Mistral raises $3.5 billion and becomes Europe's biggest bet on open AI

The French Mistral AI announced this Tuesday a Series D round of $3.5 billion, which nearly doubled its valuation to more than $24 billion. The information was reported by Crunchbase News, which described the raise as the largest AI round ever made by a European company, surpassing Mistral's own previous record.

The round was led by Samsung Electronics, with participation from the Scaleup Europe Fund (managed by EQT) and existing investor PSG Equity. With this, Mistral has raised $7.5 billion since its founding in 2023. The investment comes almost exactly one year after the $2 billion Series C, which valued the company at $13.7 billion.

Why a VC round matters to those who write code

A funding round is usually a topic for those who follow the market, not for those with the editor open. But in this case the money buys something concrete for developers: breathing room to keep an open alternative alive in a market dominated by closed APIs.

According to the statement cited by Crunchbase, the capital will "significantly expand frontier research" at Mistral and help "scale infrastructure and accelerate" commercial growth and international presence. The company currently operates in 20 countries and has more than 125 global corporate clients, including Airbus, ASML, BMW, and HSBC.

The point that matters to those who build software lies in the differentiator that Mistral itself raises against its American rivals:

Unlike many of its US competitors, Mistral promotes greater control for companies, offering models they can customize and run on their own systems, instead of depending entirely on a third-party cloud provider.

>

-- Crunchbase News

That's the axis of the dispute. On one side, OpenAI, Anthropic, and Google, whose cutting-edge models you access via API and run on their infrastructure. On the other, Mistral, which bets on models that the company can download, fine-tune, and run wherever it wants, including on-premise.

Open vs. closed-source: what's at stake in practice

For the Brazilian developer, the choice between an open model (open weights, as much of the Mistral family is) and a closed-source one (API-only) affects some very concrete variables:

| Criterion | Open model (e.g., Mistral) | Closed-source (e.g., GPT, Claude) | |---|---|---| | Where it runs | Your infra, own cloud, or third-party provider | Vendor's infra, via API | | Sensitive data | May not leave your environment | Travels to an external server | | Customization | Fine-tuning and weight adjustment | Limited to what the API exposes | | Cost | You pay for compute (GPU) | You pay per token | | Dependence | Lower vendor lock-in | High lock-in |

For those handling regulated data in Brazil, the ability to run the model within one's own environment speaks directly to requirements under the LGPD (Brazil's data protection law). A bank or a healthtech that can't send customer data to a US-based API now has in Mistral a heavyweight candidate, not a weekend experimental project.

The cash reinforcement also reduces a risk that always weighs on the decision to adopt an open model from a smaller vendor: that of the project being left without maintenance. A company that just raised $3.5 billion and has Samsung, EQT, and PSG on its cap table offers a guarantee of continuity that carries weight in architecture decisions.

The map of frontier labs

Mistral holds the position of Europe's most valuable foundation model company. According to Crunchbase's Unicorn Board, there are currently seven private frontier labs valued above $20 billion (eight, counting China's ByteDance, which also builds its own model). Aside from Mistral, all the others are in the US or China.

In other words: in terms of capital weight, the frontier model market remains concentrated around two poles. Mistral is the only relevant European counterweight, and that's why it becomes the practical reference for those in Brazil who want to diversify vendors beyond the US/China axis for technical, regulatory, or data sovereignty reasons.

The move follows a turnaround in European venture capital as a whole. Still according to Crunchbase, Europe had its strongest funding quarter in four years in the second quarter, with $24 billion raised, up about a third quarter-over-quarter and two-thirds above the $14.4 billion of the second quarter of 2025. Other big AI deals closed this year in the region, such as Isomorphic Labs ($2.1 billion, a Google spinoff) and data center provider Nscale ($2 billion at a $14.6 billion valuation).

What remains open

The statement mentions "expanding frontier research" and "scaling infrastructure," but doesn't detail which models or capabilities are coming next, nor a timeline. For those planning their stack, the practical takeaway is that the bet on a European-origin open model has gained a solid financial anchor, which reduces the risk of abandonment, but the technical decision still depends on the usual factors: benchmarking against your use case, the real GPU cost of running the model in production, and the size of your team to operate that infrastructure. A vendor's cash reserves don't replace testing in your own pipeline.