Olix raises $312M at $3.3B valuation from Netflix’s Reed Hastings, Arm, to build Nvidia rival

James Dacombe was 23 when he started Olix in London back in 2024. Two years on, the company has raised $312 million at a $3.3 billion valuation, more than triple the just-over-$1-billion mark it hit in February. Fundomo, a New York investment firm, led the round. Arm and Hudson River Trading joined in, and so did Reed Hastings, the Netflix co-founder, writing a personal check as an angel investor. A batch of existing backers, including Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court and Transition, all put more money in too.

The raise came with two notable additions to the team. Nick McKeown, the Stanford professor emeritus who helped invent software-defined networking, OpenFlow and P4, joined the board. He's the kind of name that carries weight in chip circles — he co-founded Nicira, later bought by VMware, and Barefoot Networks, acquired by Intel, where he went on to run the networking division. Alongside him comes Matt Briers as CFO, who spent nine years in the same role at Wise. Briers arrived at Wise from Google in 2015, when the company had roughly 500,000 customers and hadn't turned a profit yet. He steered it through its 2021 direct listing in London — the first tech company to do so there, and the biggest tech listing the exchange had seen — at a valuation of £8.75 billion, staying profitable the whole way through.

What Olix is actually selling is a different way of thinking about AI inference. The company frames a data centre running inference as a factory churning out tokens, where each token requires hundreds of operations spread across different kinds of hardware. Rather than throwing one general-purpose chip at the whole job, the way most companies do, Olix's X-1 platform splits the model across multiple specialised chips, each doing one task, connected by a flexible compute fabric rather than a fixed layout. The chips talk to each other over optical interconnects — light instead of copper — which Olix says cuts latency and power draw.

The first product out the door is DX-1, a decode accelerator meant for the stage where a model actually generates its output. Olix claims that for 100-billion-parameter models, DX-1 can push out more than 10,000 tokens per second per user while burning less power than general-purpose chips do on large batches. The architecture is built to stretch to models with 10 trillion parameters or beyond. With over 140 people spread across London, Bristol, Toronto, Austin and San Francisco, the new cash is earmarked for getting DX-1 into customers' hands in the second half of 2027, plus further platform work and ramping up manufacturing.

Olix isn't alone chasing this opportunity. Fractile, out of Bristol, raised $220 million in May. Etched doubled its valuation to $10.3 billion in July. SambaNova pulled in $1 billion at an $11 billion valuation. Olix's $3.3 billion tag, despite tripling since February, still trails those American rivals by a wide margin — even as the AI chip market, valued around $84 billion in 2026, is projected to grow more than 35% a year through 2030. Whether Olix closes that gap will come down to one thing: whether DX-1 actually hits its claimed throughput once real customers start running it in 2027, not how big this funding round was.