Salience Labs has raised $125 million in an oversubscribed Series B round to scale the optical switches it says will rewire how artificial intelligence data centers move data. The Oxford-based company announced the round on October 14, led by Bullhound Capital, with AMD Ventures and Intel Capital among the new investors. The money will fund volume manufacturing, larger switches and a bigger United States presence, as the company tries to turn a laboratory technology in silicon photonics into production hardware for the largest AI systems.
What the round buys
Bullhound Capital led the round, and its principal Callum Stewart joins the Salience board. SBI Group, Hudson River Trading and the British Business Bank took part as new investors, alongside backers Applied Ventures, Oxford Science Enterprises, Cambridge Innovation Capital and Silicon Catalyst Ventures. The breadth of that list matters, because it mixes venture capital, chip-industry money and a state-backed bank.
Salience says the money will accelerate deployment of its optical circuit switching technology, scale volume manufacturing and expand its United States presence in Silicon Valley, where it plans a second West Coast location. Manufacturing scale is the real test of this round, and it will decide whether the technology reaches production at all.
A switch design that works in a demonstration still has to be produced in volume, at acceptable yield, before any large data center operator will build around it. That production step is where many hardware startups stall, because yield, packaging and testing all have to work together at scale.
The company will also keep working with Keysight Technologies and Tower Semiconductor as it scales manufacturing for high volume. Those engagements are about production, not just research. Salience was founded in 2021, building on more than a decade of silicon photonics research at the University of Oxford and the University of Münster in Germany.
Why light, not copper
Modern AI training spreads one job across thousands of graphics processors, often across many racks, and the connections between those processors can decide how fast the whole system runs. Electrical switching has carried data centers for years, but it converts signals between light and electricity, and each conversion costs power and adds delay.
Optical circuit switching keeps the signal as light for longer, which is why its supporters claim better latency, throughput and energy use. If those claims hold in production systems, the gain would show up as faster training runs and lower power bills, not just as a cleaner network diagram.
Salience describes its architecture as all-optical and fully integrated on a chip, purpose-built to connect thousands of graphics processors across multiple racks for scale-up performance. That is a company claim, not an independent measurement, and it should be read as one.
The announcement gives no deployment figures, no customer names and no energy-saving numbers from a production system, so the performance case still rests on the company’s own demonstrations rather than on third-party results. Independent testing will be the real check on the claims made for this architecture.
The switch so far
The financing follows a demonstration of a 64-port all-optical photonic switch at ECOC 2026, which Salience calls an industry first for silicon photonics optical circuit switching. The funding will enable 128-port and 256-port switches, aimed at data-intensive environments where larger fabrics are needed. Bigger port counts matter because they let one switch serve more connections without adding layers of equipment.
Chief executive Vaysh Kewada said optical circuit switching will play a critical role in AI data center networking, and that the funding will accelerate growth into high-volume production. Bullhound’s Stewart said the investor base speaks to the leadership’s capability, while AMD Ventures said AI infrastructure is increasingly designed at rack scale, where networking directly affects performance, efficiency and economics.
Our read
The most telling detail is not the $125 million itself, but who put it in. AMD Ventures and Intel Capital rarely back the same small company, and their joint presence signals that both chip giants see optical switching as a serious answer to the networking bottleneck.
Even so, neither investor has committed to buying the switches, so this is investor validation, not a customer order. That distinction matters for anyone reading this round as proof of adoption, because funding and purchasing are two very different decisions.
There is also a timing question. The round was announced on October 14, and it surfaced again in trade coverage on November 2, which is how it reached this cycle after earlier cycles missed it. Nothing in the announcement has changed since then, and no new terms have emerged.
No valuation was disclosed, no customers were named, and the larger switches are still future products. Readers should treat this as a well-funded manufacturing bet, not as a finished product story, and judge it on production milestones next, not on launch-day claims.
Funding for AI hardware has stayed strong even as investors grow choosier about AI spending elsewhere, and Salience fits that pattern. Money is flowing to the physical layer that AI depends on, from chips to power to networking, because those layers set the limits.
A comparable recent hardware bet was FuriosaAI’s large inference round, covered in our FuriosaAI funding report, which shows how crowded the race to supply AI infrastructure has become. Salience is a smaller bet in the same race, aimed at a different bottleneck.
Why It Matters
If optical switching works at scale, it changes the economics of building AI data centers, because networking power and latency are becoming first-order constraints alongside chips and electricity. A cheaper, faster network fabric would lower the cost of every large training run.
Dell’Oro Group, cited in the company announcement, forecasts cumulative spending on data center switches for AI back-end networks approaching $1 trillion over 2026 to 2030, with scale-up accounting for more than half of that market by 2030. That forecast is Dell’Oro’s, relayed by Salience, not an independent finding of this report.
The financing context matters too. Large AI infrastructure deals are increasingly funded through a mix of equity, debt and credit support, as seen in the Nvidia credit talks for AI infrastructure. Equity rounds like this one are only part of that mix, and they usually arrive before larger facilities fund construction.
Salience’s equity round is smaller than those facilities, but it sits in the same story: the industry is paying heavily to remove bottlenecks, and networking is now one of the bottlenecks getting funded. That context is why a $125 million round can matter beyond its size.
Background
Salience Labs is an Oxford-based photonics company founded in 2021, and its switches aim to move data between AI processors as light rather than electricity. The Series B round, announced on October 14, was led by Bullhound Capital with AMD Ventures and Intel Capital participating, and the company plans to use it to scale manufacturing and build larger switches.
Source: Salience Labs announcement and Evertiq report
0 Comments