Positron AI Raises $875 Million at a $5 Billion Valuation

Positron AI has raised $875 million in a round structured as a $375 million Series C at a $3.5 billion pre-money valuation plus a follow-on $500 million Series C-1, pushing the company's post-money valuation to $5 billion. The round drew co-lead investment from NEA, Atreides Management, Valor Equity Partners, Andra Capital, and SemiAnalysis Capital, with additional participation from DFJ Growth, Qatar Investment Authority, Hudson River Trading, Cisco Investments, and Naver Ventures.

Five Times Its Valuation From Seven Months Earlier

Positron's most recent prior round, a $230 million raise in February 2026, valued the company at $1 billion. Thursday's round values Positron five times higher than that February figure, a rapid re-rating even by the standards of this year's AI infrastructure funding market. Forest Baskett of NEA, Gavin Baker of Atreides Management, Thomas Jermoluk from Jim Clark's office, and chip analyst Dylan Patel all joined Positron's board as part of the financing, an unusually senior board addition for a single funding round.

Betting on Memory-First Inference Silicon

Positron builds inference chips designed around a memory-first architecture, a deliberate bet that the bottleneck in running large AI models in production is memory bandwidth rather than raw compute, a different technical wager than the compute-maximalist approach most large AI chip makers have pursued. The new capital will fund the company's next chip tapeout, codenamed Asimov, the build-out of a 2-megawatt engineering data center and emulation platform, and production of Titan, a next-generation inference system combining four to eight Asimov chips into a single node.

A Wager Against the Dominant Chip Architecture

Betting that memory bandwidth, not raw compute throughput, is the real constraint on inference workloads puts Positron in direct technical opposition to the design philosophy behind most of today's dominant AI accelerators, which have generally prioritized adding more compute cores per chip generation. If Positron's thesis proves correct as inference workloads scale toward ever-larger models serving live production traffic, memory-first chips could offer a meaningfully better cost-per-query economics than compute-maximalist designs, a difference large cloud providers and AI labs running inference at scale would notice quickly in their infrastructure bills.

What This Means for Founders

Positron's round shows that investors remain willing to underwrite extraordinarily fast valuation growth for AI infrastructure companies pursuing a genuinely differentiated technical architecture, provided the team can attract senior board-level validation alongside the capital itself. AngelLinx's AI and machine learning investor directory helps founders identify investors active in AI infrastructure and chip design, and the fit-scoring match tool connects founders with the right match as technical differentiation becomes the core of their pitch. The live listings page shows current founder campaigns performing against real investor interest today, and AngelLinx's guide to managing burn rate is essential reading for founders running capital-intensive hardware development cycles. The AngelLinx newsroom tracks major AI infrastructure funding rounds as they develop. Founders building differentiated AI infrastructure can register at https://angellinx.ai/register today.

AngelLinx Intelligence | angellinx.ai