According to an update from the EU, France’s Mistral AI has participated in an early pilot phase of a European Union plan to support homegrown AI startups by providing them with access to processing power for model training on the bloc’s supercomputers, which was announced in September and kicked off last month. However, one early takeaway is that the initiative should include specialized assistance for AI companies to teach them how to make the most of the EU’s high-performance computers.
“One of the things that we have seen is the need, not only to provide access but also to provide facilities—especially the skills, knowledge, and experience that we have in the hosting centers—on how this access can be not only facilitated but also to develop training algorithms that are using the best of the architecture and the computing power that is available right now in each supercomputing center and in our machines,” a spokesperson for the European Union said at a press conference a few days ago.
According to the idea, “centers of excellence” will be established to facilitate the development of specific AI algorithms that can operate on the EU’s supercomputers. 
AI companies are more likely to train their models on specialized compute gear supplied by US hyperscalers than on the processing capacity given by supercomputers. According to EU officials speaking in the background ahead of the official ribbon cutting for MareNostrum 5, a pre-exascale supercomputer that will be inaugurated Thursday at the Barcelona Supercomputing Center in Spain, the high-performance computing access for AI training program is being augmented with a support wrapper.
“We’re developing facilities for our SMEs to be able to understand how to best use the supercomputers, how to access the supercomputers, and how to parallelize their algorithms in the case of AI to be able to develop their models,” a spokesperson for the Commission said. “As of 2024, we expect much more of these kinds of approaches than we have right now.”
“AI is now considered a strategic priority for the Union,” they went on to say. “With AI becoming a strategic priority, next to the AI Act, we’re providing the innovation capacity—or we want to provide a large innovation window for our SMEs and startups to be able to best use our machines and this public infrastructure we’ve been creating so that they can compete internationally in developing safe, trustworthy, and ethical AI algorithms.”
Another EU official stated that an “AI support center” is on the way, with “a special track” for SMEs and startups to obtain assistance making the most of the EU’s supercomputing power. “What we need to recognize is that the AI community has not been using supercomputers for the last decade,” the researchers said. “They’re not new to GPUs, but they’re new to how to interact with a supercomputer, so we need to help them.”
“In many cases, the AI community emerges from a vast knowledge of how many GPUs can be crammed into a single box.” And they’ve done an excellent job at it. But what we have on the supercomputers are a lot of boxes with GPUs, and certain additional skill sets and assistance are required in order to scale up and utilize the supercomputer to its maximum capacity.”
Over the past five years, the bloc has significantly increased its investment in supercomputers, growing the hardware to a cluster of eight machines located throughout the region, which it also plans to interconnect via terabit networks to create a federated supercomputing resource that will be accessible in the cloud and available to users all over Europe.
The EU’s first exascale supercomputers are also set to be live in the coming years, with one in Germany (likely next year) and a second in France (anticipated in 2025). The Commission also plans to invest in quantum computing, with plans to acquire a slew of quantum simulators that will be co-located with supercomputers to provide a hybrid resource that combines both types of hardware so that the quantum computers can act as “accelerators” for the classical supercomputers, according to the Commission.
Applications being developed on top of the EU’s high-performance computing hardware include Destination Earth, a project to simulate Earth’s ecosystems in order to better model climate change and weather systems, and another to create a digital twin of the human body, which is hoped to help medical science by supporting drug development and even enabling personalized medicine. Using its supercomputing capabilities to fuel AI companies has emerged as a more recent strategic objective after the launch of the compute access for AI model training initiative by the EU president last autumn.
Last month, the EU also announced a “Large AI Grand Challenge,” a competition aimed at European AI startups “with experience in large-scale AI models” that aims to select up to four promising homegrown startups that will receive a total of 4 million hours of supercomputing access to support the development of foundational models. A €1 million prize pool is also set aside for the winners, who are expected to disclose their created models under an open source license for noncommercial usage or by publishing their study results, according to the Commission.
Through a request for proposals procedure, the EU already had a program in place to offer industrial customers access to core hours of supercomputing capability. However, the EU is focusing more on commercial AI with specific programs and resources, seeing an opportunity to turn its expanding supercomputing network into a strategic power source for scaling “Made in Europe” general-purpose AI. It seems to be no coincidence that France’s Mistral, an AI company attempting to compete with U.S. foundation model behemoths like OpenAI and claiming to provide “open assets” (if not entirely open source itself), is an early recipient of the Commission’s supercomputer access initiative. However, it may raise a few questions that a software startup that just received €385 million in Series A investment, notably from US investors like Andreessen Horowitz, General Catalyst, and Salesforce, is at the head of the line for an EU computing freebie. But, well, it’s yet another example of high-level strategic investments in “large AI.”
Because the EU’s “supercompute for AI” initiative is still in its early stages, it’s uncertain if there’s any model training upside to report from dedicated access just yet. (At the time of writing, Mistral has not replied to our request for comment.) However, the Commission’s hope is that by funneling support to AI startups so that they can tap into its investment in high-performance computing, combined with building out supercomputer hardware that it claims will increasingly be procured and configured with AI model training in mind, this will translate into a competitive advantage for a local AI ecosystem that is starting on the back foot versus hyperscaler-proximate U.S. AI giants.
“Since we do not have the large hyperscalers that Americans have, in the case of training these kinds of foundational models, we use our supercomputers, and we will develop a new generation of supercomputers that will be more and more AI-compliant,” a spokesperson for the Commission said. “Not only the ones that we have right now, but, as of 2024, the purpose would be that we will go in this direction—and have even more of our SMEs use the supercomputers for developing these foundational models.”
The strategy would include obtaining “more dedicated AI supercomputing machines that will be more based on accelerators than standard CPUs,” they said.
It remains to be seen whether the EU’s AI support strategy aligns with or diverges from the ambition of certain member states to foster national AI champions, which we heard a lot about during the recent fraught talks to set the bloc’s AI rulebook, in which France led a push for a regulatory carve-out for foundational models, which drew criticism from SMEs. However, the early inclusion of Mistral in the EU’s supercomputing access initiative may indicate a convergence of thought.
