Positron AI raises up to $875 million at a reported $5 billion valuation

Seven months is a short time to reprice a chip company whose next processor has not taped out. Positron AI, the Reno inference-hardware startup, said September 10 it had closed a $375 million Series C at a $3.5 billion pre-money valuation and lined up a Series C-1 of up to $500 million — a combined financing the company and Reuters both put at a $5 billion post-money mark, more than four times the $1.06 billion PitchBook figure from February.

Two tranches behind the headline number

Positron’s own release leads with an $875 million Series C. The same document, and Reuters the same day, split that figure. The first tranche is a $375 million Series C at a $3.5 billion pre-money valuation, co-led by NEA, Andra Capital, Atreides Management, Valor Equity Partners, and Dylan Patel’s SemiAnalysis Capital. The second is a Series C-1 of up to $500 million led by NEA and Jim Clark, the Silicon Graphics founder and Netscape co-founder.

That “up to” on the C-1 is the material qualifier. The company is presenting a combined $875 million financing; it has not published a close notice showing the second tranche fully drawn. Qatar Investment Authority, which also participated in Positron’s February Series B, confirmed its participation in the new round and the $5 billion valuation in its own September 10 statement.

Additional investors named by Positron include DFJ Growth, Resilience Reserve, Arena Private Wealth, Natural Capital, Helena, 1517 Fund, Flume Ventures, Unless, Boardman Bay Capital Management, Fincadia Advisors, Banyan Ventures, and U First Capital, plus strategic checks from VentureTech Alliance, Hudson River Trading, Cisco Investments, and Naver Ventures. Forest Baskett of NEA, Gavin Baker of Atreides, Thomas Jermoluk of Jim Clark Office, and Patel are joining the board.

The chip that valuation is buying

The money is aimed at Asimov, Positron’s next-generation inference processor, and Titan, the system that packages it. The company says Asimov will tape out on TSMC’s N3P process at the end of 2026, with production targeted for the second half of 2027. Until then, the $5 billion mark is a price on architecture and a first-generation deployment — not on a shipping custom chip.

Positron describes a memory-first design that uses commodity LPDDR5X instead of high-bandwidth memory and CoWoS packaging. Asimov is specified at 288 GB to 2,304 GB of memory per chip. Titan combines four to eight Asimov chips and is designed, the company said, to serve models beyond 16 trillion parameters and context windows beyond 10 million tokens in a single node. Those capacity and performance figures, including a claim that next-generation systems realize more than 90 percent of available memory bandwidth and lead on tokens per dollar and tokens per watt, are company statements. They have not been independently benchmarked in the materials reviewed for this article.

Chief executive Mitesh Agrawal said the raise funds tape-out, Titan production, and manufacturing scale. “Deploying Atlas at scale taught us an enormous amount about what inference customers actually need, and we have carried those lessons directly into Asimov and Titan,” he said in the company release.

Atlas is already in Oracle’s cloud

Positron’s shipping product is Atlas, a first-generation inference system. The company said it is deploying more than 50 Atlas racks at Oracle Cloud Infrastructure, with Parasail using that capacity for its own inference service. Jump Trading and i3d.net are named as additional production customers. SiliconANGLE and Reuters repeated the Oracle rack figure from the company; neither published an independent count.

That deployment is the operational proof Positron is selling alongside the next-chip story. Patel, investing and joining the board, said the team “has already banked real deployment experience with its first generation inside Oracle.” Atlas and Asimov are not the same product: the LPDDR5X memory thesis is attached to the next generation, not described as the architecture of the racks already going into Oracle.

What the new money is supposed to finish

Beyond silicon design, Positron said the financing will fund a 2 megawatt-plus engineering data center and emulation platform, LPDDR5X supply commitments, production capacity, system integration, and go-to-market expansion. QIA framed its check as support for Asimov tape-out and the Titan production ramp.

The February Series B — $230 million at a valuation PitchBook put at $1.06 billion, per Reuters — is the comparison the new mark is measured against. The company’s February release called that earlier post-money figure “exceeding $1 billion.” Seven months later, the same inference-hardware story is being financed at several times that price, with the custom silicon still scheduled to tape out at year-end and to enter production in the second half of 2027.