Positron raises $875M as AI inference heats up
Positron AI raised an $875M Series C at a $5B valuation to commercialise memory-first inference chips aimed at challenging Nvidia’s dominance.
AI infrastructure spending is starting to shift from training toward the much larger long-term market for running models at scale.
What happened
Positron AI raised an $875M Series C at a $5B post-money valuation. The financing includes a $375M Series C and up to $500M Series C-1, with investors including NEA, Atreides, Valor, Andra Capital, SemiAnalysis Capital and Jim Clark. The capital will support tapeout of its next-generation Asimov inference chip and production of its Titan systems.
Why it matters
Inference workloads are heavily constrained by memory movement, power consumption and cost. Positron is betting that a memory-first architecture can improve performance without depending on scarce high-bandwidth memory, giving customers another path beyond Nvidia’s ecosystem.
The bigger picture
The AI chip market is fragmenting by workload. Training remains dominated by the largest accelerators, but inference is creating room for specialised architectures that optimise cost, memory and throughput for agents and production applications.