TLDR
- Nvidia CEO Jensen Huang told analysts Nvidia is ready to compete with Intel and AMD in the CPU market
- AI workloads are shifting from model training (GPU-heavy) to deploying “agents” (CPU-friendly)
- Nvidia’s Grace and Vera CPUs, launched in 2023, are designed for high data processing in data centers
- Meta Platforms signed a deal to use Nvidia CPUs on a standalone basis, without pairing them with GPUs
- AMD also announced a separate CPU deal with Meta, showing competition is heating up
Nvidia has long made its money from GPUs. But CEO Jensen Huang is now making a clear push into CPUs, the chips traditionally dominated by Intel and AMD.
Huang told analysts on Wednesday, during Nvidia’s fourth-quarter earnings call, that the company was not just ready for the CPU’s comeback — it planned to lead it.
“We love CPUs as well as GPUs,” Huang said on the call.
For decades, CPUs handled most computing tasks. GPUs then took over a large share of that work as AI model training exploded, requiring the kind of parallel processing GPUs do best.
But that balance is shifting again. AI companies are now moving from training their models to running them — and deploying so-called “agents” that write code, read documents, and generate reports.
That kind of work runs better on CPUs, according to analyst Ben Bajarin of Creative Strategies. He said agentic computing “is happening more and more, and sometimes primarily, on the CPU.”
The CPU vs GPU Shift in AI
Nvidia’s current flagship AI server, the NVL72, holds 36 CPUs and 72 GPUs. Bajarin suggested that ratio could move toward one-to-one — or that GPUs could be skipped entirely for some tasks.
Nvidia first released its Grace and Vera CPU chips for data centers in 2023. Huang said these chips are built differently from Intel and AMD products, focused on high data throughput rather than general-purpose flexibility.
“It is designed to be focused on very high data processing capabilities,” Huang said.
Nvidia recently announced a deal with Meta Platforms to supply large volumes of its CPUs on a standalone basis — meaning without the usual GPU pairing. That marks a shift from how Nvidia chips have traditionally been sold.
Meta is not switching away from its existing CPU suppliers. The company is simply adding Nvidia as another source. Days after the Nvidia announcement, AMD also confirmed a new CPU deal with Meta.
Intel’s Dominance No Longer Guaranteed
Analyst Dave Altavilla of HotTech Vision and Analysis said Nvidia is trying to show that the CPU “is no longer the assumed default foundation of modern compute infrastructure.”
Huang also said in January at CES that Nvidia CPU use in data centers would “explode,” and that he would not be surprised if Nvidia became “one of the largest CPU makers in the world.”
Nvidia plans to share more details about its CPU roadmap at its annual developer conference in Silicon Valley next month.





