Positron AI says it has secured up to 875 million dollars in a two-part Series C that values the company at 5 billion dollars post-money. The funding is intended to complete the Asimov inference chip's tapeout, build an engineering data center of more than 2 MW and ramp production of Titan systems. Its design emphasizes memory capacity and energy cost by using LPDDR5X rather than high-bandwidth memory.
Positron AI's financing consists of a 375 million dollar Series C and a C-1 tranche of up to 500 million dollars, producing a 5 billion dollar post-money valuation. NEA, Atreides Management, Valor Equity Partners, Andra Capital, SemiAnalysis Capital and Jim Clark are among the round's leaders, while Liberty Global Tech Ventures participated as an investor.[1], [2], [3]
Part of the funding is intended to take Positron's new Asimov inference chip through tapeout at the end of 2026, with production targeted for the second half of 2027. The Titan system is designed to combine four to eight of those chips. The plan also covers an engineering data center and emulation environment of more than 2 MW, along with manufacturing capacity for Titan. These dates and technical targets are the company's forward-looking plans rather than evidence of a finished product's performance.[1], [3]
Positron argues that moving model data during inference is constrained by memory capacity, bandwidth and power use. The Asimov design relies on widely available LPDDR5X rather than scarce and costly high-bandwidth memory, which the company says can ease supply and energy costs. More than 50 racks of its first-generation Atlas systems are being deployed at Oracle Cloud Infrastructure. Asimov has not yet taped out, however, so the new architecture's performance, efficiency and scaling claims remain to be tested on future hardware.[1], [3]
