French AI Startup Mistral Seeks €3 Billion Funding, Valuing Company at €20 Billion

If the round closes, Mistral would take its total funding raised to about €6.5 billion, including both debt and equity—up from last year’s €1.7 billion Series C that put the company at an €11.7 billion valuation.
Mistral’s in-house compute effort is branded “Mistral Compute” and is meant not only to serve Mistral’s European customers but also to provide compute access for third parties.
To fund its data-center buildout, Mistral previously secured $830 million in debt funding (earlier this year), with plans to build capacity up to 1 gigawatt (GW) across Europe by 2030.
Mistral’s recent funding history and ownership details add context to why the raise matters: it completed a €600 million Series B in June 2024 (valued at €5.8 billion) and then a €1.7 billion Series C in September 2025 led by ASML, where ASML took an 11% stake for €1.3 billion.
The reported investor roster underscores the push for “compute + capital” backing: Mistral’s backers include Nvidia, Andreessen Horowitz, General Catalyst, and France’s public investment bank Bpifrance.
French AI startup Mistral is in talks to raise €3 billion at a valuation of about €20 billion, according to GuruFocus and Sifted. That would nearly double its €11.7 billion valuation from just nine months ago and push its total funding — debt and equity combined — to roughly €6.5 billion.
The raise signals a sharp turn for Mistral. The Paris-based company, founded in 2023, built its early reputation on lean, efficient AI models. Now it is racing to build data centers, buy thousands of Nvidia chips, and compete with the world's biggest AI labs.
Mistral's funding history shows how fast the company has scaled. It raised a €600 million Series B in June 2024 at a €5.8 billion valuation. Then in September 2025, chipmaker ASML led a €1.7 billion Series C, taking an 11% stake for €1.3 billion and valuing the company at €11.7 billion, Sifted reported.
The company's backers now include Nvidia, Andreessen Horowitz, General Catalyst, and France's public investment bank Bpifrance. CEO Arthur Mensch has said that "scaling our infrastructure in Europe is critical to ensure AI innovation and autonomy remain at the heart of Europe."
In March 2026, Mistral secured $830 million in debt financing from a seven-bank consortium that included BNP Paribas, HSBC, and Bpifrance, according to Let's Data Science. The money went toward buying 13,800 Nvidia GB300 GPUs for a Paris data center. Mistral plans to reach 200 megawatts of capacity by 2027 and 1 gigawatt across Europe by 2030.
Mistral has branded this effort "Mistral Compute." It is not just for Mistral's own models — the company plans to sell compute access to outside customers too. Industrial giants Airbus and BMW are already using it to run sensitive engineering simulations, Let's Data Science noted.
A €20 billion valuation would be a record for a European AI company. But it still looks small next to U.S. rivals. OpenAI recently closed a raise at a $300 billion valuation. Anthropic reached roughly $61 billion. Mistral's entire proposed war chest of €6.5 billion is less than what those companies raise in a single round, Sifted noted.
Still, Mistral's commercial growth has been sharp. The company reported about $400 million in annual recurring revenue in early 2026 — a roughly 20-times jump from the year before — and is targeting $1.16 billion in revenue for the full year, according to Seeking Alpha.
President Emmanuel Macron has put Mistral at the center of a €100 billion French AI investment push. The logic is simple: European companies — especially in defense and government — cannot legally store sensitive data on U.S. platforms like AWS or Azure. Mistral's European-owned compute gives them an alternative, Sifted reported.
Some U.S. analysts push back on this framing. They argue that because most of Mistral's funding goes straight to Nvidia for chips, the round effectively cycles American capital back to American hardware makers — keeping Europe dependent on U.S. silicon even as it tries to break free, according to Let's Data Science.
Publishers
20
Articles
22
Reach
42