This UK AI Chip Startup Just Raised $312M in Two Years

·
Listen to this article~5 min

OLIX, a London-based AI chip startup, just raised $312 million in a Series B round at a $3.3 billion valuation, just two years after its founding. The round included Arm, Fundomo, and Reed Hastings.

It's not every day a startup hits a $3.3 billion valuation just two years after opening its doors. But that's exactly what OLIX, a London-based AI chip company, just pulled off with a massive Series B round. The company, which is building the infrastructure for frontier AI inference, announced it raised $312 million (€270.5 million) at a $3.3 billion valuation. That's a lightning-fast rise for a company that didn't exist in 2024. ### Who's backing this big bet? The funding round brought in some serious names. Fundomo, Arm, and Hudson River Trading all participated, along with angel investor Reed Hastings, the co-founder of Netflix. Existing investors also doubled down on their commitments, which is usually a good sign. But the new money isn't the only headline. OLIX also made two key hires to strengthen its leadership team. Professor Nick McKeown, a Stanford University professor emeritus and winner of the 2025 Marconi Prize, is joining the board of directors. The company also brought on Matt Briers, a former Wise executive, as its new Chief Financial Officer. ### The problem with today's AI chips OLIX's core thesis is pretty simple: the industry's current approach to building AI inference hardware is hitting a wall. The company put it this way: "A datacenter is a factory whose product is the token." Think about that for a second. Producing a single token takes hundreds of operations, each with different demands on the hardware. In any other factory, you'd give each stage a machine built specifically for it. But in today's AI datacenters, every stage runs on the same general-purpose chip. Each new chip generation tries to fix this by pushing single-chip specs higher. The result? A better generalist, but never a specialist. OLIX thinks that's the wrong approach. ### A different way to build chips OLIX designs every part of its systems: the chips, the lasers, and the network that connects them. The company's X-1 platform takes a completely different approach. Instead of running everything on one chip, models are "fully unrolled" across a large number of chips. This creates a production line where each chip focuses on one part of the model. Because AI model architectures keep evolving, each chip keeps a flexible compute fabric. Nothing is hard-coded or baked in for a specific model, so the hardware can adapt as the software changes. The system also uses a novel "slow and wide" optical interconnect. Instead of moving data between chips using copper, it uses light. This cuts latency and energy costs dramatically, thanks to rack-scale codesign across every part of the link. ### What's coming next The first chip in this platform is called DX-1, a decode accelerator designed for the stage where a model reasons and generates output. For models with 100 billion parameters, DX-1 can deliver over 10,000 tokens per second per user, all while using less energy per token than general-purpose chips running large batch sizes. Here's what makes this particularly interesting: - The architecture scales to models with 10 trillion parameters or more - It holds the model in fast on-chip SRAM memory for better efficiency - It avoids advanced packaging and high-bandwidth memory, which are in short supply across the industry That last point matters. The semiconductor industry is dealing with serious supply chain shortages, and OLIX designed its chip to scale volumes despite those constraints. ### The road ahead OLIX plans to use this funding to deliver DX-1 to its first customers by the second half of 2027. The company is also building out its wider custom silicon platform and securing the manufacturing and supply chain commitments that scaling frontier inference hardware requires. If you're interested in this space, OLIX is hiring across silicon, photonics, compiler, and systems engineering in London, Bristol, Austin, Toronto, and San Francisco. The big picture here is that OLIX believes its approach will make frontier AI more affordable and abundant. More importantly, the company thinks it will unlock the deployment of far more powerful models in the future. That's a bold claim, but with this kind of backing, it's clear the market is paying attention.

📌 Worth Reading Next