Salience Labs Raises $125M Series B for Optical Networking

Tower Semiconductor described its work with Salience Labs as moving into pre-production in February. Eight months later, the Oxford optical-switch developer has $125M in Series B funding to support deployment, manufacturing and a larger Silicon Valley presence.

Salience Labs announced the completed round on October 14, 2026, led by Bullhound Capital with AMD Ventures and Intel Capital among the participants. For CEO Vaysh Kewada and CTO Johannes Feldmann, the capital arrives after a 32-port product launch and a reported 64-port demonstration, as the company works to put silicon photonics into AI data-center networks.

The commercial proposition reaches beyond making a fast switch. Salience is assembling the manufacturing and validation relationships that can help an infrastructure buyer evaluate optical switching as part of an installation, where compatibility, workload performance and dependable supply all arrive on the same agenda.

A Series B with a production job

Bullhound Capital led the oversubscribed financing, and principal Callum Stewart is joining the Salience Labs board. Alongside AMD Ventures and Intel Capital, the announcement names SBI Group, Hudson River Trading and British Business Bank. Returning shareholders include Applied Ventures LLC, Oxford Science Enterprises, Cambridge Innovation Capital and Silicon Catalyst Ventures.

The money is intended to accelerate deployment of Salience's optical circuit switching technology, expand volume manufacturing and build its U.S. presence. The company also plans another West Coast location. Those uses put engineering, production and proximity to potential customers inside the same expansion effort.

Salience previously announced a $30M Series A in February 2025, co-led by Applied Ventures and ICM HPQC Fund. That announcement emphasized developing optical switches compatible with existing AI infrastructure. The new round follows the move from development toward a product family and a broader production plan, while financial terms such as valuation remain undisclosed in the reviewed announcements.

That progression gives the financing a useful commercial context. An AI infrastructure supplier has to support the customer's planning horizon: the system being evaluated today needs a credible path into the installations that follow it. Manufacturing capacity and a growing product roadmap become part of how a buyer assesses that relationship.

What the optical switch puts inside the rack

Salience Labs builds optical circuit switches that keep the switching path in the optical domain. Its OC-32M product page describes a 32-input, 32-output module combining a photonic integrated circuit for switching and amplification with driving electronics, firmware and standard fiber connectors. A rack-mounted evaluation platform packages the technology for customer development work.

The company lists 10 ns of latency from input to output and reconfiguration below 300 microseconds. Those figures describe different operations: the time a signal takes to pass through an established connection, and the time required to change the connection. Neither figure, on its own, tells an operator how quickly an entire AI request will finish.

Salience's technology description emphasizes reducing the conversion work associated with electrical switching and supporting optical interfaces across generations. The operating question is how that architecture behaves inside a customer's workload and physical network. Component specifications provide a starting point for evaluation; useful system performance depends on how the installation uses the component.

The product chronology also needs careful reading. Salience launched its 32-port switch in March 2026 and reports a live 64-port demonstration at ECOC in September. The 128- and 256-port configurations identified in the funding announcement are future solutions. For operators considering larger GPU groups across racks, that sequence makes the company's scaling direction visible without turning roadmap devices into products already delivered.

The foundry and the test bench enter the conversation

Tower Semiconductor's February announcement describes a partnership to manufacture photonic integrated circuit-based optical switches using its PH18DA and TPS45PH platforms. The release identifies a move from development into pre-production. That is a concrete manufacturing relationship around the architecture, with established foundry processes forming part of the route toward scale.

In March, Salience and Keysight announced an optical circuit switch testing collaboration. Keysight described extending its KAI Data Center Builder validation capabilities to optical circuit switching. The collaboration brings the product into an environment intended to help data-center operators evaluate its behavior with AI workloads.

Together, the two relationships address different responsibilities in an infrastructure purchase. A foundry helps establish how a design can be manufactured; a test environment helps investigate how it performs within the intended application. Neither announcement establishes a named customer order or a particular production volume, and the funding disclosure provides no verified revenue or shipment count.

This coordination belongs to a wider AI infrastructure conversation that also appears in DevCuration's coverage of Mueon's Cubelets systems. The architectural approaches differ, but the buyer has to make interconnected decisions about compute, interconnect and the physical installation. A promising individual device enters a system with existing constraints and people responsible for keeping it useful.

Carrying the research into an operating relationship

Founded in 2021, Salience Labs traces its research origins to Oxford and Münster. Its current leadership page identifies Vaysh Kewada as founder and CEO and Johannes Feldmann as founder and CTO. Academic co-founders Harish Bhaskaran and Wolfram Pernice are part of the scientific history, documented in university and scholarly biographies, rather than presented here as current operating executives.

That history matters because the company now has several kinds of work advancing around the same technology. The physics must survive integration into a module, the module must fit a network, and the supply relationship must support the buyer's deployment plans. The new capital supports Salience's attempt to carry those responsibilities together as its port counts and geographic presence grow.

For an operator evaluating Salience, the foundry agreement, validation environment and product roadmap create specific conversations to have. The next installation will bring its own workloads, optical links and expansion schedule, giving Salience's engineers and commercial team a continuing job of fitting the switch to the system the customer actually intends to run.