Fractile secures $720M for full-stack AI chip development

British AI-chip startup Fractile has raised a $720 million Series C to develop and commercialise purpose-built hardware for artificial intelligence inference. The round was backed by Britain’s Sovereign AI Fund and led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint.
The British government said the financing valued Fractile at more than £5 billion. Financial Times reporting and a Reuters source put the valuation at about $7 billion. Chief executive Walter Goodwin said the new capital is intended to help the company ship its first chips and systems next year.
A full-stack bid to reduce inference costs
Fractile is targeting inference, the production-stage work of running increasingly demanding AI models. Purpose-built chips are one route being explored to reduce the cost of serving those models, alongside efforts by other well-funded startups including Axelera, Etched and OLIX.
Its strategy extends beyond silicon. Fractile is building a machine-learning compiler designed to maximise the speed at which large language models execute on its own chips. The company therefore presents its offering as a full stack rather than a standalone processor.
This positioning matters because model performance in deployment depends on the interaction of hardware and software. Fractile’s compiler is meant to optimise execution for its architecture, while its chip design is intended to address memory constraints that increasingly affect large-model workloads.
DRAM design targets memory-capacity trade-off
After a pivot, Fractile centred its hardware approach on DRAM, the most common form of RAM. It plans to stack DRAM on top of each chip in a configuration intended to bypass the usual trade-off between memory and capacity.
The trade-off has become more difficult as companies pursue agentic AI systems that carry out longer chains of tasks with large models while maintaining speed. As demand for compute grows faster than available capacity, interest in alternative AI-chip architectures has intensified, even though such products remain limited in availability.
Funding creates a delivery test
The Series C gives Fractile substantial resources as it moves from architecture and software development toward shipping products. Its next stated milestone is the delivery of first chips and systems next year, when the claimed advantages of the DRAM-based design and compiler will face practical deployment requirements.
For businesses assessing AI infrastructure, the immediate implication is to distinguish funding and valuation from delivered capacity: specialised inference platforms may become relevant options, but procurement decisions should follow verified availability, model execution performance and system-level fit.

