Astera Labs expands connectivity offerings with custom solutions for AI infrastructure
Astera Labs, Inc. introduced plans to provide custom connectivity solutions focused on supporting next-generation Artificial Intelligence (AI) infrastructure that integrates heterogeneous computing resources. This development aims to address the growing complexity and varied requirements within AI system architectures.
The company identified a requirement among hyperscalers for connectivity solutions tailored to specific architectural configurations and workload performance objectives. The custom solutions are intended as a supplemental component to Astera Labs' existing Intelligent Connectivity Platform, enabling system integrators to assemble diverse accelerators in unified, rack-scale platforms.
The custom offerings will utilize the company's experience with connectivity technologies and the COSMOS software-defined architecture, incorporating recent advancements such as photonic chiplet technology. These capabilities are designed to expand the options available for deploying standard, custom, and hybrid AI infrastructures.
This initiative builds on Astera Labs’ earlier announcement of joining the NVIDIA NVLink Fusion Ecosystem, emphasizing collaboration with hyperscalers to develop NVLink-supporting solutions. With NVLink Fusion integration, the custom connectivity solutions aim to deliver multiple terabytes per second of low-latency data throughput.
“Hyperscalers are accelerating the deployment of next-generation AI infrastructure and require proven connectivity partners who can consistently deliver at cloud scale,” said Sanjay Gajendra, President and COO at Astera Labs. Ashish Karandikar, Vice President of Engineering at NVIDIA, said, “NVLink Fusion, combined with Astera Labs’ purpose-built connectivity solutions, gives customers access to the industry’s best scale-up interconnect with NVLink in a proven rack-scale architecture.”
Astera Labs outlined its focus on collaborating with partners to develop customized architectures suited to specific infrastructure needs. The company described these plans as additive to its existing portfolio, aiming to enhance the breadth of connectivity solutions for heterogeneous AI deployments.