Salience Labs Just Raised $125M to Replace the Wires Between Your GPUs With Light

On October 14, 2026, Oxford-based Salience Labs announced it had closed an oversubscribed $125 million Series B funding round to scale its photonic chips for AI datacenter connectivity. The round was led by Bullhound Capital, whose principal Callum Stewart will join the company's board of directors. What makes the raise notable is not just the size, but who showed up: both AMD Ventures and Intel Capital participated, alongside SBI Group, Hudson River Trading, and the British Business Bank. Two of the chip industry's biggest names just placed the same bet.

The bet is on a deceptively simple idea. Inside today's AI datacenters, data still shuttles between GPUs through conventional electronic switches, with repeated conversions from optical to electrical signals and back again. Salience Labs wants to eliminate the middleman entirely, keeping data in the optical domain as it moves between connections.

Why the network is the new bottleneck

Adding more GPUs does not automatically make an AI system faster. When thousands of processors have to exchange data, the network determines how much useful work those expensive chips actually complete. Modern AI clusters have tens of thousands of GPUs and other accelerators, and the interconnect bandwidth available to link them is seriously lagging behind the raw compute capacity now available, according to co-founder and CEO Vaysh Kewada.

The result is a familiar set of symptoms: latency bottlenecks and growing power consumption that drive up the cost of every AI workload. In industry terms, it is the token economics that suffer. Every wasted millisecond in the network fabric is compute you paid for and did not get to use.

This is where optical circuit switching, or OCS, enters the picture. Instead of converting light into electricity, switching it, and converting it back, Salience's approach keeps data as light through the switch itself. The company has built an all-optical, fully integrated switch architecture on silicon photonics, purpose-built to connect thousands of GPUs across multiple racks for optimal scale-up network performance.

The bet is on a deceptively simple idea: keep data as light through the switch itself, and the network stops being the bottleneck.

What Salience is actually shipping

The funding follows the company's demonstration of a 64-port all-optical photonic switch at ECOC 2026, which Salience describes as an industry first for silicon. The switch is a fully integrated on-chip solution, built on more than a decade of silicon photonics research, and designed from the start for high-volume manufacturing rather than as a lab curiosity.

The roadmap goes further: 128-port and 256-port products are in development. Port density matters because datacenter fabrics grow by adding connections; a switch that packs more ports onto a single integrated chip means fewer boxes, fewer interconnects, and less power per switched gigabit. For hyperscalers planning clusters of hundreds of thousands of GPUs, that math compounds quickly.

The technology's selling point is manufacturing compatibility. Silicon photonics lets the optical components be fabricated with the same processes used for conventional chips, which is what makes high-volume production plausible. That is the part investors seem to be underwriting: not a research result, but a manufacturing story.

The money and the believers

The full investor list tells a story of its own. New money came from Bullhound Capital, AMD Ventures, Intel Capital, SBI Group, Hudson River Trading, and the British Business Bank. Existing shareholders also returned, including Applied Ventures, Oxford Science Enterprises, Cambridge Innovation Capital, and Silicon Catalyst Ventures. When a round is oversubscribed and both chip giants' venture arms join, the market is signaling that datacenter networking has become a strategic chokepoint worth owning.

The stated use of funds is threefold: accelerate deployment of the OCS technology, scale volume manufacturing operations, and expand the company's U.S. presence in Silicon Valley. The UK-to-Silicon-Valley axis is deliberate; the customers buying this technology are overwhelmingly American hyperscalers, and being close to them matters.

What to watch

Several questions will determine whether this round ages well. First, manufacturing execution: silicon photonics at scale is notoriously hard, and the history of optical networking is littered with beautiful technology that never reached volume economics. Second, customer validation: the company's language is about customers rolling out optical links in their datacenters, but named deployments will be the real proof. Third, the competitive response: incumbent networking vendors are not standing still, and co-packaged optics and linear pluggable optics are alternative routes to the same goal.

Still, the direction of travel is hard to argue with. AI infrastructure spending is increasingly gated by power and interconnects, not by chip supply. The companies that fix the fabric between the GPUs capture value out of proportion to their size. Salience just got $125 million to try.

References

Tech.eu: Salience Labs secures $125M to scale photonic chips for AI data centres. Salience Labs: Salience Labs Secures $125 Million in Series B Funding. Unite.AI: Salience Labs Raises $125M Series B to Scale Optical Switching for AI Datacenters. SiliconANGLE: Salience Labs raises $125M after showcasing fully optical circuit-switching tech.