UK chip startup Fractile raises $220M for AI inference hardware
Fractile, a UK-based semiconductor company, closed a $220 million Series B funding round led by Accel, Factorial Funds, and Peter Thiel's Founders Fund. The capital will accelerate development and commercialization of the company's AI inference chips, expected to reach market in 2027.
Conviction, Gigascale, O1A, Felicis, Buckley Ventures, and 8VC also participated in the round. Previous investors include former Arm and Acorn Computers executive Stan Boland, and former Intel CEO Pat Gelsinger, who invested in January 2025.
In-memory compute approach
Fractile, founded in 2022 by Dr. Walter Goodwin, develops chips using in-memory compute-a method that runs calculations directly within processor memory rather than shuttling data between separate memory and processing units. This design reduces power consumption and improves performance for running generative AI and LLM inference at scale.
The approach targets the cost-versus-latency tradeoff that constrains current inference hardware. Goodwin said in a blog post that the company works "across the full stack, from foundational AI research to foundry process innovation to chip micro-architecture."
Expansion and early customer interest
In February 2026, Fractile announced £100 million ($135 million) in investment to expand UK operations over three years. The company will grow existing facilities in London and Bristol and build a new hardware engineering center in Bristol.
According to The Information, Anthropic has discussed purchasing Fractile's inference chips once they become available in 2027. The startup is hiring across London, Bristol, San Francisco, and Taipei.
For professionals working in AI for IT & Development, understanding chip architecture decisions like in-memory compute matters for infrastructure planning and cost modeling around large-scale model deployment.
Your membership also unlocks:
