Nvidia’s Ambition Goes Beyond GPUs, What Secrets Are Hidden in Jensen Huang’s CPU Grand Strategy?

TradingKey - NVIDIA ( NVDA ), has once again stunned Wall Street with results that exceeded expectations. For the quarter ending in January, which marks the start of its fiscal year cycle, revenue reached $68.1 billion, a 73% year-over-year surge that far exceeded the market consensus of $66 billion; within this, Data Center revenue grew 75% year-over-year, and gross margin reached a staggering 75%, demonstrating that its pricing power in the AI compute market remains rock-solid.

As the GPU dominant player holding approximately 90% of the global AI processor market, NVIDIA's ambitions clearly extend beyond this—the company is accelerating its layout in the AI and consumer CPU markets, seeking to replicate its success in the GPU space.

NVIDIA Positions Itself Early

Since late 2025, the global CPU market has seen a surge in interest. Intel ( INTC) and AMD ( AMD) have seen their server CPU capacity sell out, with lead times stretching to six months, as a supply-demand imbalance begins to emerge. Industry insiders generally believe this marks a return of CPU value in the AI era.

Previously, in AI computing, GPUs held absolute dominance due to their parallel processing capabilities, while CPUs only handled basic general-purpose tasks. However, with the proliferation of generative AI and multimodal models, AI computing is shifting from 'training-centric' to 'equal emphasis on training and inference.' Especially as we enter the era of AI agents, reliance on CPUs for task scheduling and tool calling has increased significantly.

During the pre-training phase of large models, CPUs are responsible for data storage, sharding, and indexing, providing support for the GPU's core operations. In multimodal inference scenarios, CPUs handle image and video decoding, alleviating the computing pressure on GPUs.

As AI penetrates edge and terminal devices in the future, a single chip will not be able to cover all scenario requirements. Full-stack heterogeneous solutions that coordinate GPUs and CPUs will become the standard.

Recognizing this trend, NVIDIA launched its self-developed data center CPU products as early as 2023. According to reports, NVIDIA CEO Jensen Huang stated during an earnings call that as AI companies move from model training to the deployment phase, NVIDIA will refocus on CPUs. He noted that their proprietary CPUs are highly competitive and have the potential to make NVIDIA a major global CPU manufacturer in the future.

Currently, NVIDIA has introduced several ARM-based processors in the AI market, and its N1X product for the consumer market is about to be released.

NVIDIA's Full-Stack Ecosystem Anchors Its CPU Confidence

The confidence behind NVIDIA's CPUs stems from its deep-rooted full-stack ecosystem capabilities. In the AI inference era, the advantages of CPU branch logic processing are amplified. While hybrid 'NVIDIA GPU + AMD CPU' solutions once appeared on the market, NVIDIA has achieved deep synergy between CPUs and GPUs through its proprietary NVLink high-speed interconnect technology, delivering system-level performance gains that far exceed the advantages of a standalone CPU.

More importantly, NVIDIA owns the CUDA ecosystem, upon which the vast majority of the world's AI development frameworks and models are built. This ecosystem 'lock-in effect' makes it difficult for customers to migrate to other platforms. Currently, NVIDIA's networking revenue accounts for approximately 15% of total Data Center revenue, with a year-over-year growth rate of 162%, further consolidating the competitiveness of its full-stack computing system.

Meta ( META) and NVIDIA's multi-year contract last week marks a major milestone for the commercialization of NVIDIA's CPUs. Meta will not only purchase millions of Blackwell and Rubin GPUs but will also adopt NVIDIA's Grace CPUs as standalone server chips. This marks the first large-scale deployment of this CPU model and validates NVIDIA's technical prowess in the CPU domain.

NVIDIA's Long-Termism

Many attribute NVIDIA's success to 'catching the AI wave,' but in fact, the company's positioning began over 20 years ago.

In the 1990s, NVIDIA established its dominance in the PC gaming hardware market with the GeForce series, accumulating core technology in GPU research and development. The launch of the CUDA parallel computing architecture in 2006 transformed the GPU from a game rendering tool into a general-purpose computing chip, laying the groundwork for the explosion of the AI era.

Today, NVIDIA's layout in the CPU sector remains a continuation of this long-term strategy. Leveraging the technical and ecosystem advantages accumulated in its GPU business, NVIDIA is building a full-stack computing system where GPUs and CPUs collaborate. This may well become the core barrier that maintains its lead over the next decade.