UK AI chip startup OLIX raises €270.5 million at €2.8 billion valuation just two years after its founding

London-based AI chip startup OLIX has announced a significant €270.5 million ($312 million) Series B funding round, pushing its valuation to €2.8 billion ($3.3 billion). This investment comes just two years after the company's inception. Notable investors in the round include Fundomo, Arm, Hudson River Trading, and prominent angel investors such as Reed Hastings, co-founder of Netflix.

Existing investors have also increased their stakes. To further bolster its leadership, OLIX has added Professor Nick McKeown, a renowned figure in computer science and electrical engineering from Stanford University and a recipient of the 2025 Marconi Prize, to its board of directors. Additionally, former Wise executive Matt Briers has joined as Chief Financial Officer.

Founded in 2024, OLIX boasts a unique approach to AI inference hardware. Unlike other datacenters that treat AI inference as a one-size-fits-all process, OLIX designs every component of its systems, including the chips, lasers, and network. The startup argues that the industry's current strategy of using general-purpose chips for every stage of AI inference is reaching its efficiency limits.

OLIX advocates for specialized chips for each stage of the "token production" process, believing this approach can lead to a significant leap in AI performance and cost reduction.

Central to OLIX's strategy is its X-1 platform, which employs a novel "slow and wide" optical interconnect, replacing traditional copper connections with light for ultra-low latency and energy-efficient data transfer between chips. This design, combined with a fully deterministic compiler and a flexible compute fabric, allows each chip in OLIX's X-1 platform to focus on a single part of the model, creating a production line for AI model reasoning and generation.

The company's DX-1 chip, part of the X-1 platform, is designed for 100B parameter models and can achieve Pareto-optimal inference performance, delivering over 10,000 Tokens per second per user. This level of performance is achieved with higher output token throughput per watt compared to general-purpose chips running large batch sizes.

OLIX's modular approach allows the DX-1 chip to scale effectively, supporting models up to 10 trillion parameters in size. The company's multi-rack scale-up domain architecture ensures that the system remains performant and cost-effective, even as model sizes grow. By avoiding the use of advanced packaging and high-bandwidth memory, components that are currently in short supply, OLIX aims to scale production despite industry-wide supply chain challenges.

The funding will be used to deliver the DX-1 chip to OLIX's first customers by the second half of 2027, as well as to expand the custom silicon platform and address the manufacturing and supply chain needs necessary for scaling frontier AI inference hardware. The company is actively hiring across various engineering disciplines in London, Bristol, Austin, Toronto, and San Francisco to support its ambitious growth plans.