Sovereign AI invests in Fractile to build the next generation of frontier AI chips

The chips powering AI have a huge influence on how quickly it can advance, how much it costs to run, and how widely it can be used. As AI models continue to become more capable, the demands on this hardware are growing rapidly.

Fractile is taking an ambitious approach to this problem. Headquartered in London and Bristol, the company is developing a new generation of advanced chips designed specifically for AI, with global ambitions and an expanding international operation.

Rethinking how AI chips are built

What makes Fractile particularly interesting is the breadth of its approach. Rather than adapting existing chip architectures to meet the demands of AI, the team is rethinking the technology from the ground up.

Its work spans the full stack: from the physical design of chips and the software that runs on them to fundamental research into machine learning. By developing these elements together, Fractile aims to remove inefficiencies that have become embedded in conventional computing architectures.

This matters because the performance of an AI system depends on much more than the raw processing power of its chips. How data moves through a system, how memory is accessed, how efficiently calculations are performed can all affect the speed and cost of running a model.

Fractile is looking at these problems together. Its approach has the potential to deliver substantial improvements in performance and efficiency, making it cheaper to run increasingly capable AI models.

The opportunity extends well beyond today's AI applications. More efficient hardware could make it economically viable to run larger models, serve more users and develop applications that would otherwise be constrained by the cost of computing.

Fractile is working on technology that could help shape the next generation of AI infrastructure.

Why we're investing

We see significant potential in Fractile's approach and the ambition of the team behind it. Designing a new chip architecture is an enormous undertaking. It requires deep expertise in semiconductor engineering, software and machine learning, as well as the ability to take complex research through to commercial products.

Fractile is bringing these disciplines together incredibly well. Its approach is technically ambitious, and we think it has the potential to make a meaningful contribution to the future of AI hardware.

Joséphine Kant, Investment Partner at Sovereign AI:

“Walter and the team at Fractile are taking on an ambitious challenge, with the potential to make a significant contribution to the global future of AI hardware. Their approach brings together deep expertise across chip design, software, and machine learning, and we believe it has the potential to compete with the very best globally.”