Positron AI Raises $230 Million to Accelerate Development of AI Inference Chips
February 6, 2026
In a significant move within the competitive AI hardware landscape, startup Positron AI has secured a substantial $230 million in a Series B funding round. The investment underscores the intensifying race to develop more efficient and powerful semiconductors for artificial intelligence, particularly for the energy-intensive inference phase where trained models generate outputs. The oversubscribed funding round, which valued the Reno, Nevada-based company at just over $1 billion, was co-led by Arena Private Wealth, Jump Trading, and Unless. It also saw participation from a mix of new and existing investors, including the Qatar Investment Authority, Arm, Valor Equity Partners, Atreides Management, and Flume Ventures. Founded in 2023, Positron AI focuses on designing energy-efficient hardware specifically for AI inference workloads. The company's first-generation chip, named Atlas and fabricated by Intel in the United States, is already shipping to customers. Positron claims the Atlas chip delivers three times the compute performance per watt compared to Nvidia's flagship H100 GPU. The capital infusion will primarily fuel the development of its next-generation accelerator, codenamed Asimov. Designed for memory-intensive AI applications, the Asimov chip is slated to support 2 terabytes of memory per accelerator and 8 terabytes per system. At a rack scale, this architecture enables a total memory capacity exceeding 100 terabytes, a figure the company states is comparable to Nvidia's upcoming Rubin GPU. Positron CEO Mitesh Agrawal framed the successful fundraise as a response to critical market bottlenecks. "We're grateful for this investor enthusiasm, which itself is a reflection of what the market is demanding," Agrawal said. "Energy availability has emerged as a key bottleneck for AI deployment. And our next-generation chip will deliver 5x more tokens per watt in our core workloads versus Nvidia’s upcoming Rubin GPU." He further highlighted memory as another major constraint, noting that the Asimov chip will ship with over 2,304 gigabytes of RAM per device, versus 384GB for Rubin, positioning it as a critical differentiator for video processing, financial trading, and large language model inference. The company is on an aggressive timeline, aiming to tape out the Asimov design in October 2026, just 16 months after beginning the design process, with production targeted for early 2027. This funding round and Positron's roadmap signal a growing challenge to Nvidia's dominance, as well-capitalized startups and tech giants alike seek to carve out niches in the high-stakes AI semiconductor market by addressing specific pain points like power efficiency and memory bandwidth.
Source: datacenterdynamics
