Fractile: €187 Million to Revolutionize AI Chips

Fractile Raises $220 Million for AI Chips

Fractile, a British startup led by Walter Goodwin, has recently announced a funding round of $220 million, approximately €187 million. This financing is led by Accel, Factorial Funds, and Founders Fund, with participation from Conviction, Felicis, 8VC, Gigascale, O1A, and Buckley Ventures.

Founded in 2022, Fractile focuses on developing hardware architectures aimed at accelerating the inference of artificial intelligence models. Inference, which is the phase where models produce results, has become a major issue with the rise of reasoning models and autonomous agents.

Fractile's thesis is based on a simple yet powerful idea: the most advanced models will soon be limited not by their theoretical capabilities, but by the time required to execute their reasoning chains. "We bet that the most advanced AI systems would ultimately be limited in their impact by the time it takes to produce useful results," explains a representative of the startup. "The only way to truly unlock this latent value was to radically reinvent the hardware on which frontier models operate."

This evolution is gradually transforming the economy of the sector. Each request made to an AI model consumes computational resources. As models become more complex, the costs of inference increase. New reasoning systems now generate long processing sequences, sometimes spanning tens of millions of tokens.

Fractile estimates that some models already produce up to 100 million tokens to solve complex problems. At execution speeds close to 40 tokens per second on current architectures, such processing can require nearly a month of continuous computation.

For the company, this constraint goes far beyond mere performance issues. "Inference is both the revenue engine of the AI industry and the main limiting factor for its expansion."

Fractile draws a parallel with the systems developed by DeepMind for AlphaGo. The system did not rely solely on a neural network producing an immediate response, but on a succession of inferences allowing the exploration of different scenarios before each decision.

According to the British startup, large language models are now evolving in this direction. "Complex intellectual work involves many sequential steps, each dependent on the previous one," explains the company, which sees reasoning models as a first step towards systems capable of maintaining long and structured analysis chains.

The main technical bottleneck identified by Fractile concerns memory bandwidth. The company believes that current architectures are not progressing quickly enough to absorb the increasing demands related to long contexts and reasoning models. "To compress this month of computation into one day, we would need to achieve around 1,200 tokens per second while managing the complexity and capacity constraints of large models operating on very long contexts," the company specifies.

To address this issue, Fractile is working across the entire technology chain: microarchitecture, system design, manufacturing processes, and hardware optimization. This vertical approach brings the company closer to players like Cerebras Systems or Groq.

This battle over inference has become one of the main industrial fronts of AI. Several groups are seeking to reduce their dependence on traditional GPU architectures dominated by NVIDIA. AMD, Google, Amazon Web Services, and Intel are accelerating their investments in AI accelerators, while startups like SambaNova Systems, Etched, Tenstorrent, and d-Matrix are working to develop specialized architectures for reasoning and AI agent workloads.

Europe is also trying to maintain a presence in this strategic layer of infrastructure. In France, SiPearl is developing processors for European supercomputers, while Kalray is working on parallel processing architectures suited for massive data flows and AI applications. Scaleway and Mistral AI are also contributing to the emergence of a European computing and inference infrastructure. In the UK, Graphcore remains one of the main industrial precedents in this segment despite commercial difficulties against NVIDIA.

However, Fractile believes that the stakes go beyond the current uses of generative AI. "The workloads that are currently pushing the limits of the frontier are already transformational. Those that lie beyond this frontier will redefine the entire economy," asserts the company.

The firm is currently hiring in London, Bristol, San Francisco, and Taipei.