The costs of AI infrastructure are rising quickly. With the newest AI models applying more computing power, running these trained models (called inference) is becoming very expensive. This has led to the development of new chip designs better suited to efficiently handle inference tquestions.
London-based startup Olix believes it has an answer. The company has secured $220 million in new funding, lifting its valuation above $1 billion, according to reports. The round was led by Hummingbird Ventures and brings Olix’s total funding to roughly $250 million since launching in 2024.
Alongside Hummingbird Ventures, Olix has attracted investors including Plural, LocalGlobe and Entrepreneurs First.
At just 25 years old, founder James Dacombe is now leading one of the UK’s newest unicorns in one of tech’s most capital-intensive sectors.
A different path from Nvidia
Olix is creating a new type of AI accelerator, the Optical Tensor Processing Unit (OTPU), that utilizes photonics rather than traditional GPU architecture, focapplying primarily on inference tquestions.
This chip combines SRAM with photonics to achieve higher efficiency and lower costs than current designs that rely on high-bandwidth memory (HBM). By steering clear of complex memory and packaging methods that pose supply chain issues, Olix aims to avoid the challenges other competitors face, the report adds.
Olix has grown from its founding team to more than 70 employees and plans to expand to over 200 staff this year. The company is recruiting across London, Bristol, Austin, San Francisco and Toronto.
Olix declares it plans to ship its first products to customers by 2027. If successful, it would mark a rare breakthrough for a UK-founded AI chip company in a market defined by scale, capital intensity and technical complexity.
Leave a Reply