Meta signs multigenerational AI infrastructure pact with Nvidia

Meta is deepening its reliance on Nvidia with a long-term agreement that extends beyond graphics processors. The arrangement centers on deploying millions of Nvidia Blackwell and next-gen Rubin GPUs across hyperscale data centers for both training and inference workloads. Meta is also standardizing additional pieces of its stack, including Nvidia CPUs and high-throughput networking, to reduce friction in the design and operation of its AI clusters. The company views the shift from occasional model training to constant inference serving as a reason to favor an integrated compute-and-network fabric at scale. The agreement signals that Nvidia's platform strategy is working, according to Business Insider. Nvidia is becoming the default supplier not only for accelerators but also for other elements inside the data center.