Positron AI, a Reno, Nevada-based startup building chips to run — not train — large AI models, said Wednesday it has raised $875 million in a Series C financing round that pushes its valuation to $5 billion, a fivefold jump from the $1 billion mark it hit just seven months ago.

The round was structured in two tranches: a $375 million Series C priced at a $3.5 billion pre-money valuation, followed by a Series C-1 of up to $500 million. It was co-led by NEA and Atreides Management, with Valor Equity Partners, Andra Capital, SemiAnalysis Capital and Netscape co-founder Jim Clark also anchoring the raise. The Qatar Investment Authority, Cisco Investments, Hudson River Trading and Naver Ventures were among the additional backers, according to the company's press release.

Betting Against Nvidia's Playbook

Positron's pitch is a direct challenge to Nvidia's dominance in AI inference: instead of the pricey, supply-constrained high-bandwidth memory (HBM) that powers Nvidia's GPUs, Positron's chips run on LPDDR5X — the same commodity memory found in smartphones. The company says that swap lets it pack far more memory per chip at a fraction of the cost.

Its next-generation chip, Asimov, is designed to carry between 288 GB and 2.3 TB of memory each and is slated to tape out on TSMC's 3-nanometer node by the end of 2026, with mass production targeted for the second half of 2027. Four to eight Asimov chips will combine into Positron's flagship system, Titan, which the company says can serve models exceeding 16 trillion parameters with context windows beyond 10 million tokens.

The company's current-generation Atlas system is already live, with more than 50 racks deployed at Oracle Cloud Infrastructure and additional customers including Jump Trading and i3d.net.

"Our focus now is to tape out Asimov, bring Titan to production, and scale manufacturing to meet the demand in front of us," said Positron CEO Mitesh Agrawal.

The new funding will go toward the Asimov tapeout, standing up a 2-megawatt-plus engineering data center and emulation platform, ramping Titan production, and locking in LPDDR5X supply commitments, the company said. New board members joining include NEA's Forest Baskett, Atreides' Gavin Baker, Thomas Jermoluk of Jim Clark's office, and independent chip analyst Dylan Patel of SemiAnalysis.

Riding — and Feeding — the Inference Boom

Positron's leap follows a February round that raised $230 million from investors including Arm and Qatar's sovereign wealth fund at a $1 billion valuation, as first reported by Bloomberg. Talks over the latest, far larger round were already underway by midsummer, according to Bloomberg's July report that Positron was in discussions to raise at a $5 billion valuation.

The whiplash valuation growth underscores investor appetite for alternatives to Nvidia as AI inference — running trained models to answer queries, rather than training them — becomes the industry's fastest-growing and most compute-hungry segment. With Asimov's tapeout still months away and mass production not expected until late 2027, Positron's next real test will be whether its memory-first architecture can deliver the promised cost advantage once it actually reaches customers at scale.