Industry
AI Minute Newsroom
2026-08-20
In May this British chip startup was worth $1 billion. One Anthropic order later, it is asking for $6.5 billion.
Bloomberg reported on 19 August that Fractile, a London chip company, is in talks to raise about $600 million at a pre-money valuation of $6.5 billion. In May it raised $220 million at roughly $1 billion, which makes this a more than sixfold repricing in three months. The thing that changed in between is a customer. Fractile has agreed an initial deal to sell Anthropic around $250 million of its chips, and both sides have said they intend to expand it. The company was founded in 2022 by Walter Goodwin, an Oxford roboticist, and builds inference silicon — chips for running trained models rather than training them — using in-memory compute, an approach that attacks the memory bottleneck limiting how fast a model can emit tokens. Fractile has pitched applications beyond chatbots, including drug and materials discovery. The chips are not expected to be production-ready until 2027.
Why it mattersThat last sentence is the whole risk. A $6.5 billion price is being put on silicon that does not exist in volume, in a market where Nvidia holds the incumbency and Cerebras, Groq and Etched are making versions of the same pitch. What makes the Anthropic order load-bearing rather than decorative is who placed it: frontier labs are the hardest customers in this market to fool, because they can measure the claims against their own workloads before signing. But $250 million is a serious evaluation, not a bet-the-company commitment — it is the size of an order you place when you want the option, not the size you place when you have decided. The number worth watching is not the valuation. It is whether that contract expands in 2027, when the chips have to stop being a thesis and start running models.
✓ Verified · 4 sources
▶ Related video: How UK chip startup Fractile is hoping to challenge Nvidia
Read in the app — free, in 9 languages
Related stories
OpenAI fired three safety staff for talking to an outside watchdog.
2026-10-05OpenAI will put ads beside the pictures you ask ChatGPT to draw.
2026-10-05Musk is dropping the word artificial from his AI unit's name.
2026-10-05Altman says the world should accept some harm to keep AI open.
2026-10-05Change.org will spend $100 million on AI that failed on its site.
2026-10-05