Episode summary
Warsh’s Fed, Hyperscaler CapEx Chaos & Trading Compute Like Oil (LIVE w/ Steve Hou)
Monetary Matters
· · Hosts: George Noble, Jack Farley, Max · Guests: Steve Hou
Summary
Steve Hou argues that the AI infrastructure buildout is still being driven by real economics, not just narrative. He says data centers are profitable today, but the more important question is whether the returns stay high enough to justify ongoing borrowing and capex as rates remain elevated. In his view, the core tension is that token prices have trended down while GPU rental rates for H100s, B200s, and B300s have surged, showing strong demand for compute even as the revenue per token compresses.
He also pushes back on the idea that the AI trade is already exhausted. Hou says Anthropic’s S-1 and enterprise adoption data suggest the business can be highly profitable, with margins still positive and customers sticky once onboarded. He frames the real debate as forward-looking: whether growth, pricing power, and productivity gains can persist once more competition and capacity arrive.
On compute futures, Hou says Silicon Data’s curves imply a market that expects rental rates to stay high and the curve to slope upward at longer maturities, reflecting scarcity and reluctance to lock in long contracts too cheaply. He also says rising rates matter mainly insofar as they hit the broader financing and capex cycle, while physical constraints like power and land may matter even more.
Key takeaways
- Hou says data centers are profitable today, but continued borrowing depends on whether AI returns remain strong enough to justify new buildouts.
- Hou says GPU rental rates for H100, B200, and B300 chips are rising sharply even as token prices drift down.
- Hou says Anthropic’s S-1 supports the view that frontier AI labs can earn positive gross margins and attract sticky enterprise adoption.
- Hou says the compute forward curve is pointing to higher future rental rates, not a simple backwardated commodity pattern.
- Hou says the biggest constraint on hyperscaler expansion may be physical capacity—power and land—more than the Fed's current rate level.
Notable quotes
“they are profitable today. Clearly um I mean um that's why you're seeing there is this you know famous chart”
“the token index uh that's telling you the usage weighted price index of all the models that are being used out there”
“They are all surging and um in most recent weeks um you know they've been actually almost going vertical”
“it seems at least directionally clear that uh we've gone from an environment where where is a typically a downward sloping”
“the question is just you know does that you know actually really scale out right is that going to be most people is that going to be most companies”
How the episode opens
This episode is brought to you by Sarmia Partners, who manages a return to Tangibles ETF, ticker LES, the Smaya thematic ETF built for a return to tangibles. Learn more at sarmaetf.com and see the show notes for disclosures. Let's get into it, Max.
All right, thanks for kicking it off, Jack. And we are joined today by Steve Hoe, head of research at Silicon Data. Steve, thank you so much for joining us on this uh live stream where it's Fed day, but really hyperscaler earnings are sort of what's what the market is really focused on.
I mean, if you just look at the move from both Google and Microsoft, pretty high expectations. Um, look, I I personally think that the biggest question in markets right now is are data centers profitable and are they so profitable that the capex boom is going to continue? That's really what everybody wants to know and that's why we wanted to have you on, Steve, because you're out there uh trying to build uh futures talking to all of these data center providers to understand what the pricing of compute is going to be.
So, that's why we wanted to have you on today, Steve. Uh, I want to start with that big question. Are data centers profitable?
And are they so profitable that people are going to continue to borrow money to build them even if the Fed continues to hike rates?
Um I mean first of all they are profitable today.
From the episode’s captions: roughly the first two minutes, lightly cleaned. Keep watching on YouTube →
Topics
- fed rates
- gpu rental pricing
- compute futures
- token pricing crunch spread
- anthropic s-1 economics
Companies and tickers mentioned
- GOOGL
- GOOG
- MSFT
- NVDA
- CME
- Alphabet
- Microsoft
- Nvidia
- CME Group
- Anthropic
- Silicon Data
- Ornn
Watch the episode
Summary, takeaways and quote selection by Octagon. Quotes are short excerpts; the full conversation is in the episode.
More from Monetary Matters
-
AI is Heading for a Massive Oversupply: Top Hedge Fund Manager’s Prediction | David Orr -
Why Market Fragility is Building into a Crisis in 2027 | Cem Karsan & Noel Smith -
Inside a Short Seller’s Private Credit Investigation | Laks Ganipathi -
The End of Cheap Capital & What it Means for Markets | Russell Clark -
Breaking Down the Markets Massive Post-Midterm Moves with Andy Constan -
Michael Howell: Pain Now, Money Printing Later
Monetary Matters, this episode and its artwork belong to Jack Farley. Octagon is not affiliated with the show; this page is an independent summary.
