Mistral increases ARR to $400 million, builds its own AI data centers
French AI startup Mistral has increased its revenue twentyfold within a year, benefiting from growing demand among European companies and governments for alternatives to American technology providers. Co-founder and CEO Arthur Mensch put the company’s annualized revenue run rate at over 400 million dollars, compared to just 20 million dollars in the previous year. The Paris-based company, which was valued at nearly 12 billion euros last year, aims to achieve recurring annual revenue of more than one billion dollars by year-end.
The rapid expansion is taking place against the backdrop of growing concern in European executive suites about possible technological decoupling due to US President Donald Trump’s foreign policy. The European Union currently relies on foreign, predominantly American providers for over 80 percent of its digital services and infrastructure. Mensch stated that Europe had recognized that its dependence on US digital services had been excessive and now stands at a breaking point. Mistral positions itself as a provider of models, software, and computing power that function completely independently of American actors.
Unfortunately, Mistral does not play a major role in international rankings such as Arena.ai or Artificial Analysis with its AI models, so it still has some catching up to do. By way of comparison, Anthropic communicated around $10 billion in 2025, while OpenAI aimed to reach around $20 billion.
Infrastructure expansion in Sweden
The company announced an investment of 1.2 billion euros to build new AI data centers in Sweden, the first such facility outside France. In collaboration with EcoDataCenter, a facility with 23 megawatts of computing power is being created, which is expected to go into operation next year. Sweden offers itself as a location for energy-intensive AI chips because energy there is both carbon-low and relatively cost-effective. Mistral projects that the facility will generate revenue of over two billion euros over the next five years.
Despite the ambitions of American competitors OpenAI and Anthropic to go public soon, Mistral does not plan an IPO this year. The availability of external financing makes this step unnecessary for now, although an IPO is being considered for the coming years to ensure long-term independence. The company has expanded its customer base of major enterprise customers to over 100, including ASML, TotalEnergies, HSBC, and several European governments such as France, Germany, Luxembourg, Greece, and Estonia. Approximately 60 percent of revenues come from Europe, the rest from the US and Asia.
Vertical integration as business strategy
Mistral is increasingly pursuing a strategy of vertical integration by building and operating its own AI data centers rather than relying on American hyperscalers such as Amazon, Microsoft, and Google. This approach enables the company to finance chips that process customer workloads during the day and train new AI systems at night. At the same time, storing data on local servers provides European customers with security regarding data sovereignty. The provision of sovereign AI infrastructure is a priority for all stakeholders, Mensch emphasized.
Regarding market developments, the CEO expressed skepticism about exaggerated expectations for AI chatbots. Many enterprise customers had been disappointed by standard chatbot solutions because these had difficulty achieving a return on investment. The notion that a single system could cover all business processes, he called unrealistic. Nevertheless, Mistral sees potential in AI’s ability to understand user intentions and generate required user interfaces, while traditional business software providers would remain relevant due to their business-critical data.
