Oxford-based Salience Labs has raised $125 million in Series B funding to scale photonic networking technology designed to reduce one of the fastest-growing bottlenecks inside AI data centers: moving information between increasingly large numbers of GPUs.
The oversubscribed round was led by Bullhound Capital, with participation from AMD Ventures, Intel Capital, SBI Group, Hudson River Trading and the British Business Bank, alongside existing investors including Applied Ventures, Oxford Science Enterprises and Cambridge Innovation Capital. Salience said the financing will support volume manufacturing, deployment of its optical circuit switching technology and expansion in Silicon Valley.
The British Business Bank separately disclosed that it invested £10 million in the round and described Salience as an Oxford spinout building optical switches intended to reduce the speed and efficiency constraints involved in transferring data between GPUs. The bank said faster communication between GPUs is becoming increasingly important as AI demand grows.
AI’s bottleneck is shifting from compute to networking
The funding reflects a broader problem emerging inside large AI clusters.
Adding GPUs increases theoretical computing power, but those processors still need to exchange enormous amounts of data. Traditional electronic network switches introduce latency, consume energy and can become an increasingly important constraint as systems scale.
Salience is developing optical circuit switches, or OCS, that keep signals in the optical domain rather than repeatedly converting them between light and electricity.
The company recently demonstrated a 64-port all-optical silicon-photonics switch at ECOC 2026 in Málaga and says future systems will include 128-port and 256-port configurations. Independent industry publication Optics.org reported that the technology is intended to improve network efficiency while reducing power consumption and cost.
CEO and co-founder Vaysh Kewada said optical circuit switching will play a critical role in AI data-center networking as operators deploy more optical links between compute resources. She said Salience’s integrated on-chip architecture is designed to scale into high-volume manufacturing.
AMD and Intel are betting on a new infrastructure layer
The presence of both AMD Ventures and Intel Capital is particularly notable because the companies sit close to the compute layer Salience is trying to connect more efficiently.
Intel Capital managing director Srini Ananth said photonics is becoming essential to scaling AI infrastructure and that optical switching could deliver significant gains in performance and efficiency.
Market research firm Cignal AI estimates the optical circuit switching market will exceed $8 billion by 2030, driven increasingly by GPU scale-up networking. The firm expects OCS revenue to grow at more than 30% annually between 2026 and 2030.
Europe is competing below the model layer
For European technology, the significance goes beyond another large funding round.
Europe’s AI conversation often focuses on models, regulation and access to chips. Salience is targeting a less visible layer: the networking fabric that determines how effectively thousands of expensive processors can operate together.
That layer becomes increasingly important as compute clusters become larger. Buying more GPUs does not automatically translate into proportional performance if communication between them becomes the limiting factor.
The $125 million investment suggests investors and chipmakers increasingly see data movement itself as part of the AI infrastructure race.
For builders, that shifts an important question. The next leap in AI performance may not come solely from faster processors or larger models. It may also depend on how quickly—and efficiently—the machines running those models can talk to one another.
