Microsoft, Meta, and OpenAI Collaborate with AMD and Nvidia on High-Speed Optical Interconnect for AI Megaclusters

• March 13, 2026

Leading cloud service providers Microsoft, Meta, and OpenAI have entered into a collaborative partnership with semiconductor and hardware companies AMD, Broadcom, and Nvidia. This joint effort is focused on developing a high-speed, protocol-independent optical interconnect designed specifically for AI megaclusters.

The initiative aims to address the growing demand for scalable and efficient communication infrastructure capable of supporting the immense data throughput and low latency required by next-generation artificial intelligence workloads.

Establishing a Universal Fiber Optic Infrastructure for AI Clusters

The collaboration has formed a Multi-Source Agreement (MSA), a framework that will define an open and standardized approach to building short-distance, high-bandwidth interconnections within AI compute clusters. This MSA is intended to complement existing technologies by providing a flexible and interoperable solution that is not dependent on a single proprietary protocol.

By focusing on protocol-agnostic compatibility, the developers seek to enable more straightforward scalability and integration across various hardware platforms and data center environments. This enhances performance and reduces complexity in the design and deployment of AI megaclusters, which consist of thousands of interconnected GPUs and compute nodes.

This development reflects a broader trend in the tech industry where cloud providers and hardware manufacturers collaborate to optimize infrastructure for the computational demands of artificial intelligence. The resulting optical interconnect technology is expected to significantly improve data transfer speeds while minimizing latency, a critical factor in training and deploying large AI models.

While specific technical details, timelines, and commercial availability have not been disclosed, the partnership signals a coordinated effort to accelerate innovation in AI infrastructure, enabling cloud providers and AI organizations to manage more extensive and complex workloads efficiently.

As AI continues to evolve rapidly, advancements in interconnect technology remain a crucial component supporting the performance and scalability of data centers that host large-scale machine learning applications.

Tech giants and hardware developers unite to create a scalable, protocol-agnostic optical interconnect for AI megaclusters.