Fractile, a London-based AI chip startup founded in 2022, has raised $720 million in a Series C round that values the company at about $7 billion post-money, according to the Financial Times and a person involved in the deal cited by Reuters. The British government, an investor through its Sovereign AI Fund, would describe the valuation only as "over £5 billion." The round was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint; Georgian has also confirmed it participated.

The number is a startling jump. Fractile was valued at roughly $1.1 billion in a May funding round, meaning its valuation multiplied more than sixfold in about six months — a sign of how desperate the market has become for anything that promises to lower the cost of running AI models.

The technology is the reason. After a pivot, Fractile's approach now centers on DRAM, the most common form of memory, stacked directly on top of each chip. The company argues this sidesteps the long-standing trade-off between memory capacity and bandwidth — a trade-off that has grown more painful as agentic AI asks models to run longer chains of tasks without losing speed. Fractile pairs the silicon with a machine learning compiler, positioning itself as a full-stack, "frontier AI" company rather than a pure hardware vendor.

There is one large caveat: Fractile has no production chip yet. 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," with a first commercial chip targeted for 2027.

That has not stopped Fractile from joining the top tier of funded challengers to Nvidia's inference dominance, alongside Etched, Axelera and OLIX. The UK government framed its backing in strategic terms, saying the investment was meant to ensure "the U.K. has a greater stake in the hardware that will underpin its future economy."

For now, the bet is being priced on trajectory rather than shipped product: a sixfold valuation jump, a cast of tier-one venture firms, and a technical claim that stacked memory can beat Nvidia on the economics of AI inference.