SAN FRANCISCO, Feb. 25 (GeokHub) — Nvidia Chief Executive Jensen Huang is preparing investors for a new competitive front in the semiconductor industry, signaling that the company’s ambitions now extend beyond its dominant graphics processing units (GPUs) into the central processing unit (CPU) market long led by rivals.
For years, Nvidia’s explosive growth has been powered by specialized GPUs that fuel artificial intelligence training in data centers worldwide. But as AI companies increasingly shift from training models to deploying them in real-world applications, CPUs are re-emerging as critical components of modern computing infrastructure.
“We love CPUs as well as GPUs,” Huang told analysts during the company’s fourth-quarter earnings call, emphasizing that Nvidia is ready to compete aggressively in the CPU space.
The Shifting Balance of AI Computing
Traditionally, CPUs — most closely associated with Intel and Advanced Micro Devices (AMD) — have served as the general-purpose “brain” of computers. GPUs, by contrast, specialize in performing large numbers of simple calculations simultaneously, making them ideal for AI model training and graphics rendering.
Huang has often noted that computing workloads have dramatically shifted in recent years, with AI-driven tasks flipping the traditional balance between CPUs and GPUs.
However, industry analysts say the next phase of AI — particularly “agentic” systems capable of autonomously performing tasks such as writing code, analyzing documents and generating reports — may rely more heavily on CPUs.
Ben Bajarin of Creative Strategies said deployment workloads are increasingly happening “more and more, and sometimes primarily, on the CPU.”
Nvidia’s current flagship AI server, the NVL72, includes 36 CPUs and 72 GPUs. Analysts suggest future systems designed for agent-based computing could see a more balanced ratio — or in some cases reduce reliance on GPUs altogether.
Nvidia Expands CPU Ambitions
Nvidia first introduced its data center CPUs in 2023 and is now positioning them as serious competitors in a market historically dominated by Intel and AMD.
Earlier this year, Nvidia announced an agreement with Meta Platforms that will see the social media giant deploy Nvidia’s Grace and Vera CPUs on a standalone basis — a notable development compared with Nvidia’s previous AI server configurations that paired each CPU with multiple GPUs.
AMD also recently disclosed a major deal with Meta involving its own CPUs, underscoring intensifying competition for hyperscale data center contracts.
On the earnings call, Huang highlighted Nvidia’s differentiated architectural strategy. Unlike some rivals that segment chips into smaller modular components, Nvidia has focused on high data throughput and memory access — features he said are crucial for AI-driven workloads.
“It is designed to be focused on very high data processing capabilities,” Huang said, emphasizing that artificial intelligence is fundamentally data-centric.
