Fractile seeks $6.5 billion valuation after Anthropic chip deal

Fractile, a startup building AI inference chips, is in advanced talks to raise about $600 million at a $6.5 billion pre-money valuation — six times its May round — after signing a supply deal with Anthropic.

The funding round, which includes some money invested at a lower valuation, comes three months after Fractile raised $220 million from Accel, Founders Fund, and Factorial Funds at a roughly $1 billion post-money valuation, according to people familiar with the matter.

Fractile has reached an initial agreement to sell about $250 million worth of chips to Anthropic, the AI lab behind the Claude model family, with plans to expand the contract. The startup designs processors optimized for running large language models at lower cost and power consumption than general-purpose GPUs from Nvidia.

The valuation surge from roughly $1 billion to $6.5 billion in three months reflects investor appetite for alternatives to Nvidia's dominant AI accelerators. Fractile's chips target the inference segment — the process of running trained models — where cost per query is the key competitive metric.

Fractile, founded by University of Oxford researchers, develops processors designed specifically for AI inference workloads. The company's approach targets the growing segment of AI computing where trained models are deployed to answer queries — a market that requires different performance characteristics than the training phase dominated by Nvidia's GPUs. Inference workloads are increasingly the bottleneck for AI companies as model usage scales, making specialized silicon an attractive investment.

The Anthropic agreement is notable because it represents a major AI lab committing to non-Nvidia hardware for inference workloads. Anthropic, which relies heavily on Nvidia GPUs for training its Claude models, has been diversifying its compute supply chain. The deal could pressure Nvidia's data center business as AI companies seek cost advantages in the inference segment. For Anthropic, locking in a dedicated chip supply at scale could reduce its dependence on the tight GPU allocation queues that have constrained AI labs since 2023.

Cerebras Systems, another challenger to Nvidia's dominance in AI compute, has also drawn substantial investor interest as AI labs seek alternatives to the incumbent's supply-constrained accelerators. The broader trend points to AI companies increasingly treating compute as a strategic asset to be secured through dedicated supply agreements rather than relying solely on the open market for GPUs.

For investors, the key question is whether Fractile can scale from initial agreements to volume production. The company has not disclosed its manufacturing partner or production timeline. Its chips would need to demonstrate meaningful cost-per-inference advantages over Nvidia's latest offerings to justify the valuation. The $250 million Anthropic commitment provides a revenue anchor, but scaling to hundreds of millions in annual sales will require additional design wins.

The funding round, expected to close in the coming months, would value Fractile at a level comparable to more established AI chip companies. The company's ability to expand the Anthropic contract and secure additional customers will determine whether the valuation holds. If Fractile delivers on its performance claims, the round could mark the beginning of a broader shift in how AI inference compute is sourced.

This article is for informational purposes only and does not constitute investment advice.