On November 11, British AI chip startup Fractile announced a $720 million Series C funding round at a valuation of more than £5 billion, roughly $7 billion.The roughly $7 billion figure was reported by both the Financial Times and Reuters. Britain's Sovereign AI Fund took part in the round, which was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint. With the money in hand, Fractile is also positioning itself as a frontier AI company, not just a chipmaker.
Fractile is betting on the cost of inference. Its design stacks DRAM — the most common type of memory — directly on top of each chip, sidestepping the long-standing trade-off between memory and capacity: agents running long chains of tasks need large models that are both fast and able to remember context, and conventional architectures struggle to deliver both. Beyond hardware, Fractile is building its own machine-learning compiler so large language models run faster on its chips. This approach is the result of a pivot: the team changed technical direction once before, and is now betting everything on the combination of stacked DRAM and its own compiler.CEO Walter Goodwin said the funding will enable the startup to ship its first chips and systems next year.
Why this round is worth watching
Purpose-built inference chips are becoming a main route to cheaper AI. Google has just launched a new generation of TPUs, Nvidia's inference chips are competing for the same market, and Fractile now stands alongside Axelera, Etched and OLIX as one of the best-funded startups in the space. The British government's involvement is also notable: the official line is that the UK should hold a bigger stake in the hardware that will underpin its future economy, giving the round an industrial-policy tint as well.At the same time, AI chips remain in limited supply, and the fact that compute cannot keep up with demand is now industry consensus — another reason purpose-built chips carry such high hopes.
Fractile, however, has not shipped a single chip yet, and its valuation is already at $7 billion. Whether its technical approach delivers will come down to the measured speed and per-unit inference cost of the first products next year. For anyone tracking AI compute, what to wait for next is not the next funding announcement, but real feedback from the first customers.
