France's Mistral AI has announced it will build large-scale AI infrastructure across Europe, creating an environment where companies can directly choose the location of their data to strengthen European AI sovereignty.
Imagine you are working on a highly confidential project at your company, and you decide to use AI to organize your materials, only to be told that your data will be transmitted across borders to servers in another country. You might worry about security, and you would likely prefer the data to be processed safely in a location protected by your own laws. This concern is vital not only for individuals but also for governments and corporations. Recently, the French AI startup Mistral AI unveiled an ambitious plan to address these issues and defend Europe’s “AI sovereignty.”
Why Is This Important?
Current AI technology is largely dominated by US corporations. This means that for European governments and companies to use the latest AI, they often have to send their data abroad. Entrusting data to external parties makes it difficult to manage or protect it in the way we want. Microsoft Vice Chair Brad Smith emphasized that “Europe must be able to access the world’s most capable AI technology without surrendering control over its data, operations, and digital future.” This move by Mistral AI signifies a commitment to helping European companies utilize top-tier AI without losing data sovereignty.
Simplifying the Concept: AI and Digital Territory
Think of AI as a “smart assistant.” In the past, this assistant could only perform well if it were constantly on a business trip to the United States. Now, Mistral AI is trying to bring this smart assistant into the European home front.
Mistral AI has announced three main strategies.
First is “regional inference endpoints.” “Inference” refers to the process by which AI generates results based on what it has learned. This feature is like having a button that lets us choose whether to have the work done in our own living room (Europe) or in an office on the other side of the planet (the US). Mistral AI stated that it would allow customers to choose where their AI workloads are performed.
Second is the construction of large-scale infrastructure. Mistral has released a roadmap to build 200 megawatts of infrastructure by 2027 and 1 gigawatt (1,000 megawatts) by 2030 across Europe. To achieve this, they plan to utilize thousands of NVIDIA Vera Rubin GPUs. “GPUs” are high-performance semiconductors specialized for AI calculations. In other words, they are building massive offices and installing cutting-edge equipment across Europe where these assistants can work.
Third is an open ecosystem. Mistral has begun supporting the use of open models created by other companies alongside their own on their platform. It is like being free to choose assistants from various firms.
Current Situation
Currently, Mistral AI is a Europe-based AI startup that already has a close partnership with Microsoft. Through this partnership, customers can operate AI not only in the public cloud but also in fully disconnected environments. A “cloud” refers to renting server resources over the internet, while a disconnected environment means creating a space where AI can be used safely within one’s own building without an internet connection.
What Lies Ahead?
Mistral AI’s goal will not be achieved in the short term. Securing 1 gigawatt of computing power by 2030 is a monumental challenge. However, if this plan succeeds, Europe will possess a robust “digital sovereignty” that no longer relies solely on the technology of AI superpowers. Just as technology advances, the concern over who manages our data and where will become even more critical. Watching how European companies choose their technology and manage their data will serve as an important indicator for the future of the global AI ecosystem.
References
-
[Mistral AI wants to build 1 gigawatt of European compute by 2030 — and lock in customers now. VentureBeat](https://venturebeat.com/infrastructure/mistral-ai-wants-to-build-1-gigawatt-of-european-compute-by-2030-and-lock-in-customers-now) - Microsoft and Mistral expand strategic partnership to give enterprises and regulated industries frontier AI they can control - Source
-
[Microsoft and Mistral expand sovereign AI partnership in Europe Digital Watch Observatory](https://dig.watch/updates/mistral-ai-partnership-europe) - Microsoft Expands Mistral Partnership for European AI - Thurrott.com
-
[What is Mistral AI? IBM](https://www.ibm.com/think/topics/mistral-ai)
- 200 megawatts
- 500 megawatts
- 1 gigawatt
- It provides faster AI models
- It allows customers to choose the processing location for their data (Europe or the US)
- It makes AI free to use
- To monopolize all AI technology
- To enable enterprises to use powerful AI while maintaining control over their data and operations
- To stop producing AI models