Positron AI Raises $875 Million at $5 Billion Valuation to Fund Next-Generation Inference Silicon

Positron AI, the AI inference hardware company focused on making model serving dramatically cheaper and more energy efficient, has raised $875 million in a Series C financing at a $5 billion post-money valuation, the company announced Thursday. The round was co-led by NEA, Atreides Management, Valor Equity Partners, Andra Capital, Dylan Patel's SemiAnalysis Capital and Jim Clark, the founder of Silicon Graphics and Netscape. As part of the financing, Forest Baskett of NEA, Gavin Baker of Atreides Management, Thomas Jermoluk from Jim Clark Office and Dylan Patel will join Positron's board of directors.

The raise comes as AI's center of gravity shifts from training models to running them. Every agent, assistant and copilot depends on inference, and serving that demand at scale is increasingly constrained by memory capacity, memory bandwidth and power. Positron builds memory-first inference systems designed around those constraints. Its next-generation systems realize more than 90 percent of their available memory bandwidth, are built on commodity LPDDR5X memory that sidesteps constrained HBM and CoWoS supply chains, and deliver leading tokens per dollar and tokens per watt. Because the architecture is highly energy efficient, Positron's products can be deployed in air-cooled or liquid-cooled data centers operating at varying rack densities.

"Speed matters in this market, both in how quickly we ship new generations of silicon and in how quickly they reach customers," said Mitesh Agrawal, CEO of Positron AI. "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. Our focus now is to tape out Asimov, bring Titan to production, and scale manufacturing to meet the demand in front of us. This financing gives us the resources to do exactly that."

Positron is deploying more than 50 racks of Atlas, its first-generation inference system, at Oracle Cloud Infrastructure. Parasail, a key partner, utilizes that capacity to power its own inference service. Additional Atlas production customers include Jump Trading and i3d.net.

The financing was raised in two tranches. The Series C, $375 million at a $3.5 billion pre-money valuation, was co-led by NEA, Andra Capital, Atreides Management, Valor Equity Partners and Dylan Patel's SemiAnalysis Capital. The Series C-1, up to $500 million, was led by NEA and Jim Clark. Additional investors include DFJ Growth, Qatar Investment Authority, 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 alongside strategic investors VentureTech Alliance, Hudson River Trading, Cisco Investments and Naver Ventures.

The financing will fully fund the tapeout of Asimov, Positron's next-generation silicon; the bring-up of a 2 MW+ engineering data center and emulation platform; and the production ramp of Titan, Positron's next-generation inference system, including LPDDR5X supply commitments, production capacity, system integration and go-to-market expansion. Asimov tapes out on TSMC N3P at the end of 2026, with production in the second half of 2027, and pairs Positron's compute architecture with 288 GB to 2,304 GB of memory per chip. Titan combines four to eight Asimov chips into a single system, designed to serve models beyond 16 trillion parameters and context windows beyond 10 million tokens in a single node, scaling to thousands of nodes.

"We spend our lives measuring what AI hardware actually delivers in production, and most inference economics struggle under that scrutiny," said Dylan Patel, founder and CEO of SemiAnalysis Capital, who joins Positron's board of directors as part of the round. "Positron's architecture addresses the real constraint, memory, without depending on HBM or advanced packaging for its next-generation systems, and the team has already banked real deployment experience with its first generation inside Oracle. That is why I am investing and joining the board."

"Positron is solving the constraint that actually matters right now," said Forest Baskett, Partner at NEA. "While the rest of the industry is racing to secure scarce HBM and packaging capacity — even Nvidia's Rubin Ultra roadmap has had to scale back, from a terabyte of HBM4E down toward 192GB, simply because the supply isn't there — Positron built Asimov and Titan to sidestep that dependence altogether. That's not incremental, that's a fundamentally different bet, and they've already proven they can execute on it: Atlas is running at scale inside Oracle's cloud today. We led this round because we believe Positron is making the boldest memory-first bet in AI hardware. We couldn't be more excited to back Mitesh, Thomas and the team as they bring Asimov and Titan to market."

"The Positron inference architecture balances compute, the enormous required memory bandwidth and extraordinarily large context and weight storage," said Jim Clark, Silicon Graphics Founder and Netscape Co-Founder. "Its optimized power consumption, cost and density is near ideal for inference requirements of the next evolution of Frontier large language models with trillions of parameters."

"AI inference is emerging as one of the world's most consequential infrastructure markets and Positron's memory-first architecture addresses the performance, power and deployment constraints of modern AI systems," said Paul Tuan, Managing Partner of Andra Capital. "We are excited to partner with Mitesh, Thomas and the Positron team as they build the foundational infrastructure for the inference era."

Positron AI builds hardware and software to make AI inference dramatically cheaper and more energy efficient. The company's shipping product, Atlas, is deployed at hyperscaler scale today, and its next-generation custom silicon, Asimov, tapes out in late 2026 with production in the second half of 2027. Asimov powers Titan, a multi-terabyte-memory inference system for long-context and next-generation AI workloads. Positron's energy-efficient architecture allows its systems to be deployed in air-cooled or liquid-cooled data centers at varying rack densities, without dependence on constrained HBM or CoWoS supply chains.