Imagine it is autumn 2028 and the challenge has ended. What would a real breakthrough look like for you?

The initiative starts with ten teams and narrows down to three. A real breakthrough would be seeing those three teams on a credible path to raising about €1B each. Not by chance, but because we will work alongside them from day one to put them on a clear, fundable trajectory. By autumn 2028, the goal is that each team has commitments in place, not just momentum. 
If that happens, it means we have succeeded in bringing AI development back to Europe and secured a place at the table of modern AI development.

What kinds of technologies could enable Next Frontier AI to produce these breakthrough teams?

One promising area is turning the complexity of European federalism into a strength. That could mean solving hard problems with thousands of small models working together in agentic systems.
Another area is manufacturing. Imagine AI powered robots that can be tuned and deployed across different production lines and sectors.
AI for science is also a major opportunity. Speeding up innovation in drug discovery and material science could be transformative. These are just examples. I encourage teams to think big and dream big in any direction that excites them.

Having worked at OpenAI, what responsibility comes with building foundation models?

These systems are extremely capable, which means developers carry real responsibility. You need to think about how models impact the people using them. We already see how attached people become to chatbots.
As AI systems begin to make decisions with or for humans, they must be robust, reliable, and able to understand their environment. The more powerful the AI, the higher the responsibility. Ultimately, humans matter, and AI directly impacts human lives.

What challenges do you see for building frontier AI in Europe?

One challenge is Europe’s fragmented market. This makes scaling harder but also creates unique opportunities. Frontier AI is too large for any single nation to support alone, so access to the full European market is essential.
Another challenge is compute. We help teams figure out how to compute in European settings to maintain sovereignty and freedom of choice. This is difficult, but it is a challenge we are ready to take on.

If the best compute solution is an American hyperscaler, is that still allowed?

Yes, absolutely. If your requirements demand massive centralised compute and a US hyperscaler is the best option, you are welcome to use it. We don’t want compute to be a gatekeeping mechanism. At the same time, SPRIND is working on supporting European compute options, and we help teams however we can.

What gets you most excited about Next Frontier AI and why?

I’m excited by the people. We are already talking to many smart teams and receiving strong positive feedback from academia, industry, and politics. There is a shared feeling that this is the right initiative at the right time.
I’m also excited about what will be built. I’m already looking forward to the end of the challenge, even though it will be a bittersweet moment, because I know what comes out of it will be great.