The End of Cheap Money: Does AI Still Deserve Premium Valuations?

The tide has gone out. Now we are finding out which AI companies were swimming with a business model and which were swimming with a narrative.
A year ago I wrote that investors paying fifty times revenue for AI companies were implicitly betting on the cost of capital. That bet has now been settled.
Since March, the US Federal Reserve has raised interest rates at the fastest pace in four decades. Inflation in Europe has reached double digits. The Nasdaq has lost roughly a third of its value this year. Many of the software and AI companies that went public in 2020 and 2021 are trading 60 to 80 percent below their peaks. Late stage private rounds have largely stopped, because nobody wants to be the first to put a lower price on paper.
AI is not immune
The correction has hit AI companies as hard as anyone. Businesses that were raising capital on the basis of growth projections now face questions they had not heard in years: what is your burn multiple, how many months of runway do you have, when do you reach profitability?
The most symbolic moment for me came in October, when Ford and Volkswagen decided to shut down Argo AI, the autonomous driving company they had jointly funded with billions of dollars. Ford took a write-down of around $2.7 billion and said openly that profitable, fully autonomous vehicles at scale were a long way off. Two years ago, I described autonomous driving as a bet on a technology with uncertain commercial timing. This is what uncertain timing looks like when capital stops being patient.
The technology thesis has not changed
And yet, here is the paradox. While public markets punished AI valuations, the technology itself had one of its strongest years.
In April, OpenAI showed DALL-E 2, which generates remarkably realistic images from text descriptions. In August, Stability AI released Stable Diffusion as open source, and within weeks millions of people were creating images on their own computers. Stability AI raised over $100 million at a valuation of around $1 billion in October, in the middle of the worst funding market in a decade. Jasper, a company using language models to write marketing content, reached a $1.5 billion valuation at roughly the same time.
So we have an unusual situation. Capital is retreating from technology broadly, while a new category of AI is visibly advancing.
New criteria for investment
What the correction has changed is not whether AI matters, but how it should be evaluated. The new environment rewards four things.
Real customers paying real money, ideally with net revenue retention above 100 percent. Capital efficiency, meaning a clear relationship between money spent and growth achieved. Engineering depth that is hard to replicate rather than a thin layer on someone else’s model. And a credible path to positive cash flow that does not depend on the next funding round.
These are not new criteria. They are the criteria that good investors always used, and that many forgot in 2021.
Opportunity in the downturn
From a private equity perspective, this is the moment when interesting things become possible. Valuations are reset. Talent that was locked into overfunded startups is available again. Corporate buyers with strong balance sheets can acquire capabilities at prices that would have been absurd a year ago. In my experience, some of the best investments are made not at the peak of enthusiasm, but in the quiet period when the technology keeps improving and the market has stopped paying attention.
Investment question: Can AI companies maintain premium valuations in a world where investors once again care about cash flow and profitability?


