Fractile raises $720M Series C at $7B valuation for inference chips
Fractile raised a $720 million Series C at a $7 billion valuation to build full-stack AI chips with stacked DRAM, with first shipments due in 2027.
Fractile has raised a $720 million Series C at a $7 billion valuation, according to reporting on the round. The company is building full-stack AI chips using stacked DRAM. The architecture targets agentic inference workloads, meaning the running of AI models on live tasks rather than training them. First chips are expected to ship in 2027. Investors named in the reporting are Thrive Capital, Founders Fund, Lightspeed Venture Partners, Redpoint and the UK Sovereign AI Fund.
The round is part of a pattern in how the money is moving. Reporting on the round described it as one of the two largest AI funding rounds of the week to target inference delivery rather than training capability. The other was Torus, a $500 million growth round for modular inference infrastructure. The same reporting says Torus can deploy 1 to 20 MW of capacity in 12 to 16 weeks using existing grid interconnections, with 47,000 GPUs targeted by the end of 2027. Taken together, the reporting reads these deals as investors betting that deployment speed and access to power are the binding constraints on enterprise AI adoption, rather than the models themselves.
The available account does not say what performance or efficiency gains Fractile expects from stacked DRAM, what workloads it has been measured on, or how its chips compare with existing accelerators. With first chips not due until 2027, the design claims cannot yet be checked against shipping hardware.
What changed
Fractile raised a $720 million Series C at a $7 billion valuation to build full-stack AI chips using stacked DRAM.
Type: Business
Why it matters
Investment in inference-specific chips signals that hardware for deployment, not only training, is becoming a competitive market.
Original source
We have no link to the original source for this story yet.
Where we saw it
The AI Report
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works