As artificial intelligence systems grow more advanced, the industry is increasingly confronting a new challenge that goes beyond model capability: speed. While AI models continue improving rapidly, the time and cost required to generate useful outputs at scale are becoming major bottlenecks limiting further progress. UK based startup Fractile believes solving that problem will require an entirely new generation of hardware, and investors are backing that vision with one of Europe’s largest AI infrastructure funding rounds.
The company has raised $220 million in a Series B funding round led by Accel, Factorial Funds, and Founders Fund. Additional participation came from Conviction, Gigascale, O1A, Felicis, Buckley Ventures, and 8VC.
Reinventing AI Inference Hardware
Founded in 2022, Fractile is developing next generation inference hardware specifically designed for frontier AI systems. The company’s core thesis is that future AI progress will increasingly depend not just on model intelligence but on how quickly and economically those models can generate outputs at scale.
As large language models and advanced AI systems begin tackling more complex tasks, they often require extremely long output sequences involving millions or even tens of millions of tokens. Existing hardware architectures struggle to process these workloads efficiently, creating technical and economic limitations that slow down AI deployment and expansion.
Fractile is attempting to address that problem by redesigning the underlying hardware infrastructure powering AI inference. Its work spans AI research, chip microarchitecture, and semiconductor foundry process innovation, with the goal of dramatically increasing inference speed while reducing operational constraints.
The Growing Bottleneck in AI
According to Walter Goodwin, the company was founded on the belief that the biggest limitation facing future AI systems would not necessarily be intelligence itself but the amount of time required for those systems to produce valuable outputs.
Goodwin explained that Fractile’s founding thesis centred on the idea that unlocking the full value of advanced AI would require radically reinventing the hardware used to run frontier models. Since then, the company has focused entirely on building chips and systems capable of solving that challenge.
He argued that AI capabilities have already reached a stage where latency and inference speed are becoming the main barriers to progress. As models improve, they are increasingly being orchestrated across far longer and more sophisticated output sequences, significantly increasing computational demands.
The Economics of AI Inference
One of the major issues Fractile is addressing involves the economics of inference. According to Goodwin, inference has become both the revenue engine of the AI industry and one of the biggest constraints on scaling it further.
He noted that some large language models are already generating outputs containing up to 100 million tokens when solving highly complex tasks. On current hardware systems, which often operate at roughly 40 tokens per second, generating a single output of that size could take nearly a month to complete.
The company believes memory bandwidth limitations in current chip architectures are among the key factors restricting inference speed and overall AI scalability. Fractile’s hardware designs are specifically aimed at overcoming those limitations through entirely new system architectures.
Looking Beyond Current AI Workloads
Goodwin said the long term opportunity for Fractile extends beyond simply accelerating today’s AI workloads. Instead, the company believes faster inference hardware will enable entirely new categories of AI applications and workflows that are currently impractical because of time and cost constraints.
The newly raised funding will support continued development of Fractile’s hardware systems as the company expands its operations and hiring efforts across London, Bristol, San Francisco, and Taipei.
As the AI industry increasingly shifts focus from model creation toward inference efficiency and deployment scalability, Fractile is positioning itself as part of a new generation of companies building the infrastructure required for the future of frontier AI.
