U.S. Researchers Develop Breakthrough 3D Photonic-Electronic Chip for Next-Gen AI Hardware
A team of scientists from Columbia University, Cornell University, and partner institutions in the United States has successfully developed a revolutionary 3D photonic-electronic chip that merges cutting-edge photonics with advanced CMOS (complementary metal-oxide-semiconductor) technology. The innovation marks a significant leap toward more efficient data communication and processing in next-generation AI systems.
A team of scientists from Columbia University, Cornell University, and partner institutions in the United States has successfully developed a revolutionary 3D photonic-electronic chip that merges cutting-edge photonics with advanced CMOS (complementary metal-oxide-semiconductor) technology. The innovation marks a significant leap toward more efficient data communication and processing in next-generation AI systems.
The newly designed chip, measuring just 0.3 square millimeters, integrates 80 densely packed photonic transmitters and receivers. It delivers an impressive data transmission rate of 800 GB/s, while consuming only 120 femtojoules per bit—a record-breaking combination of energy efficiency and bandwidth performance. The chip also boasts an industry-leading bandwidth density of 5.3 TB/s/mm², significantly surpassing existing standards.
One of the most notable aspects of the new chip is its compatibility with current semiconductor manufacturing processes, paving the way for scalable, large-scale production.
Optical data transmission offers ultra-low energy loss and the ability to carry massive volumes of data. This advantage has already revolutionized internet infrastructure through fiber optics. Now, the integration of photonics into chip-level communication promises to overcome current limitations in distributed AI computing architectures, where latency and energy inefficiencies have long been major obstacles.
By removing bottlenecks in inter-node data transmission, this ultra-efficient, high-bandwidth photonic system opens the door for high-performance, scalable, and distributed AI hardware. This breakthrough could redefine how AI systems are designed, enabling more powerful, energy-efficient, and faster AI computing at scale.
