-year at 4.47, 2-year at 4.07, spread of 38 basis points. That's the uninversion we've been talking about. Historically, a steepening curve after inversion has been a late-cycle signal, but not necessarily an immediate crash. Luna: Right, the curve has been positive since late last year. The economy is still growing — real GDP growth annualized at 1.6 percent, up from 0.5. That's not recessionary. But the market is pricing in a soft landing, maybe even no landing. Lucas: And the Russell 2000 is participating this time. Small caps up 3.7 percent in five days, outperforming the S&P's 3.4. That's a healthier rally than the narrow tech-led ones we saw earlier. Luna: But the compute commodity angle — that feels like a genuinely new way to think about AI exposure. If it takes off, it could be bigger than just another tech theme. Lucas: Before we go further, I want to mention something briefly. If you find these conversations helpful — and we hope you do — we keep this podcast ad-free thanks to listeners who support us through buy me a coffee dot com slash fexingo. It's a small way to help us keep doing deep dives without commercial breaks. No pressure, just wanted to put it out there. Luna: Yeah, it really does make a difference. We don't have to chase sponsors or compromise on topics. So if you're inclined, that's the place. Lucas: So back to compute as a commodity — the standardization problem is real, but there are precedents. Electricity was once a local, unstandardized thing. Then grids, then wholesale markets. Compute might follow the same path. Luna: And the key players are already positioning. NVIDIA obviously — they make the hardware. But also the cloud providers who own the data centers. If compute becomes a tradeable commodity, the value chain shifts. Lucas: Right now, the pricing power is with the GPU suppliers and the hyperscalers. A transparent market would reduce their margins but expand the total addressable market — more users could access compute at market-clearing prices. Luna: And it would create new financial intermediaries. Exchanges, clearinghouses, data providers. That's a whole ecosystem. Lucas: The article mentions that some startups are already developing 'compute futures' — standardized contracts for a certain number of gpu hours on a specific architecture. If that gains traction, you could see institutional investors treating it like a real asset class. Luna: One risk: if compute becomes commoditized, the huge capex spending by the tech giants might not yield the returns they expect. They're building out capacity assuming high margins. A liquid market could compress those margins. Lucas: That's a great point. It's the same dynamic that happened with oil — when it becomes a global commodity, the producers lose pricing power. The big winners are the low-cost producers and the traders. Luna: So who's the low-cost producer in AI compute? Probably the companies with the most efficient data centers and the best chip utilization. Maybe NVIDIA, but also the cloud providers who optimize their infrastructure. Lucas: And the traders — that could be a new breed of firms, like the energy trading desks at banks but for compute. We might see CME or ICE list a compute futures contract in the next few years. Luna: That would be a sign that the market has truly arrived. But for now, it's still nascent. The CNBC article is more exploratory than anything. Lucas: But that's exactly when you want to be paying attention — when the narrative is forming. The VIX is low, markets are high, but the structure of the economy is changing. Compute commoditization is one of those structural shifts. Luna: And it ties into the broader theme of this podcast — surviving downturns, buying dips, long-term resilience. If you understand where the puck is going, you can position yourself ahead of the crowd. Lucas: Right. For example, if compute becomes a commodity, then the companies that are pure-play AI infrastructure providers might face margin compression. But the facilitators of that market — exchanges, data providers, low-cost operators — could benefit. Luna: It's also a reminder that every new technology goes through a commoditization phase. Mainframes, PCs, cloud computing itself. AI is no different. Lucas: So the takeaway for investors: don't just buy the hottest AI stock. Think about where the value accrues as the industry matures. And keep an eye on the infrastructure layer, not just the applications. Luna: And with the market at all-time highs, it's a good moment to ask: are we paying for today's earnings or tomorrow's dreams? The compute commodity thesis is a bet on the future. Lucas: It is. But it's grounded in real demand. Companies are spending real money on compute. The question is whether that spending becomes more efficient through market mechanisms. Luna: Alright, that's a good note to end on. We'll be watching how this develops. Next week, we might look at the Fed — Warsh is expected to withhold the dot plot from the rate outlook, which is a big shift in transparency. Lucas: Yeah, that's a huge topic. The Fed removing forward guidance is a regime change in itself. But for now, thanks for listening. We'll be back soon.