Fractile Raises $720 Million for AI Chips Before Its First Silicon Ships
Fractile, a British startup developing AI chips, has raised a $720 million Series C funding round, and Britain's Sovereign AI Fund backed the round. TechCrunch reported the raise on 11 November 2026.
The round was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint. The British government stated that Fractile's valuation in the round was over £5 billion, while the Financial Times and Reuters reported a valuation of $7 billion.
CEO Walter Goodwin said the new funding will enable Fractile to ship its first chips and systems next year. The company was founded in the summer of 2022, according to its website. Shipping is the stated plan, not a completed milestone.
Fractile describes a full-stack approach that pairs its silicon with a machine learning compiler, the software that translates a model into instructions the chip executes. The compiler is designed to maximise the speed of large language model (LLM) execution on the chips. On the hardware side, the company's approach centres on DRAM (dynamic random-access memory) stacked on top of each chip. TechCrunch characterises that layout as a way to sidestep the tradeoff between memory and capacity, a constraint that becomes more acute as context windows grow. A context window is the amount of text a model can process in a single pass.
The company states that it is building chips and systems to generate output at thousands of tokens per second, and to operate large models at very long contexts. These are targets Fractile has set for its own products. They map onto the two bottlenecks that dominate production inference. The first is decode throughput, which is bound by how fast memory can feed the processor. The second is the key-value (KV) cache, the stored state a model keeps for the conversation so far, whose footprint grows with context length. Hitting both at once is the design problem the stacked-memory and compiler choices are meant to address.
The broader context here is the capital structure. A $720 million Series C, with a sovereign fund participating alongside growth-stage investors, is a large round for a company that has not yet shipped silicon. The valuation discrepancy between the government statement and the figures reported by the FT and Reuters is worth keeping in view, since gaps of this kind tend to narrow only when filings or further disclosures arrive.
In my view, the decisive test for Fractile will be the one Goodwin set out: whether the first chips and systems arrive next year and perform against the throughput and long-context claims the company has published. Funding closes the gap between design and tape-out, but it does not settle whether a stacked-DRAM architecture and a custom compiler deliver in production. For buyers and engineers evaluating alternatives to incumbent accelerators, the round adds a credible new option to the roadmap conversation. The next useful data point will be silicon in hand rather than another funding announcement.
