Positron, which is designing AI inference processor Asimov featuring a “memory-first architecture” and up to 2.3TB of memory, raised $875M at a $5B valuation
NEA, Atreides and Jim Clark are among investors putting $875 million into firm making chips designed for fast running of AI models
Context & Ripple Effects
Positron’s financing has accelerated from a $23.5M seed round in 2025 to a reported $230M Series B in February 2026. July reporting had already outlined a two-stage fundraising plan targeting a $5B second-tranche valuation.
The completed round supplies the capital base to bring Asimov’s memory-first inference design to market. It also puts Positron among a set of venture-backed specialists pursuing inference hardware, including d-Matrix’s inference-focused chip effort.
First-order effects
- Positron receives $875M to develop and commercialize Asimov, while NEA, Atreides and Jim Clark gain exposure to a $5B-valued inference-chip supplier.
- The round gives Positron a much larger financial cushion to turn its high-memory processor design into a product for model-serving workloads.
Second-order effects
- Inference-chip rivals such as d-Matrix face a sharper funding and execution benchmark as Positron moves from early-stage financing toward market deployment.
- Buyers seeking to run AI models gain a better-capitalized prospective alternative built around memory capacity, increasing pressure on suppliers to substantiate performance and operating-cost claims.
Third-order effects
- If heavily funded specialists can translate memory-centric designs into deployments, AI inference may support a more segmented chip market rather than concentrating demand in general-purpose accelerators.
- The funding path signals that access to large late-stage capital is becoming a prerequisite for independent AI-silicon companies to reach commercialization.
The trend: AI infrastructure finance is shifting toward large rounds for specialized inference hardware whose differentiator is memory architecture as much as raw compute.