Nvidia built its recent dominance on specialized graphics processors that power artificial intelligence systems, but chief executive Jensen Huang is now emphasizing the growing importance of central processing units as competition in data centers heats up.
For decades, CPUs were seen as the core of computing, a market long led by Intel and later challenged by Advanced Micro Devices. Huang has often noted that while most computing once relied on CPUs, recent years saw workloads shift heavily toward GPUs.
That balance, however, is beginning to change again. As artificial intelligence companies move from training models to deploying them at scale, CPUs are increasingly viewed as equally important — and sometimes better suited — for certain tasks. Nvidia says it intends to play a major role in that transition.
“We love CPUs as well as GPUs,” Huang told analysts during the company’s latest earnings call, adding that Nvidia’s data-center CPUs, first introduced in 2023, are designed to outperform existing alternatives.
Huang recently said he expects adoption of Nvidia’s high-performance CPUs in data centers to accelerate rapidly, suggesting the company could eventually rank among the world’s largest CPU suppliers.
Traditionally, CPUs and GPUs have handled different types of computing. CPUs are general-purpose chips capable of managing a wide range of tasks, while GPUs excel at performing simpler calculations simultaneously at massive scale — a strength that made them essential for graphics and AI workloads.
As AI systems increasingly rely on autonomous software agents that write code, analyze documents, and generate reports, more processing is shifting back to CPUs, according to industry analysts. Some believe future systems may rely on fewer GPUs or, in certain cases, bypass them altogether.
Nvidia’s flagship AI server currently combines dozens of its CPUs with a larger number of GPUs, though analysts say future configurations could become more balanced depending on workload demands.
Highlighting its ambitions, Nvidia recently announced that Meta Platforms will use its CPUs independently of GPUs in large volumes — a notable shift from earlier server designs. Meta has also secured CPU supply agreements with AMD, reflecting a broader diversification strategy rather than a switch away from existing partners.
Huang argues that Nvidia’s CPU architecture differs fundamentally from rivals, emphasizing high data throughput and memory access over chip-splitting techniques used elsewhere. He said the design is tailored for data-driven workloads, with artificial intelligence at the center.
Industry analysts say Nvidia is attempting to challenge the long-held assumption that CPUs from Intel form the default foundation of modern computing, positioning them instead as one option among several competing architectures.
Huang said Nvidia will share more details about its CPU roadmap at the company’s annual developer conference in Silicon Valley next month.
