Quick Facts
- Fractile raised $220 million in Series B funding co-led by Accel, Factorial Funds, and Founders Fund
- The startup claims its in-memory compute chips will increase performance from 40 to 1,200 tokens per second
- Commercial production is expected around 2027, with potential deployment by Anthropic
British AI chip startup Fractile secured $220 million in Series B funding to develop inference processors that could dramatically reduce the cost and time of running large language models. Accel, Factorial Funds, and Peter Thiel’s Founders Fund co-led the round.
The Oxford-founded company claims its chips will run AI models 100 times faster and 10 times cheaper than Nvidia’s GPUs. Based on computer simulations, Fractile projects increasing performance from roughly 40 tokens per second on existing hardware to around 1,200 tokens per second.
“Compressing a month of work into a day, a weekend of lab computation into a coffee break, will make all that work happen radically faster,” said Walter Goodwin, Fractile’s CEO and founder.
Fractile uses in-memory compute technology that stores data directly next to the transistors performing calculations. This approach eliminates the bottleneck created when data shuttles between memory chips and GPUs during AI model processing.
The company targets 20 times better performance per watt compared to existing AI hardware. However, Fractile has only tested its designs in computer simulations and has not yet manufactured test chips.
Commercial production remains at least two years away, with chips expected to reach market readiness around 2027. The Information reported that Anthropic held discussions with Fractile regarding chip purchases when the hardware becomes available.
If the partnership materializes, Fractile would become Anthropic’s fourth major chip supplier alongside Nvidia, Google, and Amazon. This represents a significant validation for the startup’s technology approach.
“Inference is dominant now, and the existing GPU architecture wasn’t built for it in ways that matter most at scale,” said Patrick Schneider-Sikorsky, director at the NATO Innovation Fund, which invested in Fractile.
The funding comes as AI chip startups globally raised $8.3 billion in 2026, with the sector expected to see record investment this year. Nvidia currently controls 80-95% of the AI accelerator market.
Fractile’s team includes former engineers from Graphcore, Nvidia, and Imagination Technologies. The startup follows a $132 million expansion announcement in February for its London and Bristol operations, including a new hardware engineering site in Bristol.
The UK government welcomed the investment as validation of British AI capabilities. AI minister Kanishka Narayan called it “a strong vote of confidence in British AI” that anchors “high value jobs and expertise” domestically.
The semiconductor sector carries risks, as evidenced by Graphcore’s 2024 acquisition by SoftBank for $600 million – less than its total venture funding. However, Fractile’s approach to inference processing represents a different technical strategy focused on the growing demand for faster AI model deployment.
Read more: British inference chip startup Fractile bags $220M to accelerate token consumption