After GPUs, Jensen Huang Aims to Build a "CPU Empire"

After dominating the AI chip market with GPUs, NVIDIA is now setting its expansion sights on a previously untapped territory: the CPU market.

NVIDIA Chief Financial Officer Colette Kress stated during Wednesday's earnings call that the company's first standalone AI-focused CPU product, Vera, has opened up a "brand-new" potential market worth up to $200 billion. She added that NVIDIA already has visibility into nearly $20 billion in CPU revenue for this year, with Vera "propelling us to become a leading global CPU supplier."

This statement quickly drew attention from Wall Street. Multiple analysts pointed out that if the projected $20 billion volume materializes, it will place NVIDIA directly among the leaders in the CPU market and surpass AMD's revenue scale in the cloud CPU sector. At the same time, it also means NVIDIA's reach will extend into a market long dominated by Intel, AMD, and Arm.

For competitors like Arm and Qualcomm that rely on standalone CPU architectures, NVIDIA's entry could squeeze already tight production capacity and deliver a substantial impact.

Vera Unlocks a Brand-New $200 Billion Market

Vera was officially launched this March as NVIDIA's first standalone CPU product in history, designed specifically for AI inference and agentic workloads, and built on the Arm architecture. During the earnings call, Kress stated that hyperscale cloud computing clients and AI system vendors have already begun deploying the chip, noting that CPUs represent a market segment NVIDIA has "never previously ventured into."

According to Kress's disclosure, NVIDIA already has visibility into nearly $20 billion in CPU revenue for this year. If this figure proves accurate, it will surpass UBS analyst Timothy Arcuri's forecast of approximately $16 billion in AMD's CPU revenue for this year—while AMD currently holds over a 50% share of the cloud CPU market.

In a Thursday research note, Bernstein analyst Stacy Rasgon wrote that NVIDIA "appears poised to become the king of the CPU space, with a business scale that could match or even surpass more traditional competitors in this market."

Wall Street: The $20 Billion Figure Requires Interpretation

Analysts generally agree that the $20 billion figure leaves room for interpretation and should not be directly benchmarked against traditional CPU revenue metrics.

Arcuri noted that Vera can be sold either as a standalone chip paired with non-NVIDIA AI chips, or as an integrated system solution that includes CPU and memory components. Because expensive memory chips are also bundled into the latter package, the $20 billion projection is not "as straightforward as it appears." He estimates that Vera accounts for about 30% of the value in a complete system, and sets NVIDIA's actual CPU revenue forecast for this year in the range of $5 billion to $7 billion.

Rasgon suggested that the high $20 billion figure might stem from NVIDIA's pricing model of selling CPUs by the rack rather than per individual chip. Morgan Stanley analyst Joseph Moore shared a similar view, stating that the figure "may include CPUs shipped alongside GPU cards, which counts but is relatively easier to achieve." Nevertheless, Moore indicated that the $20 billion target places NVIDIA "at the tipping point of market leadership," while his revenue modeling for Intel's data center business (including some networking products) remains slightly higher than NVIDIA's.

Inference Wave Reignites Demand for CPUs

NVIDIA's entry into the CPU market comes at a time when AI applications are rapidly shifting from model training to inference and the agentic AI paradigm. Unlike the training phase, which relies heavily on GPUs, the inference stage is seeing a significant resurgence in demand for server CPUs.

This trend has already sparked signs of recovery for Intel, which was previously mired in difficulties. Arm Holdings also announced earlier this year the launch of its first-ever self-developed CPU chip, specifically tailored for inference and agentic AI workloads.

Moore stated that the Arm-architecture CPUs NVIDIA is entering compete on different dimensions compared to Intel's x86-architecture CPUs, meaning they "will not bring additional negative impacts" to x86-based chips.

In the competitive landscape outside of NVIDIA, Arcuri warned that NVIDIA's large-scale push into the CPU space will further consume already tight wafer and memory capacity, "posing negative impacts" on the market expansion of companies like Arm and Qualcomm.

Arcuri specifically pointed out that Arm's collaborative relationships at the wafer foundry and memory supplier levels are relatively less mature than NVIDIA's, meaning it may face a more direct impact. Meanwhile, Moore believes that the supply chain bargaining power and procurement priority NVIDIA has built up over the years will keep it highly competitive even in a capacity-constrained environment.