Positron AI raises $875M for memory-first inference chips
The Nevada startup's valuation jumped to $5 billion as it prepares next-generation hardware designed to handle trillion-parameter models.
Positron AI secures massive funding round
Positron AI closed an $875 million financing round on Thursday that values the Reno-based chip startup at $5 billion post-money, reflecting surging investor appetite for specialized AI inference hardware.
The capital came in two pieces: a $375 million Series C at a $3.5 billion pre-money valuation, co-led by NEA, Atreides Management, Valor Equity Partners, Andra Capital, and SemiAnalysis Capital, followed by a Series C-1 of up to $500 million anchored by NEA and Netscape co-founder Jim Clark. The investor roster also includes DFJ Growth, Qatar Investment Authority, Hudson River Trading, Cisco Investments, and Naver Ventures.
The valuation represents a fivefold increase from Positron's February funding round, which brought in $230 million at a $1 billion valuation, according to the Wall Street Journal.
Why it matters
Positron's approach tackles a critical bottleneck in AI infrastructure: memory capacity and bandwidth for inference workloads. By designing around commodity LPDDR5X memory rather than competing for constrained high-bandwidth memory supplies, the company offers a potential path around supply chain constraints that have limited deployment of large language models. The rapid valuation climb signals investor confidence that memory-centric architectures may complement or challenge compute-focused designs from established players.
Memory-first architecture for massive models
Positron's strategy diverges from conventional AI chip design by prioritizing memory over raw computational power. The company's next-generation Asimov chip will pair its compute architecture with between 288 GB and 2,304 GB of memory per chip using commodity LPDDR5X memory, avoiding the supply constraints and advanced packaging requirements associated with high-bandwidth memory.
Asimov is scheduled to tape out on TSMC's N3P process at the end of 2026, with production targeted for the second half of 2027.
The funding will support the Asimov tapeout, construction of a 2-megawatt engineering data center and emulation platform, and the production ramp of Titan. Titan represents Positron's next-generation inference system, combining four to eight Asimov chips into a single node designed to serve models exceeding 16 trillion parameters and context windows beyond 10 million tokens.
Early deployments inform roadmap
Positron has already deployed its Atlas server rack at Oracle Cloud Infrastructure, with over 50 racks now operational. Customers using that capacity include Parasail, Jump Trading, and i3d.net.
CEO Mitesh Agrawal said the Oracle deployment directly informed the design decisions behind 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," Agrawal said.
The financing brings new board members from the investor group: Forest Baskett of NEA, Gavin Baker of Atreides Management, Thomas Jermoluk from Jim Clark Office, and analyst Dylan Patel of SemiAnalysis Capital.
Details of the funding round were first reported by Quartz.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call