British AI chip startup Fractile's valuation soars to $6.5bn after securing Anthropic deal

Fractile Secures Massive Valuation Jump

Fractile, a British AI chip startup, is in advanced talks to raise funds at a $6.5bn valuation, marking a dramatic surge from the $1bn valuation it achieved just three months ago in May1

1

. The AI hardware start-up expects to bring in approximately $600mn in this funding round, with Lightspeed Venture Partners and Redpoint Ventures set to co-lead, alongside possible participation from Thrive Capital and Founders Fund2

2

. This pre-money valuation represents more than a six-fold increase, driven primarily by an initial agreement to supply chips to Anthropic3

3

.

Anthropic Deal Drives Growth

The valuation leap stems from Fractile's initial deal to sell roughly $250mn of chips to Anthropic, with intentions to expand that contract in the future1

1

. However, the chips won't be ready for deployment until 2027, reflecting the typical timeline in the chip business where designs take years to reach production2

2

. This early commitment from Anthropic signals confidence in Fractile's approach to AI inference tasks, even though the supply line remains years away. The deal fits a wider pattern of AI labs spreading their chip bets beyond Nvidia to control costs and mitigate supply chain risks1

1

.

Targeting the AI Inference Market

Fractile, founded in 2022 by Oxford University roboticist Walter Goodwin, designs systems to cut the time AI takes to generate answers and solve complex problems1

1

. The startup focuses specifically on running existing AI models—work known as inference that occurs after training, each time a model responds to a request. The company aims to enable scalable, low-latency AI token processing for faster AI model responses by building hardware that offers low latency and high throughput2

2

. Its operations span transistor-level circuit design up to cloud inference servers2

2

. Beyond chatbots, Fractile has pointed to applications in drug discovery and materials science, tasks that lean on fast, repeated computation1

1

.

Competing in the AI Chip Market

Fractile joins several newcomers designing chips tuned for AI inference tasks, all attempting to claim parts of the AI chip market dominated by Nvidia1

1

. Competitors include Cerebras Systems and venture-backed startups like Etched, which recently raised $700mn at a $21bn valuation, and Groq, which closed a funding round at $3.5bn with Nvidia itself joining1

1

. Each startup bets that specialized inference chips can undercut Nvidia's general-purpose processors on cost and speed. The wager centers on chips built for one job beating flexible ones at that specific task, despite Nvidia still supplying most of the industry's AI hardware1

1

.

British Presence in American-Led Race

Fractile's British base stands out in a race led by American firms. The startup operates from two primary locations in London and Bristol, though it has also hired in Taiwan and San Francisco2

2

. It sits alongside a small cluster of UK chip startups pushing against Nvidia, including OLIX, which raised money at a $3.3bn valuation1

1

. Founders Fund, co-founded by Peter Thiel, is among Fractile's backers, demonstrating American venture money flowing into British AI hardware1

1

. The May funding round was led by Accel, Factorial Funds, and Founders Fund, with participation from Conviction, Gigascale, 01A, Felicis, Buckley Ventures, and 8VC2

2

.

What This Means for the Industry

The Fractile funding round underscores intensifying competition to break Nvidia's grip on AI computing. As chatbots and agents handle more queries, inference has become a growing share of computing costs that AI firms pay for, making it a prime target for cheaper, faster hardware1

1

. Anthropic's decision to back Fractile early, despite a 2027 delivery timeline, suggests major AI labs are actively hedging against processor scarcity and cost pressures. With global demand for AI chips surging and shortages threatening supply chains, startups offering specialized solutions are attracting substantial investor interest2

2

. Watch for how Fractile executes on its 2027 timeline and whether it can deliver on its promise of faster, more efficient inference as AI systems move from single answers to long chains of reasoning.