Nvidia’s Deal With Meta Signals a New Era in Computing Power (www.wired.com)

Nvidia has announced a significant multiyear deal with Meta, where the social media giant will purchase billions of dollars' worth of Nvidia chips, including its newly promoted CPUs. This collaboration aims to enhance Meta's AI infrastructure by building hyperscale data centers optimized for both training and inference. Notably, this marks Meta as the first major tech firm to make a large-scale acquisition of Nvidia's Grace CPUs, reflecting a growing shift towards integrating CPU architectures in AI workflows, as agentic AI demands more general-purpose computing power. This development is crucial for the AI/ML community as it underscores the evolving nature of AI workloads, which now require both GPUs and CPUs. Analysts suggest that there is an escalating need for CPU usage in data centers, driven by complex AI processing requirements. As companies like Meta anticipate spending dramatically on AI infrastructure—between $115 billion and $135 billion this year—Nvidia is positioning itself to adapt to these changing demands amid increasing competition from custom chip initiatives by other tech giants such as Microsoft and Google. The partnership not only strengthens Nvidia's market position but also signals a potential shift in how AI computation is approached in the industry, balancing the need for traditional GPU-heavy processing with efficiency-driven CPU utilization.