Positron AI raises $230 million in Series B funding, achieving unicorn status as it accelerates deployment of energy-efficient AI inference infrastructure for enterprise and cloud customers.
Positron AI, a company focused on energy-efficient AI inference hardware, has announced a $230 million Series B funding round that was oversubscribed, pushing its post-money valuation beyond $1 billion. The round was co-led by ARENA Private Wealth, Jump Trading, and Unless, with new and strategic participation from the Qatar Investment Authority, Arm, and Helena. Existing investors including Valor Equity Partners, Atreides Management, DFJ Growth, Resilience Reserve, Flume Ventures, and 1517 also joined the round, signaling strong confidence in Positron’s approach to lowering the cost and power demands of AI inference.
According to CEO Mitesh Agrawal, the funding reflects growing demand for infrastructure that can address power and memory bottlenecks in AI systems. He noted that Positron’s upcoming chips are designed to deliver up to five times more tokens per watt than NVIDIA’s next-generation Rubin GPU for certain workloads. The company’s next custom silicon platform, Asimov, is expected to ship next year with more than 2,304GB of RAM per device—far exceeding the 384GB projected for Rubin—making it particularly suited for video processing, trading systems, multi-trillion-parameter models, and applications requiring extremely large context windows. The company also expects improved cost efficiency for memory-intensive use cases.
Positron’s current product, Atlas, is already shipping. It is an inference system built for rapid deployment and scaling, with manufacturing based in the United States to support reliable supply and faster capacity ramp-ups. The company positions itself as building the infrastructure layer needed to run modern AI models more efficiently at scale.
Industry observers say memory capacity and bandwidth are emerging as major constraints for next-generation inference workloads. Dilan Patel, founder and CEO of semiconductor research firm Semianalysis, said Positron’s memory-centric approach could deliver more than ten times the high-speed memory capacity per chip compared with existing and upcoming semiconductor vendors.
A notable part of the funding round is Jump Trading’s decision to co-lead after first adopting Atlas as a customer. The firm reported that in its testing, Atlas delivered roughly one-third the end-to-end latency for targeted inference workloads compared with comparable H100-based systems. It also highlighted the system’s air-cooled, production-ready design and supply-chain advantages. After evaluating Positron’s roadmap for its upcoming Asimov chips and the Titan system, Jump Trading chose to invest, citing the company’s potential to reshape the cost and performance curve for AI inference.
The next-generation Asimov platform is designed around a memory-first architecture. Each accelerator is expected to support up to 2 terabytes of memory, with the Titan system reaching 8 terabytes and rack-scale deployments exceeding 100 terabytes. The design targets memory bandwidth comparable to NVIDIA’s Rubin GPUs while focusing on efficiency and system-level performance.
Arm, one of the strategic investors, said the partnership reflects how tightly integrated systems and ecosystems are becoming in next-generation AI infrastructure. Positron is also working with supply-chain and technology partners such as Supermicro to bring its platforms to market.
The company plans to tape out the Asimov chip roughly 16 months after starting its design process following the Series A round, and it aims to maintain a similar development pace for future chips. Agrawal said speed of development is a key competitive factor, especially when competing with established players like NVIDIA.
Looking ahead, Positron expects strong revenue growth in 2026 and believes it could become one of the fastest-growing silicon companies, driven by commercial traction roughly two and a half years after its founding. The company is already working with multiple customers across cloud computing, advanced computing, and other performance-intensive industries as it expands deployments and customer programs.
About Positron AI
Positron AI develops purpose-built hardware and software to dramatically lower the cost and improve the energy efficiency of AI inference. Positron’s currently shipping product, Atlas, is designed for rapid and scalable deployments. The company’s next-generation custom silicon, Asimov, is targeted for tapeout in late 2026 with production start in early 2027. Positron’s systems are designed to deliver superior economics for long-form contextual and next-generation AI workloads. Learn more at positron.ai .
The official version of this press release is the original language version. Translated versions are provided for the convenience of readers and have no legal effect. When using translated versions as reference material, please refer to the original language version, which is the only legally effective version.
