AI data-center operators pay for faster, lower-power interconnects that keep GPU clusters utilised. iPronics is positioning its silicon-photonics optical circuit switching as a way to remove a growing pain point in AI infrastructure: networking contention inside and across racks as clusters scale.
Spain’s iPronics has raised EUR 125 million in a Series B round, recently announced. The syndicate includes Maverick Silicon, Light Street Capital and NVIDIA, alongside Triatomic Capital, Bosch Ventures, Catalight Capital, the European Innovation Council (EIC) Fund, The Tate Family Trust, Fine Structure Ventures, Amadeus Capital Partners, Build Collective and Criteria Venture Tech.
Why this round fits the current AI infrastructure funding wave
The deal is another signal that venture attention is shifting from training-model headlines to the physical constraints of running AI at scale. iPronics says its core product is silicon-photonics optical circuit switching for AI infrastructure and data-center architectures, explicitly targeting network bottlenecks inside AI clusters.
As GPU counts per cluster rise, the cost of inefficient connectivity shows up quickly as underutilised compute, higher power draw and more complex network design. In that context, iPronics’ pitch is practical: its ONE platform provides programmable optical connectivity within and across racks, aiming for more bandwidth, lower latency and reduced power consumption, while simplifying infrastructure.
Product focus: programmable optical connectivity, not just “faster links”
Optical circuit switching is attractive to hyperscalers and AI infrastructure builders because it can change how traffic is steered without relying purely on ever-larger electronic switching fabrics. iPronics has been explicit about being production-oriented. At OFC 2026, the company said it would showcase the first commercially available silicon-photonics optical circuit switch for AI-driven data-center infrastructures.
Reuters reported the funding will help iPronics shrink a data-center networking chip intended to connect hundreds of thousands of computing chips. That detail matters commercially: moving from lab-grade photonics into data-center form factors and deployment realities is where adoption either accelerates or stalls.
Strategic syndicate: ecosystem alignment around deployment
The mix of backers suggests more than capital. iPronics itself describes support from both strategic and financial investors, and the round includes infrastructure-relevant names such as NVIDIA and Bosch Ventures alongside institutional venture firms.
NVIDIA’s participation is notable in a market where GPU platform roadmaps and networking architectures are tightly coupled. While iPronics has not disclosed commercial agreements in the announcement, the presence of strategic investors typically helps with design-in conversations, partner validation and access to integration pathways. For deep-tech infrastructure components, these routes can be as important as the cheque size.
The EIC Fund’s participation also reinforces iPronics’ European deep-tech profile. iPronics is headquartered in Valencia and was founded in 2019 as a spin-off from the Universitat Politècnica de València.
Likely use of proceeds: productisation and commercial deployment
The company has said it is scaling commercial deployment of ONE across AI data-center architectures. Based on the round’s stated focus and typical requirements for photonics hardware, likely priorities include:
- Engineering to reduce size, power and manufacturability of the switching silicon-photonics package (aligned with Reuters’ note on shrinking the chip).
- Expanding field engineering and integration support for data-center deployments, where qualification cycles and interoperability testing drive timelines.
- Building partnerships across the optical and data-center supply chain to de-risk volume ramp.
Competitive context: crowded interconnect spend, differentiated approach
Data-center networking spend is already intense, but much of it still concentrates on electronic switching and incremental optical upgrades. iPronics is pushing a more architectural change: programmable optical connectivity to relieve congestion and improve cluster scaling. The key commercial question will be whether its system can deliver measurable utilisation and power benefits in real workloads, with operational simplicity that justifies switching costs.
If iPronics can demonstrate repeatable performance gains and a clear integration playbook, it will be selling into a budget line that AI operators increasingly treat as mission-critical: interconnect as the limiter of compute ROI.
What this enables
- Larger AI clusters with less network-induced GPU idle time
- Lower power and complexity pressure as bandwidth demands rise
- A pathway to optical switching that is positioned as production-ready
What to watch
- Evidence of pilot-to-production conversions and deployment scale
- Packaging and manufacturability milestones as the chip is miniaturised
- How strategic investors translate into ecosystem partnerships and design-ins