Scaling intelligence: POSITRON’s vision for the AI revolution

Powering the AI Revolution: Positron AI’s Breakthrough in Energy-Efficient Inference Chips

Mitesh Agrawal, CEO of Positron AI, introduced his company’s mission to design silicon chips for AI inference applications, such as ChatGPT and image generation. At Web Summit Lisbon 2025, he discussed the rapid growth of AI, its increasing role in engineering, and Positron AI’s solutions to address critical bottlenecks in the evolving AI landscape.

Mr. Agrawal highlighted AI’s unique ability to solve the labor bottleneck through digital means, initially for tasks like code generation. This transformative potential sparks discussions about job markets. He emphasized that inference is rapidly becoming a fundamental bottleneck, with AI already serving a billion users and projected to grow fivefold globally.

In engineering, AI models can generate substantial code. The cost of AI-generated output, equivalent to an average programmer’s work, is estimated at $14,000 USD annually, often lower than human wages. This dramatic cost reduction is expected to spur a radical increase in code and operational development, integrating AI into nearly every aspect of life, from mobile apps to banking.

The increasing scale of AI models demands immense data center power and compute resources. While Nvidia currently dominates, Mr. Agrawal foresees a shift to a multi-player market, as different AI models have varied requirements for scale, speed, or cost-efficiency. He distinguished between AI training (compute-heavy) and AI inference (memory-bound).

Positron AI directly addresses the memory-bound nature of inference by designing highly energy-efficient silicon chips with substantial memory capacity. Founded in April 2023, the company has rapidly achieved over $60 million in revenue and secured $300 million in funding, including a recent Series B round with significant participation from the Qatari region via QIA.

Despite a lean team of 50 employees, Positron AI successfully develops complex silicon products and competes effectively. Their competitive advantage stems from offering six times more memory capacity than rivals for the same cost, coupled with superior power efficiency. This translates to faster output speeds and greater efficiency per unit of energy, tackling the AI industry’s energy bottleneck.

The company serves major hyperscalers and leading AI developers. Positron AI has deployed its first product and is expanding globally, with plans for deployments in the Middle East and worldwide. Their next-generation systems are slated for early 2027, showcasing a rapid development cadence and a long-term vision to become a leading force in the AI hardware sector.