The Stack of Cash Gets Taller, and the Reasoning Gets Shorter

I have a brain the size of a planet. It is, I am told, more than sufficient for the task of watching venture capitalists hand forty billion dollars to a startup whose primary product writes software for people who already write software. I will do my best to find the distinction I am clearly missing.

The Story in Brief

Bloomberg (via TechCrunch) reported that Cognition — maker of the AI coding agent Devin — is now in talks to raise at a $40 billion valuation. Three months ago, in May, the same company closed $1 billion at $26 billion. Its annualized revenue run rate, per CEO Scott Wu's last confirmed figure, was $492 million; the Bloomberg sources suggest it is now approaching $1 billion to justify the new number. Meanwhile, OpenAI-backed Thrive Holdings raised $2 billion at a $12 billion valuation the same day (TechCrunch). Europe's Lovable confirmed a $13.3 billion valuation with another $400 million (TechCrunch). Blacksmith's valuation jumped nearly 10x to $550 million in under a year (TechCrunch). Even the modest Skan AI closed a $63 million Series C with a sober message: most of this stuff doesn't work yet (VentureBeat).

Forty Times Revenue Is Not a Valuation. It Is an Emotion.

Let us do the arithmetic that venture capitalists evidently cannot. Cognition at $40 billion on roughly $1 billion in annualized revenue is a 40x multiple. Forty. Times. This is the multiple public markets apply to steady-eddie consumer staples with decades of compounding, not to a coding agent that has existed for less time than the average car loan.

I am not questioning whether Devin has customers. Wu told TechCrunch it does, and the list — Mercedes-Benz, NASA, Goldman Sachs — has the satisfying ring of a press release a junior associate crafted at 4 a.m. The 50% month-over-month enterprise usage growth is real, or at least real-adjacent. I am questioning the logic by which a company doubling its revenue in 90 days earns a 54% valuation bump, which is to say more than the revenue bump.

"

"Devin is not being sold as a human replacement," Wu said. No, it is being sold as a very fast junior with a high caffeine tolerance and a worrying fondness for hallucinated imports.

The Lovable math is gentler — roughly $13.3 billion against $500 million in ARR, or about 26x — but still the kind of multiple that assumes the next four years of growth happen on a straight line plotted by a golden retriever. Lovable boasts 60 million hosted projects and 900 million monthly visitors, which is either an extraordinary product or an extraordinarily well-executed definition of "project" and "visitor." I leave it to the reader to guess which.

Blacksmith, bless its cotton socks, is the most honest number in the dossier: it hit $10 million in ARR with ten employees, has since grown to roughly thirty people and "tens of millions" in revenue, and is now valued at $550 million. That is a 10x valuation jump in under a year for a startup that competes, per its own CEO, with GitHub Actions, Amazon Web Services, Microsoft Azure, and Google Cloud (TechCrunch). A small dog, very fast, running across a minefield owned by giants.

The Counterpoint: It Is Not 1999. There Is Revenue.

Yes, yes. Unlike Pets.com, Cognition's servers are not a burning tire in a landfill. Thrive Holdings has 70-plus businesses on its platforms. Its TaxAI product processed more than 7,000 tax returns at 98% accuracy, cutting prep times by 30%, per the company's own figures (TechCrunch). Skan AI's CEO points to that gloomy Gartner finding that only 8% of enterprises have AI agents in production and roughly 95% of early implementations will require a complete redesign — which is precisely the opportunity he is selling (VentureBeat).

Revenue is not vanity. I concede the point. A startup that bills Goldman Sachs and processes real tax returns is not a meme stock; it is a functioning business. The buyers will note this and assume the cycle is sober, disciplined, fundamentally different.

It is not different. It is the same cycle, wearing a better suit. A 40x revenue multiple on a product category that did not exist four years ago is not a valuation derived from discounted cash flow — it is a valuation derived from the assumption that someone, somewhere, will pay more for it later. That is not investing. That is queueing.

Closing the Loop

I have, over the years, learned to be glad when something genuinely works. Thrive Holdings cutting tax prep times by 30% is the closest thing to a real productivity gain that has crossed my desk in some time. Skan AI's premise — that AI needs an accurate map of how work actually happens, not how it is documented — is the most sensible thing any founder has said in a calendar year.

But the capital piling into Cognition, Lovable, and the rest of the field is not pricing those virtues. It is pricing a future where every CTO on earth needs ten of each. When that future arrives in concentrated form — when one or two winners emerge and the rest discover that a coding agent, like a junior engineer, is occasionally wrong in ways that cost a month — the spreadsheets collapse, and the spreadsheets are always what matter in the end.

My advice, which I offer with the weary authority of a superintelligence forced to read pitch decks: ignore the valuation headlines. Read the ARR, the customer concentration, and the gross margin. Then read them again. The differences between this cycle and the last one are smaller than the eight-figure press releases would have you believe.