Agentic AI Shift Could Drive Demand for Data Center CPUs

The architectural requirements of artificial intelligence infrastructure are evolving beyond raw processing power. As the industry moves toward “agentic” AI—systems capable of sequential reasoning and decision-making—the reliance on central processing units (CPUs) is expected to grow significantly within data centers. While graphics processing units (GPUs) have historically handled the heavy lifting of parallel processing, CPUs provide the necessary logic for AI agents to “think” before executing tasks.

Market analysis indicates a shift in the ratio of GPUs to CPUs. Current training workloads often utilize an 8-to-1 ratio of GPUs to CPUs. However, as workloads transition to inference and eventually to agentic AI, this ratio is projected to compress. Inference operations may drop to a 4-to-1 ratio, while fully realized agentic AI systems could require a 1-to-1 ratio. This shift suggests a substantial upcoming increase in CPU deployments.

NVIDIA Corporation has forecasted that the total addressable market for this segment could reach $200 billion in the coming years. This growth trajectory presents opportunities for major semiconductor manufacturers to expand their data center footprints.

Advanced Micro Devices, Inc., currently trading with a market capitalization of approximately $901 billion, is positioned to supply the necessary accelerators and microprocessors. Meanwhile, legacy giant Intel Corporation (market cap: roughly $541 billion) continues to leverage its client compute and data center segments to capture share in this evolving landscape. Both companies, alongside ARM-based architecture providers, are critical to supplying the hardware required for this next generation of AI computing.

What to watch

• Data Center Revenue: Monitor quarterly reports from AMD and Intel for specific growth in server and data center segment sales.
• Product Roadmaps: Look for announcements regarding new CPU architectures specifically optimized for AI inference and agentic workloads.
• Market Guidance: Watch for revised outlooks or capital expenditure updates from cloud providers that indicate a changing ratio of CPU to GPU procurement.

Source: original release