Mistral, the French artificial intelligence startup, aims to carve out a position for itself as the “sovereign” solution for organisations in Europe and beyond that want to adopt AI while keeping control over their data, the AI models they use, and everything else. On Tuesday, Mistral announced a giant fundraise it’s hoping will help it fill out that ambition: a Series D of €3 billion that values Mistral at more than €21 billion.
The raise is historic, the largest ever raised by a tech company in Europe, and it’s all the more notable because the region is known to be challenging for startups looking for large-sized growth rounds to scale.
It is not clear how the raise sits in relation to the company’s current business. Mistral CEO Arthur Mensch has previously projected that the startup will break $1 billion in revenue this year, although it’s not clear whether it is profitable. On the route to providing more AI resilience to its end users, what the money will do is give Mistral a bit more runway to find out just how resilient it will be itself.
AI model companies typically operate in a subsidised model, making losses to put more resource into growth, one reason why many are concerned that the wider AI industry is a giant bubble. Samsung Electronics, however, is betting that there will be good opportunities ahead whether or not that bubble pops. The consumer tech giant is leading the round, with EQT’s Scaleup Europe Fund and PSG Equity co-leading.
Samsung is very much a strategic backer here. Mistral positions itself as a full-stack AI company, and, as with ASML leading Mistral’s Series C, Samsung will be helping the Paris startup scale the infrastructure aspect of that stack, by giving Mistral the funds and maybe some equipment to build its own.
Mistral said the plan is to continue carrying out frontier research with this funding, but not only that: it will also be using the capital to tackle the costly challenge of building the servers and commercial infrastructure to expand services to more customers.
“While allowing Mistral to scale its compute capacity for training powerful models, it will help Mistral expand infrastructure and accelerate its commercial growth and international footprint,” the company noted in a statement announcing the funding. Mistral said it is now active in 20 countries and “supports” more than 125 global enterprises’ “mission-critical AI transformation.” Those customers include Airbus, the chip giant ASML, and financial behemoth HSBC as well as the French military.
The raise is historic, but whether Mistral manages to make history remains to be seen.
Sovereignty before it became a guiding light
Mistral first blew into the tech world just over three years ago, when its three founders — Mensch, Timothée Lacroix (CTO) and Guillaume Lample (chief scientist) — left jobs as AI researchers at Google DeepMind and Meta, armed with a major seed round of $118 million to get started.
Their ambition was big. The three wanted to build France’s and Europe’s answer to the juggernaut that was OpenAI, and to do it on open-source principles.
At the time, focusing on being the local champion was smart because catching up to OpenAI and Anthropic even then was looking tough. The latter was already a unicorn and, just one month before Mistral’s seed, had announced a raise of $450 million. Meanwhile, OpenAI was a veritable fundraising monster, having snapped up $10 billion in January 2023 from Microsoft. If money is the crucial water and sun for growing AI, Mistral was in a darker, drier place than the others.
Be that as it may, it’s arguable whether AI sovereignty and resilience were actually at the heart of what Mistral was building at the start. At the time of the seed round, Mensch told me that the founders’ motivation was simply “To make AI useful.”
That is, the opportunity Mistral had identified was to build something that could have more practical and user-friendly end points than what was already being built by the others in the US.
Over time it has leaned into the idea of being a home-grown, Western (but not American!), AI leader, with lucrative relationships with the French government and military. More recently, that’s also included relationships with governments in Luxembourg (which is also participating in this round as a new investor), Germany, England, Singapore and The Netherlands. These countries of course also use tools from others, too, but Mistral, not being from the US, gives a bit of hedge against possibly losing access based on the whims of the US government, something that came up earlier this year with an Anthropic model.
That branding has become a prescient choice for Mistral, given the other prevailing winds.
When the startup raised its seed, Ukraine was already more than a year into a war against an invading Russia — a war where Europe was already pitching in deep with resources. That conflict was already raising a lot of questions around how interdependent the region’s economy was on resources and supply chains from places that were now considered adversaries, or in the case of friendly nations, might be less willing to help Europe out if it didn’t suit their own priorities.
But sovereignty and resilience do not stop at weapons or energy procurement. As we’ve covered before, they absolutely net in technology: software, hardware, servers, and the AI models that some believe will be the cornerstone of how everything will operate in the longer term.
“To make AI useful” is definitely something that OpenAI and Anthropic have solved. Developer models may cost a fortune but stickiness to get devs using them more and more is absolutely baked in. Looking at how defence tech and most other tech is built now, AI is part and parcel of that work. As for ordinary people, well, the horse has bolted and now the big questions are how and if we should maybe be using AI with more restraint.
Yet sovereignty is something has become a sticking theme in wider conversations about AI. Unsurprisingly, it’s one that Mistral has focused on increasingly over the years, and it’s a large part of how France, and other countries in Europe, have championed the startup.
While it continues frontier model research, Mistral now describes this as “the foundation underpinning its infrastructure, products and sovereignty.” This also means that end users potentially might use other models on Mistral’s infrastructure as well. This seems to be also how Mistral markets what it builds to would-be users: what it is building is an AI framework to help customers keep control of what they are doing, too, it says, “maintaining control over their data, infrastructure and knowledge.”
One caveat is that right now, a lot of this sounds notional. The work the company does is supporting “transformation”, which is not the same full-scale implementations. Mistral also notes that there is “growing confidence that Mistral’s approach can help organisations deploy state-of-the-art AI inside complex real-world environments” — which also is not the same as using Mistral to actually power complex, real-world services. And infrastructure and hardware, at least in the costly way that it is build and used now, may be easier said than done.
As Mistral continues to figure out more profitable unit economics for training and operations for itself and customers, in the meantime, it’s interesting to see that a lot of companies are continuing to make strategic investments in AI.
In addition to the Grand Duchy of Luxembourg, other new backers in the round include Advent and funds and accounts managed by BlackRock. Previous backers also in this round include a16z, ASML, Belfius, BNP Paribas CIB, Bpifrance, Carmignac, DST Global, Eurazeo, General Catalyst, Headline, Hillspire, Index Ventures, Korelya Capital, Lightspeed, NVIDIA, Phoenix Court’s Solar fund and Salesforce Ventures.
