Ayar Labs, a leader in co packaged optical (CPO) solutions, announced its joining of the NVIDIA NVLink Fusion ecosystem and making its products optically and SerDes compatible with NVIDIA. This move enables large-scale customers and system innovators to build AI infrastructure based on optical connectivity around the NVIDIA NVLink Fusion platform and its partner ecosystem.
Data mobility and power consumption become key bottlenecks
As AI factories move towards larger GPU clusters and increasingly heterogeneous architectures, data mobility and power consumption are becoming key bottlenecks. System designers must rethink the connection methods between computing units, how bandwidth expands with distance, and how power is distributed throughout the entire rack.
Ayar Labs’ CPO solution aims to address scalability limitations by bringing high bandwidth, low latency, and energy-efficient optical connections to the NVLink Fusion architecture, thereby helping customers expand design space as bandwidth demands grow and electrical constraints tighten. By deploying optical technology in critical locations, Ayar Labs’ CPO supports flexible system architecture and efficient AI scaling beyond copper wire limitations.
Both parties’ executives discuss the significance of cooperation
Ayar Labs CEO Mark Wade said, “AI infrastructure is being co designed from scratch, and as bandwidth continues to rise, customers need more options to efficiently scale performance. By joining the NVIDIA NVLink Fusion ecosystem, we are using co packaged optical technology as the foundational building block for customers to deploy heterogeneous computing in NVIDIA AI factories.”
NVIDIA Vice President Ashish Karandikar said, “The combination of NVLink Fusion and Ayar Labs’ CPO technology provides customers with more options for building heterogeneous AI factories. By extending the NVLink Fusion ecosystem with Ayar Labs’ optical connectivity, innovators can expand bandwidth and bring heterogeneous AI infrastructure to market faster.”
Assist customers in deploying heterogeneous computing
This collaboration aims to help customers deploy heterogeneous computing (including customized silicon wafers) within NVIDIA rack level platforms while retaining investments based on NVLink architecture. NVLink Fusion provides a path to integrate custom CPUs and XPUs into NVIDIA rack level architectures and ecosystems, enabling customers to move faster from system concepts to large-scale deployments.
Ayar Labs’ CPO solution complements NVIDIA’s broader AI factory technology stack by combining NVLink Fusion with NVIDIA’s end-to-end network platform. These technologies collectively support the expansion of next-generation AI factories within and across racks, where bandwidth density, latency, and energy efficiency directly affect utilization and total cost of ownership.
Ayar Labs will collaborate with customers and ecosystem partners to coordinate matters related to CPO integration and NVLink Fusion deployment, including system architecture, validation requirements, and platform timelines.
Continuing the momentum of $500 million financing
Today’s announcement is based on Ayar Labs’ successful completion of a $500 million Series E funding round, which includes investors including NVIDIA.
About Ayar Labs
Ayar Labs is revolutionizing AI infrastructure by partnering with the world’s leading semiconductor ecosystem to manufacture the industry’s first validated co packaged optical solution. By unleashing performance improvements and reducing workload costs in power constrained environments, Ayar Labs’ light engine is key to supporting next-generation AI scaling. The company was founded in 2015 and is supported by domestic and foreign venture capital firms, as well as strategic investors including AMD, Applied Ventures, MediaTek, NVIDIA, and VentureTech Alliance. For more information, please visit www.ayarlabs.com.
