1 hour ago
17
Olix Computing, a UK-based startup building photonic AI chips, just closed a $312 million Series B that values the company at $3.3 billion. That is a staggering jump from roughly $1 billion just months ago, when the company raised $220 million in February 2026.
Olix, formerly known as Flux Computing before rebranding in January 2026, makes what it calls Optical Tensor Processing Units, or OTPUs. Think of them as chips that use light instead of electricity to process the math that powers AI models.
Traditional GPUs from Nvidia crunch AI workloads using electrons flowing through silicon. Olix wants to do the same thing with photons, which travel faster and generate less heat. The company claims this approach is faster, cheaper, and more energy-efficient than conventional GPU setups.
The key innovation is eliminating the need for high-bandwidth memory, or HBM, which is the expensive and supply-constrained component that makes Nvidia’s data center GPUs so costly. Olix’s architecture reimagines the data center as what the company describes as a “token factory,” essentially a facility purpose-built for churning out AI inference at scale.
The founder behind all of this is James Dacombe, a 25-year-old Thiel Fellow who started the company in March 2024. He is now sitting on a $3.3 billion valuation roughly 15 months after founding the venture.
The round was backed by Hummingbird Ventures, Plural, Vertex Ventures, and Entrepreneurs First, among others. Total funding now surpasses $300 million, and the company has plans to expand into the US and Canada with a focus on what it calls “AI infrastructure sovereignty.”
Olix is not the only company trying to dethrone Nvidia’s dominance in AI compute. Cerebras, Groq, and several other startups have pitched alternative chip architectures. What makes Olix distinct is its photonic approach, which is genuinely different from the silicon-based alternatives that still rely on similar memory architectures as Nvidia.
The $3.3 billion valuation after a Series B is aggressive by any standard. For context, Cerebras was valued at around $4 billion before its IPO filing, and that company had years more revenue history.
The risk is straightforward: photonic computing has been “five years away” for roughly two decades. Academic labs have demonstrated the physics, but manufacturing photonic chips at scale with competitive yields remains an unsolved engineering challenge.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·