Docket #: S25-308
System for Optical Interconnects
Stanford researchers have developed a low-power optical interconnect system that can reduce the energy required to move data within and between advanced computer chips. As AI, machine learning, large language models, and data center workloads continue to scale, chips increasingly spend more power on communication than computation, creating a major barrier to energy-efficient computing. This technology uses optical communication channels to transfer information within and between chips, reducing reliance on conventional electrical interconnects and supporting more efficient integration with optical computing accelerators. The approach may enable next-generation AI and high-performance computing hardware with lower power consumption, improved scalability, and simplified connections between electronic and optical computing systems.
Stage of Development: Proof of Concept
Applications
- AI and machine learning hardware
- Large language model computing infrastructure
- High-performance computing systems
- Optical interconnects for integrated circuits
- Chip-to-chip optical communication
- Optical computing accelerators
Advantages
- Reduces power consumption from data communication in advanced chips
- Supports more energy-efficient AI and machine learning hardware
- Enables optical communication within and between integrated circuits
Related Links
Similar Technologies
-
CMOS-Compatible Single Crystal Metal Growth S15-228CMOS-Compatible Single Crystal Metal Growth
-
Integrated Laser Stabilization with Built-In Isolation S22-318Integrated Laser Stabilization with Built-In Isolation
-
Selective Area Growth of Germanium for On-chip Optical Interconnect Silicon Waveguides S12-030Selective Area Growth of Germanium for On-chip Optical Interconnect Silicon Waveguides