iPRONICS Raises $125M for Programmable Optical Networking

An AI cluster can contain thousands of expensive accelerators and still lose time to the paths between them. Training jobs expand, inference traffic arrives with a different shape, and the physical network has to connect the right resources before the hardware can stay productive. iPRONICS has raised a $125M Series B to make that network programmable, bringing total disclosed funding to $177M.

Maverick Silicon and Light Street Capital co-led the September 2, 2026 financing, with NVIDIA participating. The syndicate also includes Triatomic Capital, Bosch Ventures, Catalight Capital, the European Innovation Council Fund, The Tate Family Trust, Fine Structure Ventures, Amadeus Capital Partners, Build Collective, and Criteria Venture Tech. iPRONICS plans to use the capital to scale operations and commercial deployment, expand product and customer programs, deepen AI-infrastructure partnerships, and grow its U.S. presence in Santa Clara.

What iPRONICS raised and who backed it

The $125M Series B is the company's largest disclosed financing and follows a €20M Series A announced in January 2025. The new round brings total funding to $177M. iPRONICS did not disclose a valuation, revenue figure, customer roster, deployment count, or timetable for the next commercial milestone.

The board is changing alongside the capitalization. Manish Muthal, senior managing director at Maverick Silicon, and Geoffrey Tate, an adviser to Light Street Capital, will join the board. Young Sohn, a founding member of Catalight Capital and former Samsung corporate president and chief strategy officer, becomes an adviser. Those additions put semiconductor, data-center, and company-building experience closer to a business moving from photonics research into industrial deployment.

Why AI clusters need a network that can change

AI infrastructure is often discussed as a race for more capable accelerators. The network determines whether those accelerators can be assembled into the right system for a changing job. A training workload may need one pattern of connections, while inference or a different model architecture may require another. When the fabric is treated as fixed, expensive compute can wait for data or remain stranded inside an inefficient topology.

Optical circuit switching is designed to move data through light rather than repeatedly converting every path through electrical packet switches. That can create a more direct connection between resources, but useful hardware also needs a control layer that can change those connections as the workload changes. iPRONICS is building at that intersection of silicon photonics, rack-level hardware, telemetry, and network software.

Inside the iPRONICS Optical Networking Engine

The iPRONICS Optical Networking Engine packages a silicon-photonics optical circuit switch with control software, telemetry, and APIs. The rack-mounted system is intended to let data-center teams program optical paths and observe the network as one operating system rather than as a collection of static links. The company says ONE can reconfigure those paths in real time for AI training and inference.

The company's ONE-32 launch materials report 32 optical ports, less than 30 nanoseconds of latency, and reconfiguration in under 300 microseconds. iPRONICS also says the design can reduce switching power by cutting the number of transceivers required in some architectures. Those are company-reported specifications, not independent production benchmarks, and their commercial value will depend on integration, reliability, and behavior under customer workloads.

That qualification matters because optical networking has to fit into an existing operational stack. Data-center teams need APIs, telemetry, failure handling, orchestration, support, and predictable manufacturing in addition to fast photonic hardware. A switch that performs in a controlled test still has to prove that it can improve utilization and simplify operations inside a live cluster.

From a university spinout to a commercial system

iPRONICS was founded in 2019 as a spinout from Universitat Politècnica de València. Founder and CTO Daniel Pérez-López has led the technical work from programmable photonic circuits toward a complete networking system. Christian Dupont joined as CEO in 2024 after operating roles across wireless semiconductors and optical networking companies, giving the business a separate executive focused on commercialization.

The company's earlier Series A funded the move from core technology into products including ONE-32. The Series B assigns much more capital to the next step: scaling the organization, supporting deployments, expanding the product roadmap, and building relationships across the AI infrastructure ecosystem. Its careers page lists the hiring surface created by that ambition.

The investor group also reflects the breadth of the task. Maverick Silicon and NVIDIA bring direct proximity to semiconductors and accelerated computing. Bosch Ventures, the European Innovation Council Fund, and the returning venture investors add industrial, institutional, and earlier-stage continuity. Their participation does not prove product-market fit, but it does show that the financing thesis reaches beyond a single optical component.

What the $125M changes

The capital gives iPRONICS room to treat deployment as part of the product. Silicon photonics requires design, fabrication, packaging, testing, hardware integration, and software coordination. Expanding from promising technical specifications to repeatable customer installations means solving those disciplines together while building a support organization in both Europe and the United States.

It also gives the company a chance to prove a larger category claim. If AI clusters need to be reconfigured continually, optical switching can move from a specialized layer into a strategic control point. The customer would not simply buy lower latency. It would buy the ability to connect accelerators according to the workload and potentially recover more useful time from equipment that already represents a large capital commitment.

The open questions are commercial. iPRONICS has not named customers or disclosed revenue, deployment volume, utilization gains, or independently measured savings. The next evidence should show whether ONE integrates cleanly with data-center software, maintains reliable performance at scale, and produces an economic improvement operators can measure.

AI infrastructure will continue to be judged by model capability and accelerator performance. iPRONICS is placing its $125M Series B one layer between them, where the network decides which resources can work together at any moment. If programmable optics becomes part of the cluster's control plane, the competitive advantage may belong to the system whose fabric can change its mind before its most expensive hardware has to wait.