Tower Semiconductor and Salience Labs Limited announced a partnership to jointly manufacture photonic integrated circuit-based optical circuit switches for artificial intelligence infrastructure. This collaboration leverages Tower's silicon photonics platforms, including the PH18DA with integrated III-V lasers and the TPS45PH with low-loss nitride waveguides, to advance products from the development phase into pre-production, supporting large-scale deployment in AI data centers.
Artificial intelligence workloads are driving growth in data center scale and network complexity, increasing the demand for higher bandwidth, lower latency, and more energy-efficient optical interconnects. Optical circuit switching architectures offer an alternative to current electronic packet-switching architectures based on OEO conversion by moving more connections and switching into the optical domain. According to Dell'Oro Group, data center switch spending for AI backend networks is projected to exceed $100 billion by 2030.
Vaysh Kewada, Founder, CEO, and Director of Salience Labs, stated: "Tower is a key partner for Salience Labs, supporting our roadmap through their silicon photonics and switching technology platforms. Our collaboration is built on shared expertise in silicon photonics and specialized platforms, enhancing our ability to deliver optical switching technologies optimized for the performance and power consumption demands of AI data centers."
Dr. Ed Preisler, Vice President and General Manager of the RF Business Unit at Tower Semiconductor, said: "Silicon photonics with integrated light sources is a key enabler for scaling next-generation optical connectivity, and our partnership with Salience Labs strengthens progress in the AI and infrastructure sectors. Salience Labs brings an optical circuit switching approach to AI infrastructure, and we are excited to support their scaling. By combining our silicon photonics and specialized process platform capabilities, customers can industrialize silicon photonics-based switching architectures while ensuring a path from development to high-volume manufacturing."
