James Dacombe founded Flux Computing in March 2024, later rebranding it to Olix Computing Ltd. in January 2026 after securing a $220 million funding round. The company has now raised a total of $250 million to develop its Optical Tensor Processing Unit (OTPU), an AI accelerator designed specifically for inference workloads. Olix integrates SRAM memory architecture with photonics to achieve higher throughput per megawatt and lower total cost of ownership compared to HBM-based designs. By deliberately avoiding HBM, advanced packaging, and other supply-constrained technologies, Olix aims to sidestep the bottlenecks that limit competitors. The company claims its optical digital processor remains compatible with existing AI models while addressing the rising compute demands of modern AI agents and reasoning tools. Dacombe also leads CoMind, a separate brain-monitoring start-up he started as a teenager that has raised $100 million. At 25, he now runs two distinct companies with a combined total of roughly $350 million in funding.
Flux Computing (originally registered as Flux Corp Ltd in March 2024 by 25-year-old founder James Dacombe) changed its name to Olix Computing Ltd. on January 19, 2026. The rebrand was associated with securing a significant $220 million funding round to scale its AI hardware and chip design efforts.
That is a staggering milestone for a company that launched in 2024 and has now raised a total of $250 million, aiming to build AI chips faster and cheaper than Nvidia’s.
Olix is not trying to beat Nvidia at its own game. Instead of tweaking the existing GPU architecture that has powered AI for years, the company is building what it calls the Optical Tensor Processing Unit (OTPU), a new type of AI accelerator designed for inference workloads.
The OTPU integrates SRAM memory architecture with photonics, an approach that Olix says enables it to surpass HBM-based architectures in both throughput per megawatt and total cost of ownership. By deliberately avoiding HBM, Olix thinks it has also sidestepped the supply chain bottlenecks that have constrained virtually every other chip company trying to compete with Nvidia.
As Olix wrote in its recently published Compute Manifesto: From a supply chain and fabrication perspective, a new architecture must avoid HBM, advanced packaging, or any other technology that is supply-chain-constrained by current incumbents. When even the biggest hyperscalers are struggling to secure capacity, a start-up simply cannot compete on those terms.
This matters because inference, the process of actually running trained AI models, is becoming increasingly expensive. The latest generation of AI agents and reasoning tools consumes far more computing power than earlier chatbots. Every time an AI tool reasons through a problem or conducts deeper research before responding, the compute bill climbs significantly.
Olix says it is building a new class of accelerator designed for high performance that is free from the architectural and supply chain constraints of today’s AI processors.
The company claims it is developing an optical digital processor with a novel memory and interconnect architecture that will still be compatible with existing AI models.
“One of the biggest constraints in AI today is the compute required to run these models at scale,” said Jonathan Heiliger, general partner at Vertex Ventures, a former Facebook infrastructure executive. “Today’s GPU-based approach forces a compromise between speed and cost,” he added. “Olix is taking a radically different approach designed to deliver a step change in both, and it has huge promise.”
Olix, hopes to deliver its first products to customers as soon as next year.
No start-up has yet succeeded in significantly loosening Nvidia’s grip on the AI market. But Nvidia’s recent deal with start-up rival Groq, which brought in one of the original founders of Google’s AI chip program, has helped to reignite investor interest in a category long written off as too risky and capital-intensive.
“AI inference demands a ground-up reconsideration” of how chips are built, said Heiliger, who previously helped launch Meta’s Open Compute Project. “This magnitude of systems-level rearchitecting is brutally hard, and James and team are executing faster than companies with 10 times their resources.”
Figure 1. James Dacombe is also chief executive of brain monitoring start-up CoMind. (Source: Charlie Bibby/FT)
Olix isn’t the only or the first to pursue optical computing and coupling. Neurophos, Lightmatter, OptoML, Rayd, and Volantis, which are discussed in our AI Processors Market report, have similar ideas. The question is, will the problem they solve be the one people care about by the time they do it?
Dacombe does not qualify as a first-time founder. He also serves as chief executive of CoMind, a brain-monitoring company he launched as a teenager that has raised $100 million in funding. At 25, he now runs two companies with a combined capital raise of roughly $350 million.
Dacombe’s execution at CoMind gave investors confidence in his capacity to handle a similarly complex effort at a young age. Plural characterizes his method as pursuing contrarian theses and recruiting top-tier talent to convert those theses into mainstream positions. CoMind does not get any of the proceeds from the $220 million funding round recently secured by Olix (an AI chip start-up).
While both companies are founded by Dacombe, they operate as separate entities with distinct investors, purposes, and funding pools.
Here’s why CoMind does not get proceeds from Olix’s investment:
Distinct legal entities: CoMind and Olix are separate companies founded by James Dacombe.
Separate funding rounds: The $220 million investment mentioned in the context of a “100m” (sometimes cited as “$220mn”) round is for Olix, which is developing AI chips. CoMind had a separate funding round of $102.5 million to advance its non-invasive brain-monitoring device.
Different purposes: The Olix funds are dedicated to developing “Optical Tensor Processing Units.” The CoMind funds are earmarked for developing “CoMind One” (brain monitoring technology) and securing US regulatory approval.
Separate investor bases: While some investors (like Plural, LocalGlobe, and Entrepreneurs First) have backed both, the capital pools are not commingled.
The investors backing Olix include Plural, Vertex Ventures, LocalGlobe, and Entrepreneurs First, alongside lead investor Hummingbird Ventures. Hummingbird’s portfolio includes names like Kraken, Revolut, and Deliveroo. Taavet Hinrikus, co-founder of Wise and general partner at Plural, confirmed he led Plural’s investment into Olix, stating that the world needs vastly more compute and that current hardware is reaching a physical wall.
Olix has grown from a founding team to over 70 employees and is aiming to scale to over 200 this year.
LIKE WHAT YOU’RE READING? INTRODUCE US TO YOUR FRIENDS AND COLLEAGUES.