Amid rising demand for faster and more energy-efficient data movement in AI-driven data centers, HyperLight Corporation has introduced a new family of 400G-per-lane photonic integrated circuits (PICs) built on its TFLN Chiplet Platform.
The newly developed devices leverage thin-film lithium niobate (TFLN) technology to support next-generation optical interconnects, a critical component in scaling AI infrastructure. As enterprises and hyperscalers push for higher bandwidth and system density, the shift toward 400G-per-lane connectivity is becoming essential to meet performance requirements.
HyperLight’s PICs are designed to overcome the limitations of conventional electronic integrated circuits, particularly in managing high data rates and power consumption. The devices deliver low insertion loss, reduced drive voltage, and high electro-optic bandwidth, enabling efficient and high-performance optical links suited for large-scale AI deployments.
Manufactured using the company’s TFLN Chiplet Platform, the PICs combine high modulation efficiency with extremely low optical loss. This architecture supports flexible transmitter designs, including single- and dual-laser configurations, while also improving scalability in production.
Mian Zhang, Chief Executive Officer of HyperLight, noted that the transition to 400G-per-lane highlights the advantages of TFLN technology, particularly in maintaining bandwidth performance while reducing power requirements and improving manufacturability.
Industry collaboration is also shaping the ecosystem around these advancements. Broadcom indicated that integrating HyperLight’s high-bandwidth PICs with its Taurus DSP platform enhances signal integrity and energy efficiency in optical interconnects. Meanwhile, Eoptolink emphasized that the solution can reduce dependence on external drivers, lower laser counts, and simplify module integration, contributing to improved cost efficiency and reliability.
