Salience Labs raises $125M for photonic AI data centre chips

Salience Labs, a developer of photonic chips for AI data centre networking, announced on 14 October 2026 that it has closed a $125 million Series B round. The company said the round was oversubscribed and led by Bullhound Capital. The money will go towards deploying its optical switching technology and expanding manufacturing.

New backers include AMD Ventures, Intel Capital, SBI Group, Hudson River Trading and British Business Bank. Returning investors Applied Ventures, Oxford Science Enterprises, Cambridge Innovation Capital and Silicon Catalyst Ventures also took part. Callum Stewart, a principal at Bullhound Capital, will join the company's board as part of the deal.

Linking GPUs with light

The company builds optical circuit switching (OCS) technology aimed at improving how GPUs communicate inside AI facilities. Its integrated, all-optical architecture is meant to connect thousands of GPUs spread across multiple racks, targeting the rising strain on network performance, efficiency and power use as AI infrastructure grows. Salience Labs says the design draws on more than ten years of silicon photonics research and is suited to high-volume production.

"OCS will play a critical role in AI datacenter networking," said Vaysh Kewada, the company's CEO and co-founder. She argued that as customers adopt optical links, high-performance optical switches will be needed to deliver the token economics that AI workloads require.

The raise follows the company's showing of a 64-port all-optical photonic switch at ECOC 2026, which Salience Labs describes as an industry first for optical circuit switching based on silicon photonics. It is also working on 128-port and 256-port versions for larger networks and other data-heavy uses.

Manufacturing and US expansion

To scale production, Salience Labs is working with partners including Keysight Technologies and Tower Semiconductor. The company plans to add a second West Coast site to strengthen its Silicon Valley presence, move into further geographic markets and hire more engineers. The funding will also support development of its next-generation switches.

Training and serving large AI models depends on thousands of GPUs exchanging data quickly and with minimal energy, which has turned networking into a growing bottleneck for data centre operators. The participation of AMD's and Intel's venture arms suggests major chipmakers are watching optical interconnects closely as an alternative to conventional electrical switching.