The Reno-based startup already has a billion-dollar valuation and claims 3.5x better performance-per-watt than Nvidia's chips, and now it wants significantly more capital to press that advantage.
Positron AI is reportedly negotiating to raise roughly $750 million across two funding rounds as it positions itself to compete with Nvidia’s grip on AI inference hardware. The talks, first reported by Bloomberg, signal that the two-year-old startup is moving fast to scale its energy-efficient chip business.
Founded in 2023 and headquartered in Reno, Nevada, Positron has already raised approximately $309 million in total funding. The most significant chunk came from a $230 million Series B round completed in February 2026, which pushed the company’s valuation past the $1 billion mark.
That round attracted some heavy hitters. QIA, the Qatar Investment Authority’s sovereign wealth fund, participated alongside Arm Holdings. Positron’s current product, the Atlas inference system, is already shipping to customers. Cloudflare is among its buyers.
CEO Mitesh Agrawal has been leading the company since before its seed round, with a stated emphasis on US-made hardware.
Positron’s core pitch is energy efficiency. The company claims its hardware delivers more than 3.5 times the performance-per-watt compared to Nvidia’s latest offerings in certain AI inference workloads.
Positron’s next-generation chip, called Asimov, is targeting tape-out by late 2026 with production slated for early 2027.
GPU pricing and availability directly affect proof-of-work mining economics and the growing number of crypto projects that rely on GPU compute for AI-related services. Render Network, Akash, and similar decentralized compute protocols all depend on GPU supply and demand curves.
Bitcoin miners have increasingly pivoted toward hosting AI workloads as a revenue diversification strategy. Companies like Core Scientific and Hut 8 have signed major AI hosting deals.
The fact that Positron has no crypto tokens, no blockchain integration, and no apparent interest in decentralized anything is itself informative. It suggests that the serious money in AI infrastructure sees hardware innovation as a standalone thesis, not one that needs tokenization to attract capital.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our