Fractile, a London-based tech startup, has raised $220 million to move its innovative inference chip into production. The company designs chips that combine compute and memory on the same silicon, aiming to improve how artificial intelligence models are run. This funding comes at a time when the European tech scene is gaining momentum in the hardware space, with larger companies and investors paying close attention.

Innovative In-Memory Compute Technology

Fractile’s approach differs from traditional AI accelerators. Most current chips, like those from Nvidia, separate memory from processing units, which adds energy use and latency. Fractile’s chip performs the main calculations directly inside SRAM memory cells placed alongside the compute logic. This design reduces data movement and boosts efficiency.

The company claims its chips can run large AI models up to 25 times faster and at a tenth of the cost of current GPU setups. While these figures are based on simulations and small-scale tests, the technology shows promise for transforming inference workloads, which are the most common tasks once AI models are deployed.

Funding and Future Plans

The $220 million round was led by venture firm Accel, with notable involvement from Pat Gelsinger, former CEO of Intel, who joined as an angel investor and adviser. Existing backers like Kindred Capital, the NATO Innovation Fund, and Oxford Science Enterprises also participated. The funding will support the company’s next steps, including tape-out of its first commercial chip and early customer integration.

Fractile’s timeline targets a product launch around 2027. The company’s goal is to build a hardware stack that’s ready for customer testing and deployment. Early discussions suggest that AI company Anthropic is interested in buying Fractile chips once they’re available. If the partnership solidifies, Fractile could become a key supplier alongside giants like Nvidia and Google.

Overall, Fractile’s focus on in-memory compute could give it an edge in the niche market of inference chips. As more companies look for energy-efficient and cost-effective AI solutions, European startups like Fractile are positioning themselves as strong contenders. The recent funding boost highlights growing confidence in the region’s ability to innovate in hardware for AI applications.

Inspired by

• https://thenextweb.com/news/fractile-220m-inference-chip

Sources

• fractile.ai
• electronicsweekly.com