TLDR
- Jensen Huang announced Nvidia’s intention to compete directly in the CPU space alongside Intel and AMD
- The AI industry is transitioning from GPU-intensive model training toward CPU-optimized agent deployment
- Grace and Vera processors, introduced in 2023, focus on high-throughput data processing for data center environments
- Meta Platforms agreed to purchase Nvidia CPUs independently, without bundling them with GPU hardware
- AMD secured its own standalone CPU agreement with Meta, intensifying competition in the processor market
Nvidia built its empire selling GPUs. CEO Jensen Huang now signals a strategic expansion into the CPU domain, where Intel and AMD have historically reigned supreme.
During Nvidia’s fourth-quarter earnings call on Wednesday, Huang conveyed to analysts that the company stands prepared to capitalize on the resurgence of CPU-centric computing.
“We love CPUs as well as GPUs,” Huang stated during the analyst call.
CPUs powered the majority of computing operations for many years. The explosion of AI model training shifted considerable workload to GPUs, which excel at the parallel processing demands these tasks require.
The pendulum swings once more. AI organizations are progressing from the training phase to deployment, running what industry experts call “agents” — automated systems that generate code, analyze documents, and produce reports.
Creative Strategies analyst Ben Bajarin explains that these agent-based workloads favor CPU architecture. He noted that agentic computing “is happening more and more, and sometimes primarily, on the CPU.”
The CPU vs GPU Shift in AI
Nvidia’s premier AI server configuration, the NVL72, currently features 36 CPUs paired with 72 GPUs. Bajarin forecasts this ratio could evolve toward parity — or certain applications might rely exclusively on CPUs.
The company introduced its Grace and Vera CPU products for data center applications in 2023. Huang emphasized these processors follow a distinct design philosophy compared to Intel and AMD offerings, prioritizing data throughput over general-purpose versatility.
“It is designed to be focused on very high data processing capabilities,” Huang explained.
Nvidia disclosed an agreement with Meta Platforms for substantial CPU volume shipments as standalone products — representing a departure from the traditional GPU-bundled sales approach.
Meta continues working with its established CPU vendors. The company views Nvidia as an additional supplier rather than a replacement. Shortly following the Nvidia announcement, AMD revealed its own standalone CPU contract with Meta.
Intel’s Dominance No Longer Guaranteed
Dave Altavilla, analyst at HotTech Vision and Analysis, believes Nvidia aims to demonstrate that CPUs “is no longer the assumed default foundation of modern compute infrastructure.”
During his January CES presentation, Huang predicted Nvidia CPU adoption in data centers would “explode,” expressing confidence that Nvidia could emerge as “one of the largest CPU makers in the world.”
Nvidia will provide additional information about its CPU development timeline at its annual developer conference in Silicon Valley next month.
