London-based OLIX has raised a $312 million Series B at a $3.3 billion valuation, one of the largest in the European/UK AI space today. The raise included Fundomo, Arm, and Hudson River Trading as well as angels like Reed Hastings, the co-founder of Netflix.
The funding will go towards delivering the OLIX Decode Accelerator 1 (DX-1), a custom silicon solution that assists in AI inference. The company plans to expand manufacturing and supply chain systems that their hardware and software solutions will require.
“OLIX believes the industry’s approach to building AI inference hardware is reaching its efficiency limits. A datacenter is a factory whose product is the token,” wrote the company. “Producing a single token takes hundreds of operations, each placing different demands on hardware. Any other factory would give each stage a machine built for it. Instead, the token factory runs every stage on the same general-purpose chip. Each new chip generation has improved by pushing single-chip specifications higher: an even better generalist but never a specialist.”
OLIX also added Professor Nick McKeown to its board of directors. McKeown is one of the pioneers of modern computer networking. He is a professor emeritus at Stanford University and visiting professor at the University of Oxford. He helped develop key technologies behind software-defined networking (SDN), OpenFlow, and programmable network switches, ideas that reshaped how data centres and cloud infrastructure operate.
McKeown has founded several successful networking companies, including Nicira and Barefoot Networks, both of which were acquired by major technology firms and received the 2025 Marconi Prize for his research in networking and communications.
The Decode Accelerator 1
OLIX’s goal is to reduce the complexity of systems needed to run AI software. Their theory is that standard GPU usage is inefficient and wasteful and that simpler, AI-first chips can help reduce the infrastructure costs.
The company’s DX-1 is OLIX’s first AI chip. It is built for one specific job: generating the next token during an AI inference process, which essentially means it processes and outputs data during a standard AI session. Instead of trying to be a general-purpose AI processor, the DX-1 is designed to make large language models respond faster while serving many users at the same time. OLIX says it achieves this by redesigning how data moves through the system, combining cheaper, faster memory, photonics, and commodity hardware rather than relying on the expensive architecture used in today’s GPU solutions. In short, it greases the skids for AI interaction.
Interestingly, because of the DX-1’s design, the company can scale their chip fabrication quickly without running into memory shortages that other manufacturers are currently facing. They have reduced the complexity of AI chips considerably, a feat that promises to create a unique way to handle the expensive and complex AI interactions needed in military and government.
“OLIX’s systems are built on the belief that using specialised chips for each stage of the token production process will unlock a step change in AI performance and cost,” wrote the company. “In OLIX’s X-1 platform, models are fully unrolled across a large number of chips, creating a production line which allows each chip to focus on a single part of the model.”
OLIX’s 25-year-old CEO, James Dacombe, founded OLIX after dropping out of college. Originally from Harrogate, North Yorkshire, he taught himself programming as a teenager and left sixth form after receiving the Thiel Fellowship. Before launching OLIX in 2024, Dacombe founded the neurotechnology company CoMind, which develops brain monitoring systems.
