Agentic AI Drives Surge in Data Center CPU Demand as Intel Unveils 36,864-Core Rack Systems

Data Center CPUs Reclaim the Spotlight as Agentic AI Reshapes Infrastructure

While GPUs have dominated headlines throughout the generative AI revolution, data center CPUs are experiencing an unexpected resurgence. The surge in demand for data center CPUs stems from a fundamental shift in how AI systems operate, moving beyond simple chatbot interactions to complex agentic AI systems that autonomously plan, decide, and execute multi-step tasks 1

1

. AMD now forecasts the CPU market will reach $120 billion by the end of the decade, with annual growth doubling from 18% to 35% 1

1

.

Why Agentic AI Demands More CPU Power

Agentic AI represents a departure from traditional AI assistants that simply respond to queries. Instead of waiting for commands one by one, AI agents can plan actions, make decisions, and complete tasks automatically with minimal human input 3

3

. Jason Beckett, CTO at Hitachi Vantara, explains that "always-on, multi-step reasoning systems don't create brief orchestration bursts around GPU workloads. They demand high-core-count CPUs running at sustained loads, continuously" 1

1

. These systems need constant coordination of tasks, memory management, workflow handling, and communication with other software—operations where CPUs excel.

The CPU to GPU Ratio Is Shifting Dramatically

Early generative AI deployments prioritized GPUs heavily, with ratios ranging from four to eight GPUs per single CPU 1

1

. But as AI infrastructure evolves to support inference and agentic workloads, this balance is changing. Analyst Ben Bajarin suggests the industry could move toward a one-CPU-to-one-GPU setup in the future 3

3

. Jeff Moore from Aegis Cooling notes that "as GPU clusters scale, CPUs are taking on larger roles in orchestration, memory management, networking, storage coordination, and inference handling" 1

1

. Recent TrendForce analysis reveals CPUs contribute nearly 91% of latency in AI responses, making their optimization critical for agentic systems 1

1

.

Intel Unveils Massive Rack-Scale Reference Designs at Computex

At Computex 2026, Intel announced rack-scale reference designs developed with Foxconn and other infrastructure providers, specifically targeting CPUs for agentic workloads 2

2

. Intel CEO Lip-Bu Tan revealed two blueprints: one optimized for latency-sensitive operations and another for maximum compute densities 2

2

. Both designs support up to 128 Intel Xeon processors—either the 128-core Granite Rapids Xeon 6 or 288-core Clearwater Forest Xeon 6+—delivering between 16,384 P-cores and 36,864 E-cores alongside up to 384 TB of DDR5 memory in a 100kW power envelope 2

2

.

Hyperscalers Drive Structural Shift in AI Infrastructure

Hyperscalers recognize the integral role CPUs play in developing AI clusters that will power future economies. Roger Cummings, CEO of PEAK:AIO, identifies "two structural shifts driving the demand surge: the rise of agentic AI and the need for deterministic, predictable performance at rack scale" 1

1

. Intel's new Xeon 6+ processors, built using the company's 18A process technology, are designed specifically for cloud-native AI workloads, networking, and large-scale inference systems 3

3

. While AI models predominantly run on GPUs and accelerators, the agent harnesses connecting them to tools, terminal shells, code interpreters, and APIs still run on CPUs 2

2

. Nvidia and Arm are also developing competing rack-scale CPU platforms, with Nvidia announcing a system packing 256 of its 88-core Vera CPUs, and Arm working on designs with up to 45,696 cores 2

2

. As millions of users deploy AI agents that continuously perform tasks and coordinate applications in real time, data centers must handle significantly more orchestration and system communication—operations where CPUs serve as the critical "manager" keeping infrastructure running smoothly.🟡,