AMD, OpenAI Partner on AI Networking With MRC
AMD and OpenAI announced a collaboration to advance AI networking infrastructure at scale using MRC technology, according to a press release from AMD published Tuesday.
The partnership between the Santa Clara, California-based chipmaker and San Francisco-based AI lab expands hardware ties between the two companies as demand for large-scale AI training infrastructure grows.
While specific technical details of the MRC protocol were not immediately available, the initiative targets the networking layer that connects thousands of accelerators in AI data centers — a bottleneck in training frontier AI models.
The announcement has implications for the competitive landscape in AI infrastructure. Nvidia currently dominates AI networking through its acquisition of Mellanox Technologies and its InfiniBand and Spectrum-X networking products, which are widely deployed in major AI training clusters worldwide.
OpenAI is among the largest consumers of AI compute and has worked with multiple hardware providers to build out data center infrastructure. AMD has reported growth in its data center segment in recent quarters.
For AMD, the partnership adds a networking component to the company’s AI infrastructure portfolio, complementing its Instinct MI series accelerators rather than competing on chips alone.
The collaboration comes as hyperscalers and AI labs have pursued open and interoperable AI infrastructure standards.
Both AMD and OpenAI are U.S.-headquartered companies, and their partnership touches on the domestic AI infrastructure supply chain that has become a focus of federal policy discussions around maintaining American competitiveness in artificial intelligence.
