iPronics ONE Silicon Photonics Optical Circuit Switch

iPronics is a programmable photonics company founded in 2019 as a spin-off from Universitat Politècnica de València, with operations in Valencia, Spain and Santa Clara, California. It builds optical networking hardware for AI data centres, where fixed electrical switching limits bandwidth, latency and energy use between GPU racks.

Its iPronics ONE is an optical circuit switching (OCS) platform built on silicon photonics, using flip-chip photonic integrated circuit bonding and semiconductor optical amplifier integration. It is delivered as OCS modules and 1U rack-mounted systems with integrated control, telemetry and APIs, so operators can reconfigure connectivity within and across racks in real time for training, inference and GPU failover. Customers are hyperscalers and AI-cluster OEMs.

• Rack-mounted optical switch: 1U OCS systems and modules for AI clusters.
• Silicon photonics core: flip-chip PIC bonding with integrated semiconductor optical amplifiers.
• Programmable optical layer: real-time reconfiguration of connectivity within and across racks.
• Control and telemetry: built-in monitoring of switch state and optical links.
• API-driven integration: software interfaces for orchestration with cluster management tools.
• GPU failover: automatic redirection of links to available GPUs.

Vietnamese data centre and cloud operators building GPU clusters can pilot iPronics ONE to reconfigure links between racks. Telecom groups running AI services can use optical switching for GPU failover and resource pooling. Research supercomputing centres can evaluate programmable optical networking as BSC does.

A programmable optical layer lets operators reroute GPU connectivity without recabling, supporting higher bandwidth and lower latency. Optical switching reduces power consumption compared with electrical packet switching paths. GPU failover redirects links to an available GPU to keep workloads running. Validation with Lumentum showed near-transparent performance for 1.6 Tb/s, 200G-per-lane transceivers through the switch.

AI clusters are growing to tens of thousands of GPUs, and hyperscalers are adopting optical circuit switching to cut power and latency in scale-out networks. Silicon photonics makes these switches manufacturable in volume. New AI data centre investment across Southeast Asia, including Vietnam, creates demand for energy-efficient interconnect options.

Founded in 2019 as a spin-off from Universitat Politècnica de València; offices in Valencia, Spain and Santa Clara, California. Raised a €20M Series A (January 2025) and a $125M Series B (September 2026) co-led by Maverick Silicon and Light Street Capital with NVIDIA, bringing total funding to $177M. Manufacturing line at Fabrinet; partners include Lumentum, Ansys and Barcelona Supercomputing Center. Listed in EE Times Silicon 100 (2025).