Positron AI announced an $875 million Series C at a $5 billion post-money valuation — roughly five times the $1 billion-plus valuation after its $230 million Series B in February 2026 — to fund Asimov, a custom inference ASIC built around commodity LPDDR5X instead of scarce High Bandwidth Memory, TechTimes reported.
Asimov is planned on TSMC N3P with tapeout targeted for late 2026 and production in the second half of 2027. The design calls for 288GB to 2,304GB of LPDDR5X per chip (the high end via CXL expansion) and claims over 90% realized memory bandwidth on Transformer workloads, versus under 30% typical GPU utilization in inference decode, according to the company. Those Asimov numbers are design targets: the chip has not taped out, and earlier Atlas benchmarks still need third-party verification, TechTimes noted.
The raise came two days after Qualcomm and Amazon disclosed a multi-generational LPDDR-based inference partnership worth up to $60 billion in purchases. Positron’s first-generation Atlas systems are already deployed in more than 50 racks at Oracle Cloud Infrastructure. The round included NEA, Andra Capital, Atreides Management, Valor Equity Partners, Dylan Patel’s SemiAnalysis Capital, and Jim Clark. Gartner reported AI inference spending at $23.3 billion this year versus $19 billion for training — the first time inference has led.
Leave a Reply