AI inference know-how startup Positron has raised USD 23.5 million in funding from a gaggle of traders, together with Flume Ventures, Valor Fairness Companions, Atreides Administration, and Resilience Reserve. The corporate goals to scale manufacturing of its US-manufactured AI chips, providing a cheap and energy-efficient various to Nvidia’s AI {hardware}.
Additionally Learn: Fluidstack to Construct AI Supercomputer in France with EUR 10 Billion Funding
Atlas Programs: Excessive-Efficiency AI Compute
The corporate mentioned it’s already transport merchandise to knowledge facilities and neoclouds throughout the US, calling it a feat “that’s nearly insurmountable to chip startups with tons of of thousands and thousands or extra in backing.”
Positron claims its Atlas techniques delivers 3.5x higher efficiency per greenback and three.5x larger energy effectivity than Nvidia’s H100 GPUs for AI inference. Its memory-optimised structure achieves over 93 % bandwidth utilisation, considerably outperforming GPUs. These techniques ship 70 % quicker inference at 66 % decrease energy consumption than H100/H200 setups, decreasing knowledge middle capital expenditures by 50 %.
“Positron’s FPGA-powered servers help trillion-parameter fashions whereas providing plug-and-play compatibility with Hugging Face and OpenAI APIs,” the official launch mentioned.
Made in America AI Chip
Past the specs, the corporate highlighted that its first-generation Atlas techniques are totally designed, manufactured, and assembled within the US.
“With this funding, we’re scaling at a tempo that AI {hardware} has by no means seen earlier than–from increasing shipments of our first-generation merchandise to bringing our second-generation accelerators to market in 2026,” mentioned Mitesh Agrawal, CEO of Positron. “Our resolution is rising quickly as a result of it outperforms typical GPUs in each price and vitality effectivity, whereas delivering AI {hardware} that eliminates reliance on international provide chains.”
Additionally Learn: Authorities Plans to Carry Again Prime Indian-Origin AI Consultants for Native Innovation: Report
Home Various to World AI {Hardware}
In contrast to many semiconductor corporations that depend on offshore manufacturing, Positron mentioned it has constructed a totally American provide chain, guaranteeing that its AI {hardware} is designed, fabricated, and assembled inside america.
“Investing in home AI {hardware} is a strategic crucial with regards to securing America’s world AI posture,” mentioned Scott McNealy, Working Companion at Flume Ventures. “Positron is proving that world-class AI compute does not have to come back from abroad, and we’re excited to help their mission to make the US a pacesetter in AI {hardware} manufacturing.”
Past price financial savings, Positron mentioned its know-how is addressing the rising energy constraints of AI infrastructure. “Many legacy knowledge facilities battle to help high-power GPUs, which devour as a lot as 10,000 watts per server—far past the capability of conventional knowledge facilities. Positron’s energy-efficient structure permits these services to take part in AI computing with out requiring large infrastructure upgrades,” the AI chip startup mentioned.
Additionally Learn: LG and Tenstorrent Strengthen Partnership to Improve AI Chip Capabilities
“The demand for AI compute is skyrocketing, and enterprises are looking for viable alternate options which might be vitality and cost-efficient for the long run,” mentioned Rob Reid, Co-founder of Resilience Reserve. “What units Positron aside is not only its price effectivity, however its capacity to deliver AI {hardware} to market at an unprecedented velocity and supply a excessive efficiency per watt. Their progressive strategy is enabling companies to scale AI workloads with out the everyday limitations of price and energy consumption.”
