iPronics and BSC collaborate on optical networking for AI infrastructure

The silicon photonics firm will integrate its programmable optical switching platform into BSC's GPU and HPC infrastructure over two years.

By The Datatech Times Editorial Team AI level 2 of 5
AI level 2 of 5: drafted by our AI editorial assistant from the company's announcement; fact-checked, edited and signed off by Mark Walker, Editorial Director. What the levels mean
A robotic arm grips a circular silicon wafer patterned with multiple integrated circuits, set in a brightly lit cleanroom with a blurred person in a cleanroom suit in the background.

iPronics, the Valencia-founded silicon photonics startup, has announced a two-year strategic technology collaboration with the Barcelona Supercomputing Center (BSC-CNS) to advance programmable optical networking for GPU-based AI and high-performance computing infrastructure. The partnership will see iPronics' optical circuit switching platform deployed within BSC's research-class hardware estate, with both teams jointly developing workload-aware networking approaches targeting AI training and inference workloads, including large language models and mixture-of-experts architectures.

The collaboration is centred on integrating optical hardware and software into a unified AI networking architecture. iPronics will contribute its programmable switching platform, low-level software and APIs; BSC will develop higher-level software above the switch-management layer. The stated goal is dynamic, workload-aware network control that can reconfigure optical connectivity in response to shifting AI application demands, without requiring cloud operators to replace their existing software stacks.

The deal

Christian Dupont, chief executive of iPronics, said the partnership marks "an important milestone in bringing programmable optical networking deeper into the AI infrastructure stack," adding that scaling compute now requires rethinking the network. Antonio J. Peña, leading researcher and group leader at BSC, said the collaboration offers the opportunity to explore how programmable optics and the software layer can work together to support more dynamic computing infrastructures.

iPronics' flagship product, iPronics ONE, is positioned by the company as a rack-ready, plug-and-play optical circuit switch built on silicon photonics, with integrated control, telemetry and APIs for real-time reconfiguration. The release claims it can improve GPU utilisation and reduce energy consumption without replacing incumbents' network software stacks, though no benchmark figures or measured energy savings were published alongside the announcement.

The BSC collaboration follows iPronics' $125 million Series B round, co-led by Maverick Silicon and Light Street Capital with participation from NVIDIA, which brought the company's total disclosed funding to $177 million. iPronics has also recently opened a US office in Santa Clara to expand customer deployments and AI infrastructure partnerships.

Market context

Optical circuit switching for AI data centres is a fast-moving area attracting significant venture and strategic capital. The core proposition is that conventional electrical packet-switching fabrics introduce latency and power overhead that become increasingly costly as GPU cluster sizes scale into the tens of thousands for frontier model training. Several well-funded startups are pursuing silicon photonics and related approaches for intra- and inter-rack connectivity, and hyperscalers including Google and Microsoft have published research on in-house optical switching architectures, indicating that the category is maturing from research curiosity to near-term procurement consideration.

NVIDIA's participation as an investor in iPronics is notable given that NVIDIA's GPU hardware is the primary compute substrate the platform is designed to serve. Strategic alignment between the interconnect layer and the compute layer is increasingly seen as a competitive differentiator for AI infrastructure vendors, and BSC's role as host of the MareNostrum 5 supercomputer and coordinator of EuroHPC JU gives the collaboration credibility as a proving ground at meaningful scale.

Regulatory and standards read-across

EU digital sovereignty concerns are relevant context here. BSC is majority-funded by the Spanish government and coordinates pan-European HPC investment under EuroHPC JU, meaning any technology integrated into that infrastructure faces implicit expectations around supply-chain transparency and European strategic autonomy. iPronics, as a European-origin company with US venture backing and a newly opened Santa Clara office, sits at the intersection of these considerations. As the EU AI Act's infrastructure-adjacent provisions and the NIS2 directive's requirements for critical digital infrastructure operators continue to phase in, the software APIs and control plane that govern dynamic optical switching in national research facilities may attract regulatory scrutiny that commercial deployments do not.

Sector

Topic

Geography