Particle.news
Download on the App Store

Positron Raises $230 Million to Challenge Nvidia in AI Inference

Fresh capital backs a memory‑first design that swaps HBM for expandable LPDDR5x to cut inference power use.

Overview

  • Positron closed a $230 million Series B with backers including Qatar Investment Authority, with the round reported to value the company above $1 billion.
  • The raise was co-led by ARENA Private Wealth, Jump Trading, and Unless, with strategic participation from Arm and Helena and follow-on funding from Valor Equity Partners, Atreides Management, DFJ Growth, and 1517.
  • Positron says its Atlas inference systems are already shipping and fabricated and manufactured in the United States, and it claims Atlas can match Nvidia H100 performance at under one-third the power.
  • The next-generation Asimov accelerators target tape-out in late 2026 and production in early 2027, using LPDDR5x memory expandable via CXL from roughly 864 GB to about 2.3 TB per chip and expected to begin shipping next year.
  • Company claims for Asimov include five times more tokens per watt versus Nvidia’s Rubin and a 16 Tbps chip-to-chip mesh forming the Titan platform that can scale to 4,096 systems with more than 32 PB of pooled memory, which remains to be independently validated.