Markets are getting AI right

The paper's late-July analysis argues that investors have stopped buying the category wholesale and started buying the underlying businesses — a subtle distinction that changes the gravity for everyone building in the space, from chip designers to vertical SaaS founders pricing their next round.

What the FT is actually saying

The Financial Times piece, published July 31, carries the confident title "Markets are getting AI right." Read closely, that single sentence is doing a lot of work. It implies that the whiplash era of AI valuations — where a model demo could reroute billions overnight and a missed earnings call could vaporize the same — is finally giving way to something colder and more discerning. Capital, the argument goes, is no longer rewarding the label. It's rewarding the line items.

The timing is deliberate. We're past the breathless phase of the cycle, past the moment when every startup with "AI" in its deck commanded a premium multiple regardless of what it actually shipped. What comes next is the part that separates the durable businesses from the slides. Hemorrhaging cash on a clever demo will no longer carry a company through a Series C — the math has to work, and the market, finally, is doing the math.

The visual record from the trading floor

AP's market wrap from the same day captured the mood in fragments rather than fireworks. Photographers documented the usual global mosaic: SK Hynix's stock price ticking across screens in a Seoul dealing room, the KOSPI board updating in the background, the yen-to-dollar exchange rate flashing on an electronic board in Tokyo, traders hunched over terminals on the New York Stock Exchange floor the day before. No single dramatic move in any frame. Just the steady, grinding work of capital finding a price — the visual opposite of a mania.

For operators on the other side of the spreadsheet, that visual restraint is the real story. It's the difference between fundraising on category narrative and fundraising on unit economics. Companies whose infrastructure — chips, data pipelines, verticalized applications — maps cleanly to a paying customer will keep their runway. Those whose pitch decks lean on the word "intelligence" alone will feel the ground shift under them in the next round.

It's the kind of market that punishes sloppiness and rewards boring excellence. Margins matter. Retention matters. The mechanics of recurring revenue matter.

What to watch next

The proof will show up in the data before it shows up in the headlines. Watch the hyperscalers' capex guidance in the next earnings cycle. Watch the chip vendors' book-to-bill ratios for any sign of demand cooling. Watch the late-stage private rounds closing — or not closing — across Q3 and Q4. Either the FT thesis hardens into the new orthodoxy, or it becomes another comfortable narrative that markets eventually shrug off when the next catalyst lands.