Meta expands its agreement with Nvidia to implement millions of chips in its AI data centers.

Meta has announced an extensive partnership with Nvidia, which will involve the deployment of millions of chips in its data centers dedicated to artificial intelligence. This agreement includes the Grace CPU processors and the next-generation Vera Rubin systems, with the contract value estimated to reach tens of billions of dollars. Meta intends to invest up to 135 billion dollars in AI infrastructure by 2026, as part of a broader commitment of 600 billion dollars for infrastructure in the US by 2028. Among the highlights of this agreement is the large-scale deployment of Grace CPU processors, which will be used as standalone processors, and the provision of a roadmap for next-generation GPUs. Additionally, Meta will integrate Nvidia's Spectrum-X networking technology and will expand the partnership into consumer AI applications, such as WhatsApp. Meta plans to build 30 data centers, most of them in the US, with two of the largest facilities already under construction. Although it collaborates closely with Nvidia, Meta is also exploring alternatives such as Google's Tensor processing units and the development of in-house silicon. This collaboration underscores emerging trends in AI, including complete AI infrastructure, hardware-software co-design, and supply chain control as a competitive advantage.