AivexaNewsSearch
AI news for builders and product teamsChecked every hour

With $720M round, Britain’s Fractile gets a leg up in the race for AI chips

Collected Nov 11, 2026

Fractile, a UK-based startup designing chips for AI workloads, has closed a $720 million Series C. Britain's Sovereign AI Fund participated in the round, and the government framed the investment around giving the UK a larger role in the hardware underpinning its future economy. The round was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint. According to the British government, the post-money valuation came in at "over £5 billion" — a figure the Financial Times and a Reuters source both put at possibly around $7 billion.

The raise places Fractile among the best-capitalized startups pursuing purpose-built AI silicon, a group that also includes Axelera, Etched and OLIX. These companies are chasing one of several routes toward lowering the cost of inference — running trained models in production, where demand keeps climbing as models grow more capable and more expensive to serve.

Fractile's pitch is not hardware alone. The company is building a full-stack offering that pairs its chips with a machine learning compiler intended to maximize how quickly large language models execute on its silicon. That positioning — chip plus software stack — is why the company describes itself as a frontier AI company rather than a component supplier, and the pricing of the round reflects that framing.

The design itself is the main differentiator. After a pivot, Fractile's approach centers on DRAM, the most common form of RAM, stacked directly on top of each chip in a way the company hopes will sidestep the usual trade-off between memory and capacity.

That trade-off is a familiar headache in accelerator design. Fast on-chip memory is scarce and expensive; large pools of external memory are capacious but slower to reach. In practice, models that need to hold long contexts, caches or many parameters while still generating tokens quickly run into this wall, and the gap has become more acute with agentic AI, where systems chain together long sequences of tasks using large models without giving up speed. Stacking DRAM on the die is one way to shorten the distance between compute and the data it needs — a direction the broader industry has explored with high-bandwidth memory, though Fractile's specific implementation and its claimed advantages have not been detailed publicly.

Hopes for dedicated AI chips are high even though such parts remain in limited supply to date. Incumbent GPUs still dominate training and much of inference, and a new entrant faces the long, capital-intensive path of taping out silicon, validating it in software, and persuading customers to port workloads onto an unfamiliar stack. The compiler is meant to reduce exactly that friction.

CEO Walter Goodwin said in a LinkedIn post that "there is still a lot of work ahead" and that the new funding will let the startup "ship [its] first chips and systems next year." That timeline is the concrete near-term milestone to watch: first silicon plus systems, which in this market usually means boards or servers rather than standalone parts.

Why it matters: this is one of the largest rounds yet for a European AI chip challenger, and the sovereign fund's participation signals that governments now see inference hardware as strategic infrastructure, not just a private-market bet. For developers, a credible full-stack alternative to GPUs would mean another target for model deployment and another compiler to learn; for product teams, cheaper inference is the lever that decides which agentic features are economically viable. The near-term test is whether Fractile's first chips and systems actually arrive next year and deliver on the memory-capacity promise — an inference on my part, since no performance benchmarks have been published.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Fractile has now become one of the top-funded players in the race for purpose-built AI chips.