Nvidia's GTC will mark an AI chip pivot. Here's why the CPU is taking center stage

Nvidia, a leader in graphics processing units (GPUs), is gearing up for a significant revelation at its annual GTC conference, set to begin on Monday. The spotlight this year will be on the often-overlooked central processing unit (CPU), which has gained newfound relevance due to the rise of agentic artificial intelligence. As AI evolves, the demand for robust CPUs has surged, and Nvidia’s head of AI infrastructure, Dion Harris, believes this is an "exciting opportunity" for the company. The tech giant is anticipated to showcase its latest agentic-optimized CPUs, with a dedicated CPU-only rack expected to be featured prominently on the showroom floor. Harris notes that CPUs are becoming a bottleneck in expanding AI capabilities, emphasizing the need for enhanced computational power to support complex workflows. This shift comes as Nvidia prepares to roll out its next-generation CPU, Vera, following the earlier introduction of its first data center CPU, Grace, in 2021. Nvidia's rise to prominence is underscored by its staggering $4.4 trillion market cap, largely fueled by the soaring demand for GPUs. However, the company's strategy has evolved significantly, particularly after securing a multiyear agreement with Meta, which includes the large-scale deployment of Grace CPUs independently, with plans for Vera to follow in 2027. Furthermore, standalone Nvidia CPUs are currently aiding supercomputers at institutions like the Texas Advanced Computing Center. Market analysts predict that the CPU sector could experience explosive growth, potentially doubling from $27 billion in 2025 to $60 billion by 2030. Nvidia's recent quarterly report highlights a staggering 75% increase in data center revenue year-over-year, totaling over $62 billion. This surge is attributed to a fundamental shift in computing requirements as agentic AI transitions from simple response systems to more complex task-oriented applications. While GPUs excel in training AI models due to their numerous cores designed for parallel processing, CPUs, with their powerful cores, are essential for managing general-purpose tasks and orchestrating data flow across agents. Nvidia's CEO, Jensen Huang, has highlighted the exponential growth in token generation, necessitating faster inference capabilities to keep up with demand. As Nvidia's CPUs take center stage, the broader CPU market is facing what some experts describe as a "quiet supply crisis." Leading manufacturers like AMD and Intel have alerted customers about potential supply shortages, with delivery lead times stretching up to six months and prices rising over 10%. AMD's Forrest Norrod acknowledges unprecedented demand over the past several months and reassures that the company is proactively addressing these challenges. Nvidia's approach to CPU design sets it apart from competitors. While traditional CPUs like AMD's EPYC and Intel's Xeon focus on maximizing core counts, Nvidia prioritizes optimizing performance to complement its GPUs in AI workloads. Unlike its rivals, Nvidia’s CPUs are based on Arm architecture, typically associated with mobile devices, rather than the x86 architecture that dominates the market. In a landscape where hyperscalers are increasingly developing their own Arm-based processors, Nvidia's entry into standalone CPUs reflects a broader trend in the industry. Major players like Amazon and Microsoft are also venturing into in-house CPU production, highlighting the shift towards custom silicon solutions. As Nvidia continues to innovate and adapt its strategy, it maintains a platform-agnostic approach, supporting a variety of architectures, including RISC-V. This flexibility positions Nvidia to thrive amid the evolving demands of the compute landscape, ensuring its offerings remain relevant across diverse workloads.