Fractile, a UK-based company founded by Walter Goodwin in 2022, has raised £12 million in seed funding to develop a chip that can deliver exponential performance improvements for AI models. The company claims to have developed a new, radically different approach to chip design for AI inference that can provide transformational improvements in performance for frontier AI models in deployment. The funding round was co-led by Kindred Capital, NATO Innovation Fund, and Oxford Science Enterprises, with participation from Cocoa and Inovia Capital, as well as several notable angel investors.
The problem with current chips and tools is that they are optimised for training large language models (LLMs) and are not suitable for inference, which is the process of running live data through a specific model to produce results. This means that the models are expensive to provision and run at scale, and performance is inhibited, with limited opportunity for AI model providers to drive differentiation.
Fractile’s approach is to fundamentally change the way computational operations are performed. The company has created completely different chips from new building blocks, using novel circuits to execute 99.99% of the operations needed to run model inference. A key aspect is a shift to in-memory compute, which removes the need to shuttle model parameters to and from processor chips, instead baking computational operations into memory directly. Fractile has already assembled a world-class team and filed patents protecting key circuits and its unique approach to in-memory compute.
The company is in discussions with potential partners and expects to sign partnerships ahead of production of its first commercial AI accelerator hardware. Fractile plans to use the funding to continue to grow its team and accelerate progress towards its first product. The AI inference chip market is projected to be worth $91 billion by 2030, making it a major growth opportunity.
“In today’s AI race, the limitations of existing hardware – nearly all of which is provided by a single company – represent the biggest barrier to better performance, reduced cost, and wider adoption,” said Dr Goodwin, CEO. “Fractile’s approach supercharges inference, delivering astonishing improvements in terms of speed and cost. This is more than just a speed-up – changing the performance point for inference allows us to explore completely new ways to use today’s leading AI models to solve the world’s most complex problems.”
Kindred Capital’s John Cassidy said: “AI is evolving so rapidly that building hardware for it is akin to shooting at a moving target in the dark. It’s a major technical challenge but, with the AI inference chip market projected to be worth $91 billion by 2030, there’s also huge growth potential. We’re excited to partner with Walter and the team on this journey.”
