Europe's Biggest Round Ever: Mistral Raises €3B Led by Samsung to Build the Sovereign AI Stack

Three years after its launch, Mistral AI has closed the largest equity funding round in the history of European technology. The Paris-based company announced today that it has raised €3 billion in a Series D at a post-money valuation of more than €21 billion (roughly $24–25 billion), with Samsung Electronics leading the round. The EQT-managed Scaleup Europe Fund and existing investor PSG Equity joined as co-leads, alongside new investors Advent, funds and accounts managed by BlackRock, and the Grand Duchy of Luxembourg. A long roster of existing shareholders — a16z, ASML, BNP Paribas CIB, Bpifrance, DST Global, General Catalyst, Index Ventures, Lightspeed, NVIDIA, Salesforce Ventures and others — participated.

To put the number in perspective: Mistral’s seed round in June 2023 was €105 million, then Europe’s largest-ever seed. The company has now raised a single round nearly thirty times that size, and it did so as a European company whose flagship pitch is that customers never have to depend on a single American vendor’s roadmap, pricing, or availability.

What the money is for

Mistral says the Series D will fund three things in parallel: frontier research as the foundation of everything it ships, expanded compute capacity for training larger models, and accelerated commercial growth and international expansion. The company now operates across 20 countries and counts 125+ global enterprises as customers for mission-critical AI — Airbus, ASML, and HSBC among them.

The round is also a strategic endorsement spanning three continents. Samsung, the world’s largest memory and chipmaker outside the US-China duopoly, now leads the cap table, just as ASML — the Dutch lithography giant — led Mistral’s Series C. Mistral has effectively attracted the two most strategically important equipment and manufacturing companies in the global semiconductor supply chain as lead investors. That is not a coincidence: both companies have a direct interest in European AI workloads demanding European-built silicon and European-run infrastructure.

The thesis: sovereignty as a product

The most interesting part of the announcement is not the number but the framing. Mistral argues that the first wave of generative AI centered on one question — who can build the most powerful model — and that the market has moved on. The question enterprises and governments now ask, according to the company, is how to harness frontier AI for mission-critical work without surrendering control over the infrastructure and intelligence loop.

Mistral’s answer is what it calls “the sovereign AI layer,” built on a full stack that no other AI company in the world offers end-to-end: open-weight models, the infrastructure and compute those models run on, and the products that bring them into production. Sovereignty, in Mistral’s definition, means control across four dimensions:

• Data that stays inside the organization’s boundaries
• Models that are controllable and customizable
• Compute that is private and predictable
• Systems in production that are fully controllable and auditable

The open-weight approach is central to this pitch. Because the model weights are available, customers can run, fine-tune, and audit the models inside their own walls, without exposing their most valuable data, workflows, and institutional knowledge to anyone outside. It is the architectural opposite of the closed-API model that dominates the US frontier labs.

From model lab to “neocloud”

What makes the €21 billion valuation defensible — or at least arguable — is that Mistral is no longer just a model lab. Over the past year it has been aggressively building physical infrastructure:

• €830 million in debt financing (March 2026) for a Paris data center cluster with 13,800 Nvidia GB300 GPUs and 44 MW of capacity
• €1.2 billion committed with EcoDataCenter for an AI-optimized facility in Borlänge, Sweden, operational from 2027
• A Microsoft partnership announced in July 2026 expanding European data center capacity for sovereign cloud deployments
• A stated target of 200 MW of capacity across Europe by 2027

Add the product layer — the Mistral API, the Le Chat assistant, coding agents under the “Vibe” brand, the Forge fine-tuning platform, and vertical solutions for finance, public institutions, manufacturing, and energy — and Mistral’s ambition is closer to “sovereign neocloud plus frontier lab plus enterprise software vendor” than to a pure research company. Trending Topics went as far as calling the round a step toward making Mistral “the world’s biggest neocloud.”

Why Samsung

Samsung’s lead position is the detail worth watching. The Korean conglomerate gets three things from this deal: exposure to European AI demand, a foothold in a market that is actively legislating for technological autonomy, and a potential anchor customer and partner for its memory and foundry businesses as AI inference workloads in Europe scale. For Mistral, Samsung brings capital that is patient and strategic rather than purely financial, plus deep expertise in the hardware layer that sovereign AI ultimately depends on.

There is also a defensive logic. European policymakers have spent two years talking about “AI sovereignty” while the region’s compute remained overwhelmingly American. Mistral is now the single best-capitalized vehicle for turning that rhetoric into physical infrastructure — and its investor syndicate (ASML, Samsung, BNP Paribas, Bpifrance, Luxembourg) reads like a coordinated European industrial alliance with Asian manufacturing muscle attached.

The caveats

Skepticism is warranted on at least three fronts. First, €21 billion is a rich multiple for a company whose disclosed customer count is 125+ enterprises; revenue figures were not part of the announcement. Second, “sovereign AI” is partly a marketing category — plenty of Mistral’s stack still runs on Nvidia GPUs, and its Paris cluster is built with American silicon. Third, the competitive gap is real: US labs are spending hundreds of billions on training compute, and Mistral’s open-weight models have historically trailed the absolute frontier even as they win on cost, licensing, and deployability.

But the market has clearly decided that the combination Mistral offers — frontier-adjacent open weights, European data residency, and now a balance sheet to build compute — is worth a European-record price. Whether it earns that valuation will depend on execution in data centers over the next 24 months, not on press releases.

What it means

For the AI industry, the round is a signal that capital is no longer flowing only toward the biggest training runs. A second axis of competition has opened: who can deliver frontier capability with guaranteed control. For Europe, it is the first time the continent’s AI champion has the resources to build at infrastructure scale rather than rent it. And for enterprises watching from the sidelines, it means the “build on US hyperscalers or build nothing” era is ending — though whether Mistral’s sovereign stack actually delivers on its auditable, controllable promise will take years to verify.

One round does not close a competitiveness gap. But €3 billion, led by Samsung, on top of an ASML-led Series C, gives Europe’s most-watched AI company a genuine runway — and gives the sovereign-AI thesis its first serious stress test.