Astera Labs, a leader in semiconductor-based connectivity for AI and cloud infrastructure, has announced a collaboration with NVIDIA to advance the NVIDIA NVLink Fusion ecosystem. This partnership aims to provide scale-up connectivity solutions critical for supporting next-generation AI workloads featuring trillions of parameters and hyperscale compute demands.
Accelerating AI Infrastructure with NVLink Fusion
Sanjay Gajendra, President and COO of Astera Labs, emphasized the importance of adopting scale-up platforms that deliver both high performance and efficiency to meet rapidly growing AI model sizes. By integrating NVLink solutions into its Intelligent Connectivity Platform, Astera Labs will enhance the options and speed to market for hyperscaler customers while expanding its reach in the AI infrastructure market.
Next-generation AI workloads require ultra-high bandwidth, low latency connectivity, and native memory semantic support to synchronize hundreds of accelerators at rack-scale. Astera Labs’ NVLink solutions will extend its platform, which already integrates PCIe®, CXL®, and Ethernet silicon with its COSMOS software suite for advanced data center visibility and optimized system performance.
Deep Collaboration Spanning Multiple NVIDIA Platforms
Astera Labs has a proven track record with NVIDIA across several platform generations, including HGX, MGX, and NVL72. Its Aries PCIe/CXL Smart DSP Retimers have been widely deployed on NVIDIA Hopper and HGX platforms and in custom rack-level solutions by hyperscalers.
Recently, Astera Labs demonstrated industry-first end-to-end PCIe 6 interoperability using Scorpio P-Series Fabric Switches, Aries 6 Retimers, and an NVIDIA Blackwell GPU at NVIDIA GTC 2025. These innovations underline the company’s leadership in delivering interoperable, scalable connectivity solutions for AI data centers.
Industry Impact and Future Outlook
Ashish Karandikar, NVIDIA’s Vice President of SoC Products, highlighted that collaborations like this empower the NVLink Fusion ecosystem to build high-performance AI “factories” capable of rapidly scaling to meet the needs of complex model training and agentic AI inference.
By integrating NVLink solutions into its Intelligent Connectivity Platform, Astera Labs is poised to deliver the scalable, efficient connectivity that hyperscalers require for heterogeneous AI compute infrastructure combining NVIDIA GPUs and custom accelerators
This collaboration marks a key step toward meeting the evolving connectivity challenges posed by rapidly expanding AI workloads. Astera Labs and NVIDIA continue to push the boundaries of high-performance AI infrastructure, enabling the next generation of scalable, efficient AI compute platforms.
