London, 13 August 2026 — EBM Newsdesk Analysis — By Katie Winearls
Databricks closed a $5bn funding round on Thursday at a $190bn post-money valuation, having crossed a $7bn revenue run-rate and grown more than 80% year on year in its second quarter. Coatue led, alongside Blackstone, Abu Dhabi’s MGX and T. Rowe Price, with Sixth Street Growth joining as a new investor.
The trajectory is steep even by current standards. In December the company was raising at $134bn on a $4.8bn run-rate. Eight months later it is worth $190bn on $7bn. The round was signed at $188bn in July and closed higher.
Join The European Business Briefing
New subscribers this quarter are entered into a draw to win a Rolex Submariner. Join 40,000+ founders, investors and executives who read EBM every day.
SubscribeNone of that is the interesting part.
The Margins Are Going Down
Ali Ghodsi told CNBC in June that Databricks’ gross margin was falling and would fall further. He declined to say what it currently is.
That sentence should stop anyone valuing software companies. The entire model rests on near-zero marginal cost: build the product once, sell it a million times, run gross margins around 80% forever. It is why software trades at multiples manufacturers never see.
Consumption-based AI does not work that way. Ghodsi’s explanation is that agents generate far more queries than humans do, and every query costs compute. Growth and cost now rise together. Databricks is getting bigger and, per unit, less profitable.
So a company being valued at roughly 27 times revenue on software assumptions is quietly converting into something closer to a utility — one that buys its main input at market price and resells it. Utilities do not trade at 27 times revenue.
What Customers Are Actually Doing
The most useful detail in the announcement is a product rather than a number.
Unity AI Gateway, one of the three things this money is funding, exists to warn companies when they are close to exhausting their AI budgets. That is not a growth feature. It is a cost-control feature, and it is being built because customers asked for it.
Ghodsi’s description of the shift is blunt: enterprises have stopped what he calls tokenmaxxing — encouraging staff to use as much AI as possible — and moved to value-maxxing, meaning optimisation. They want frontier models for hard problems and cheap ones for routine work.
Read that alongside the valuation. The customers are entering a cost-discipline phase at precisely the moment investors are pricing in unlimited consumption growth.
The European Position
Nowhere again, and the pattern is becoming monotonous.
More than 60% of the Fortune 500 runs on Databricks, and a substantial share of Europe’s large caps do too. Every enterprise AI infrastructure decision made in a German, French or Nordic company this year commits budget to an American vendor for a decade, because migration costs make these choices effectively permanent.
There is no European alternative at this scale, and there will not be one, because Europe’s growth capital does not write $5bn cheques into private software companies. Note also who does: Coatue, Blackstone, T. Rowe Price and MGX — American funds and Gulf sovereign money. The Chips Act addresses fabrication. Nobody in Brussels is addressing this.
The Verdict
My view is that Databricks is the healthiest business in the AI boom and its valuation is still built on an assumption the company itself is contradicting.
The fundamentals are real and unusually so. Free cash flow positive, net dollar retention above 140%, $1.7bn of AI product revenue, and 80% growth at $7bn scale. Compared with Anthropic’s projected $2tn listing or a robotics IPO drawing $1.2tn of orders, this is a company with actual customers paying actual money.
But 27 times revenue is a software multiple, and Ghodsi has just told everyone the software economics are eroding. If gross margin settles at 60% rather than 80%, the same revenue is worth materially less, and no announcement has addressed where it lands.
The signal worth watching is Unity AI Gateway. When your customers ask you to build them a tool for spending less on your product, the consumption curve everyone is extrapolating has already started to bend. Ghodsi has also said Databricks is unlikely to list before Anthropic or OpenAI — which is a sensible position for a private company that would rather not have this discussion quarterly.
Related Analysis
