British AI chip startup Fractile secures $720 million to bring its first chips to market

Fractile, a British startup building processors for artificial intelligence, has raised $720 million in a Series C funding round, TechCrunch reports. The deal places the company among the best-funded young firms working on chips designed specifically for AI workloads.

Investors and valuation

The round was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint, with backing from Britain's Sovereign AI Fund. The UK government said its support is intended to give the country a larger role in the hardware that its future economy is expected to rely on. According to the government, the round values Fractile at more than £5 billion. The Financial Times and a source cited by Reuters have put the figure at about $7 billion.

A memory-focused design

Purpose-built AI chips are one of several approaches being pursued to lower the cost of inference, the stage at which trained AI models are put to work in real products. Fractile's hardware, reworked after an earlier change of direction, places DRAM, the most widely used kind of computer memory, stacked directly on top of each chip. The company hopes this arrangement will avoid the usual compromise between memory and capacity.

That compromise has become harder to manage as agentic AI spreads and businesses try to run long sequences of tasks on large models without losing speed. Fractile is also pairing its hardware with a machine learning compiler designed to make large language models run as quickly as possible on its chips, which is why it describes itself as a frontier AI company rather than only a chipmaker.

The company has not yet shipped products. Chief executive Walter Goodwin wrote on LinkedIn that the funding will allow Fractile to "ship [its] first chips and systems next year," while acknowledging that much work remains.

What it means for computing and internet users

For people following computing and online services, the round shows how much capital is flowing into the hardware layer behind AI tools. If specialised chips succeed in cutting inference costs, the AI features built into websites, software and cloud platforms could become faster and cheaper to run, although such chips have so far been available only in limited quantities.