9 SEP 2026 — Mistral has raised €3 billion in a Series D led by Samsung Electronics at a post-money valuation above €21 billion, which the company calls the largest equity round ever completed by a European technology company. The money is for compute, and the stated ambition is up to one gigawatt by 2030. Three billion euros does not buy a gigawatt.
The round, and who is in it
Samsung Electronics leads, with the Scaleup Europe Fund managed by EQT and PSG Equity as co-leads. New investors include Advent, funds managed by BlackRock, and the Grand Duchy of Luxembourg as a state investor.
The valuation is roughly double the €11.7 billion Series C completed a year earlier, at a company founded about three years ago. The stated uses are compute expansion, frontier model research and international growth.
Samsung leading is the detail to notice. A memory and foundry supplier leading a model developer's round is a customer relationship dressed as an investment, and it tells you which way the compute is expected to flow.
The gigawatt arithmetic does not close
We covered a directly comparable figure this month. TCS's HyperVault committed up to 700 billion rupees, about 7.4 billion dollars, for a campus with a stated capacity of one gigawatt — and we noted there that an announced gigawatt is a design ceiling rather than a build.
Mistral's €3 billion is roughly €3.5 billion in dollars. Against a comparator that put more than twice that against the same nameplate figure, the round funds a portion of the ambition and not the ambition.
That is not a criticism of the raise. It means one of three things: the gigawatt is a 2030 target dependent on further rounds, or on partner capital, or the capacity will be leased rather than owned. All three are ordinary, and none of them is what "raises €3 billion to build a gigawatt" conveys.
A model lab becoming an infrastructure company
One framing of this round is a strategic pivot toward being a compute provider rather than only a model developer. That reading has support in where the money is going.
It also changes what kind of company this is to value. A frontier lab is valued on model capability and the licensing that follows; an infrastructure operator is valued on utilisation, power contracts and depreciation schedules. Those are different businesses with different multiples, and €21 billion is a lab's number.
Sovereign positioning reconciles the two. European buyers who want inference to run on European soil under European law are buying jurisdiction as much as capability, and jurisdiction is an infrastructure property. Luxembourg appearing on the cap table as a state investor is that thesis made explicit.
The capability question the round does not answer
A raise prices expectations rather than performance. On the independent scoreboard the frontier is currently contested between American and Chinese labs, and we have spent this month noting that even the launches driving markets have not always moved the neutral measurements — Asian chip stocks rallied on a model that scored level with its predecessor.
Mistral's case does not depend on topping that scoreboard. It depends on being good enough, open enough and European enough for a buyer whose constraint is regulatory rather than technical, which is a real and defensible market.
The risk in that position is that it is a policy bet. If European procurement preferences soften, or if a larger lab offers a compliant regional deployment, the moat is administrative rather than technical and administrative moats move quickly.
Doubling a valuation in a year
€11.7 billion to more than €21 billion in twelve months is a near-doubling, and it happened without a public revenue disclosure to anchor it. That is normal for private AI rounds and it is worth naming rather than passing over.
A round this size prices the option on a market that does not fully exist yet: European buyers required by regulation or preference to keep inference inside a jurisdiction. If that market forms at the size investors expect, the valuation is defensible. If procurement rules relax, it is a bet on an administrative condition.
The presence of a sovereign investor on the cap table cuts both ways here. It is validation of the thesis and it is also an interested party, since a state that invests in a national champion has reason to prefer procurement rules that favour it.
What would make the gigawatt credible
A named site and a grid connection date, which is the number that separates announced capacity from built capacity in every project we have looked at. A power purchase agreement, because a gigawatt in Europe is a serious electricity conversation before it is a serious computing one.
And whether Samsung's involvement extends to supply commitments on memory, since that is the constraint that has bound this cycle and the one a strategic lead investor is uniquely able to relieve.
What €3 billion buys is several years of runway and a credible claim on European demand. The gigawatt is a 2030 sentence, and 2030 is at least two more rounds away.
