Salience Labs Raises $125M for Optical AI Networking

Salience Labs has raised $125 million to scale silicon-photonic switching technology aimed at one of AI computing's most important bottlenecks: moving data efficiently between accelerators.

What happened

The Oxford-based company closed an oversubscribed Series B led by Bullhound Capital, with participation from investors including AMD Ventures, Intel Capital, SBI Group, Hudson River Trading, the British Business Bank and existing shareholders. Salience develops optical circuit switches that route information using light rather than conventional electronic switching. Its technology is designed for large AI clusters connecting thousands of GPUs across multiple racks. The company demonstrated a 64-port switch in September 2026 and plans to use the new capital for volume manufacturing, commercial deployments and development of larger 128- and 256-port systems.

Why it matters

AI cluster performance depends heavily on networking. GPUs can sit idle if data cannot move between them fast enough, while electronic interconnects consume significant power. Optical switching could improve bandwidth and energy efficiency, especially as clusters become larger.

The bigger picture

The AI hardware opportunity extends well beyond processors into networking, memory and power delivery. Salience is targeting this supporting infrastructure layer. Commercial success will depend on manufacturing yield, reliability and whether optical systems deliver enough economic advantage to justify changes in data-centre architecture.