Olix Computing Raises $220M for Photonic AI Chips
Olix Computing, a startup founded by a 25-year-old entrepreneur, has secured $220 million to develop photonic AI inference chips. The company aims to directly challenge Nvidia's market dominance by leveraging photonics for ultra-low-latency, high-bandwidth processing with reduced power consumption.
- The company's founder, James Dacombe, dropped out of school at 16 to work as a software engineer and previously founded the brain healthcare company CoMind before receiving a $200,000 grant from the Thiel Foundation in 2022. - Olix's chip is called the Optical Tensor Processing Unit (OTPU) and it is designed to circumvent the "memory wall" bottleneck by completely avoiding High-Bandwidth Memory (HBM). Instead, the architecture relies on faster SRAM integrated directly onto the chip, which reduces data travel distance and latency. - This HBM-free design is a strategic choice to sidestep supply chain constraints that are largely controlled by established industry players, a major hurdle for new hardware startups. - The new funding round was led by Hummingbird Ventures, a backer of Revolut and Deliveroo, and brings Olix's total valuation to over $1 billion. - Olix is also developing a software compiler designed to adapt existing AI models to run on its new photonic hardware, a critical step for enabling adoption of the new architecture. - The challenge to Nvidia is significant, as the established company controls more than 90% of the market for AI GPUs used in both training and inference. - The AI inference market is forecast to grow from approximately $106 billion in 2025 to over $254 billion by 2030, driven by the enterprise deployment of generative AI and large language models. - Other startups, such as Ayar Labs, are also developing photonic interconnects, but Olix is focused on delivering a complete processing unit that integrates photonics with a novel memory architecture.
