At last month’s Optica + APC Photonic-Enabled Cloud Computing (PECC) Summit in Silicon Valley, Daniel Pérez-López, CTO and Co-founder of iPronics, outlined how the company’s ONE Series Optical Networking Engine brings programmability and ultra-low-loss silicon photonics to AI-scale data-center interconnects.
He described a new generation of software-defined photonic fabrics that can reconfigure optical paths in microseconds—allowing AI clusters to evolve dynamically without rewiring or replacing optics.
“AI training workloads change constantly. The network itself must adapt just as quickly,” Pérez-López told attendees. “With programmable photonics, the optical layer finally becomes software-defined.”
Key Highlights from iPronics PECC Presentation
• Low-Power Architecture:
iPronics’ latest measurements show a base system power of 30 W plus 0.78 W per active optical channel, with a target below 150 W by 2027 and an ultimate goal below 100 W.
This represents an order-of-magnitude reduction compared with traditional electronic switches that include front-panel optics.
• Broadband, Low-Loss Silicon Photonics:
The company’s standard-foundry building blocks deliver state-of-the-art results:
– 3 dB splitters measured 0.060 dB IL (DR4 @1310 nm) and ≈ 0.51 SR.
– Monolayer crossings measured 0.014–0.016 dB IL with crosstalk < –50 dB.
iPronics reports 10× better performance than foundry PDKs available today, optimized for DR, LWDM and CWDM O-band operation.
• High-Speed Reconfiguration & BER Integrity:
The ONE32 engine maintains sub-millisecond reconfiguration time with less than one decade BER degradation for FR, DR and LR links—confirming stability across the full O-band.
• Programmable Photonic Core:
Based on a mesh of tunable couplers and phase shifters, the ONE architecture allows microsecond-scale path switching, optical-gain balancing, and embedded telemetry—paving the way for adaptive AI cluster fabrics.
• Scalability:
The current ONE-32 platform supports 32 ports and is designed for future 64-port / 1.6 T configurations, enabling stepwise expansion of optical connectivity without redesigning switch hardware.
Strategic Context
By combining standard silicon-photonics manufacturing with a programmable control plane, iPronics positions its technology as a software-defined optical layer for hyperscale AI networks.
The company is headquartered in Valencia, Spain, and works with international foundry partners to scale volume production of its ONE Series Optical Networking Engines.
“We’re merging the flexibility of software with the speed of light,” Pérez-López said. “That’s how you build the adaptive networks AI factories need.”
🌐 Follow the latest developments in AI photonics and optical interconnects at https://convergedigest.com/category/photonics
