Ayar Labs announced that it has joined the NVIDIA NVLink Fusion ecosystem, positioning its co-packaged optics (CPO) technology as a connectivity option for future rack-scale AI systems built around NVIDIA’s expanding heterogeneous computing architecture. The move aligns Ayar Labs’ optical interconnect technology with NVIDIA’s optical and SerDes technologies, enabling hyperscalers and system developers to integrate optical connectivity into NVLink Fusion-based AI infrastructure.
The announcement addresses one of the most significant challenges facing next-generation AI clusters: moving massive amounts of data efficiently as GPU counts scale higher and systems become increasingly heterogeneous. Traditional copper interconnects face growing constraints in bandwidth density, reach, latency, and power consumption. Ayar Labs aims to overcome these limitations by integrating optical connectivity directly alongside compute silicon, enabling higher bandwidth and longer reach while reducing power consumption.
Under the collaboration, Ayar Labs will work with customers and ecosystem partners to support deployment of CPO within NVLink Fusion architectures. NVIDIA’s NVLink Fusion platform enables integration of custom CPUs and accelerators into NVIDIA-based rack-scale systems while preserving compatibility with the broader NVLink ecosystem. The addition of optical connectivity provides another option for architects designing large-scale AI factories that span multiple racks and increasingly complex compute topologies.
- Ayar Labs will make its optical interconnect products electrically and optically compatible with NVIDIA optical and SerDes technologies.
- The collaboration targets hyperscalers and system innovators building heterogeneous AI infrastructure.
- Co-packaged optics seeks to address bandwidth, latency, power, and distance limitations associated with conventional electrical interconnects.
- NVLink Fusion enables customers to integrate custom CPUs and XPUs into NVIDIA rack-scale architectures.
- Ayar Labs said its CPO technology supports higher bandwidth density and lower power consumption for AI scale-up applications.
- The announcement follows Ayar Labs’ recent $500 million Series E financing round, which included participation from NVIDIA.
- Ayar Labs was founded in 2015 and develops optical I/O technology for AI and high-performance computing systems.
“By joining the NVIDIA NVLink Fusion ecosystem, we’re introducing co-packaged optics as a foundational building block for customers deploying heterogeneous compute in NVIDIA AI factories,” said Mark Wade, CEO of Ayar Labs.
🌐 Analysis: This announcement represents a notable milestone for co-packaged optics as the technology moves closer to mainstream deployment in AI infrastructure. While CPO has been discussed for years as a potential successor to power-hungry pluggable optics and copper interconnects, AI training and inference clusters are now reaching bandwidth and power levels where optical integration is becoming increasingly attractive. NVIDIA’s decision to include Ayar Labs within the NVLink Fusion ecosystem signals growing industry recognition that future rack-scale systems may require optical connectivity inside the rack, not just between racks.
