Three hundred and twelve million dollars for a chip that computes with light. OLIX, a London startup most people had not heard of a year ago, has closed the largest semiconductor venture round ever raised by a European company: a $312 million Series B valuing the firm at $3.3 billion.
The pace is the story. OLIX raised $220 million as recently as February 2026, at a valuation of roughly $1 billion. Six months later its price has more than tripled, and the company still does not have a product in customers’ hands.
Chips that trade electrons for photons
OLIX builds what it calls Optical Tensor Processing Units, or OTPUs: chips that use light rather than electricity to run the matrix arithmetic behind AI models. The company pairs that silicon with photonic interconnects and what it describes as deterministic rack-scale computing, an architecture aimed squarely at large-scale inference, the business of running trained models rather than training them.
The design also does away with high-bandwidth memory, the scarce and expensive component that has become a chokepoint for conventional accelerators. If OLIX delivers on that claim, it sidesteps one of the tightest supply constraints in the entire industry, and one of the biggest line items on any chip maker’s bill of materials.
Optical computing itself is not new. Researchers have chased light-based processors since the 1980s, because photons generate less heat and move faster than electrons through copper. What has changed is the customer: AI inference workloads are uniform, predictable and gigantic, which gives a specialised architecture something it never had before, a market large enough to justify the manufacturing gamble.
Who wrote the cheques
The round brought in Fundomo, Arm and Hudson River Trading as new investors, alongside existing backers Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court and Transition. Netflix co-founder Reed Hastings joined personally. The UK government’s Sovereign AI venture fund also backed the deal, a signal that Westminster sees the company as industrial strategy as much as a private bet.
Arm’s presence deserves a second look. Britain’s most successful chip company taking a position in what could become Britain’s next one reads as either a hedge or an endorsement, and quite possibly both. Hudson River Trading’s involvement is telling in a different way: high-frequency trading firms buy speed, and deterministic low-latency inference is precisely what OLIX is promising.
The 2027 test
The money is earmarked for finishing DX-1, the company’s first platform, and expanding its custom silicon effort. Initial customer access is planned for the second half of 2027. That is a long runway in a market moving this fast, and it is where the risk lives. Photonic computing has a decades-long history of dazzling demos that never survived contact with manufacturing. OLIX has to prove it can produce OTPUs at volume, at yield, and at a cost that beats the brute-force alternative of simply buying more GPUs.
The prize, if it works, is enormous. Inference is where AI’s operating costs actually live. Every chatbot answer, every generated image, every agent grinding away in a terminal lands on an inference bill somewhere. A chip that meaningfully cuts the energy and cost of that arithmetic attacks the industry’s largest recurring expense.
Competition will not stand still either. Incumbent GPU makers iterate on an annual cadence, and a crowd of other inference-focused silicon startups is chasing the same customers. By late 2027 OLIX will be measured not against today’s accelerators but against whatever ships alongside DX-1, a moving target that has buried more than one ambitious chip company before it reached volume production.
Europe’s bet on hard tech
The round says as much about the market as about the company. Investors in 2026 are still writing enormous cheques, but increasingly for difficult technical problems rather than thin application layers built on someone else’s model. Capital is flowing toward teams that own physics, fabs and infrastructure.
For European tech, which has spent years watching its most promising AI companies get acquired or outgunned, a homegrown semiconductor firm valued at $3.3 billion before its first customer shipment is rare leverage in an industry dominated by American and Taiwanese giants. Whether OLIX becomes Europe’s answer to the inference problem or another entry in photonics’ long list of near-misses will start to become clear when DX-1 reaches its first customers in late 2027. For more coverage of the startups building AI’s infrastructure, visit Mylistingo.
