British AI-chip startup Fractile has raised a $720 million Series C, giving the company a major boost as countries and investors race to reduce the cost of running increasingly capable AI systems. The round was backed by Britain’s Sovereign AI Fund and led by Thrive Capital, Founders Fund, Lightspeed Venture Partners, and Redpoint.
Fractile is reportedly valued at more than £5 billion, or roughly $7 billion, according to reports cited by TechCrunch. That valuation reflects how strategically important AI hardware has become. Training large models gets attention, but the longer-term business challenge may be inference: the cost of running those models repeatedly in real products, agents, search systems, and enterprise workflows.
Fractile is building specialized AI chips and a machine-learning compiler designed to maximize how efficiently large language models run on its hardware. Its technical approach centers on DRAM memory stacked directly onto the chip. The company hopes this will reduce a major bottleneck in AI computing: the tradeoff between memory capacity and speed.
That tradeoff becomes more serious as AI agents take on longer, multi-step tasks. Agents need to hold more context, make repeated tool calls, and process large amounts of information without slowing down. If chips can move and access data more efficiently, the cost of operating advanced AI applications could fall significantly.
The funding is also a national-strategy story. Governments increasingly see AI infrastructure as a strategic capability, not just a private-sector market. The United Kingdom’s involvement signals an effort to keep a stake in the hardware layer of the AI economy, rather than relying entirely on foreign chipmakers and cloud platforms.
Fractile still has a long way to go. Designing a promising chip is very different from manufacturing at scale, competing with dominant incumbents, building software support, and persuading customers to change their infrastructure. The company says the new funding should help it ship its first chips and systems next year.
Fractile raises a useful question: will the next AI winners be the companies building better models, or the ones making those models dramatically cheaper to run?
