Fractile Raises $720M for AI Chips, Hits $7B Valuation

British startup Fractile has raised $720 million in a Series C round, pushing the AI chipmaker's valuation past £5 billion, or roughly $7 billion, according to TechCrunch. The round positions the company among the most heavily funded startups building specialized hardware for running large language models.

What the $720 Million Series C Covers

The raise, reported by TechCrunch, funds Fractile's push to move from chip design to manufactured systems. Fractile CEO Walter Goodwin wrote in a LinkedIn post that the money will let the company "ship our first chips and systems next year," while cautioning that "there is still a lot of work ahead."

Who Is Backing Fractile's Round

The Series C was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners, and Redpoint, according to TechCrunch. Britain's Sovereign AI Fund also backed the round. The U.K. government said the investment is meant to ensure "the U.K. has a greater stake in the hardware that will underpin its future economy," a statement TechCrunch cited directly. The roughly $7 billion valuation figure was reported by both the Financial Times and a Reuters source, per TechCrunch's account.

How Fractile's Chip Design Works

Fractile's hardware approach centers on DRAM, the most common form of computer memory, stacked directly on top of each chip. TechCrunch reports the design is meant to get around the usual trade-off between memory capacity and processing speed that limits how fast AI models can run. The company pivoted to this architecture after earlier work, and it has paired the hardware with its own machine-learning compiler designed to speed up execution of large language models on its chips — a full-stack approach TechCrunch says sets Fractile apart from pure hardware plays.

Why Memory Is the Bottleneck for AI Inference

The memory-capacity trade-off has become a bigger problem as companies push AI systems toward longer, more autonomous task chains — so-called agentic AI — without sacrificing speed, according to TechCrunch. Inference, the process of running trained AI models in production rather than training them, has become one of the costliest parts of deploying large models at scale. Purpose-built chips are one of several approaches companies are testing to bring that cost down, TechCrunch notes, alongside Fractile peers Axelera, Etched, and OLIX, which TechCrunch lists among the top-funded competitors in the same race.

What Comes Next for Fractile

Goodwin's stated timeline puts Fractile's first chips and systems on track for next year, a milestone that would mark the company's transition from design work to physical product. TechCrunch frames the compiler and hardware combination as central to Fractile's pitch that it operates as what it calls a frontier AI company rather than a conventional chip vendor. Whether that full-stack bet pays off will depend on execution timelines that, by Goodwin's own account, remain unfinished. For now, the scale of the raise — and the direct involvement of a government-backed sovereign fund — signals how much weight is being placed on domestic chip capacity as AI compute demand keeps climbing.