Growing use of autonomous AI agents could drive a four-fold increase in data center CPU core requirements. CPUs have been declared all but irrelevant in the era of artificial intelligence, but they are receiving a new lease on life thanks to agentic AI creating a new source of demand.
According to semiconductor research firm TrendForce, the growing computational requirements of autonomous AI agents could force operators to substantially increase CPU deployments and reconsider the balance between CPUs and GPUs in their server infrastructure.
TrendForce said AI agents introduce a more CPU-intensive workload than conventional large language model (LLM) inference. While GPUs remain essential for executing the training portion of AI models, CPUs are responsible for coordinating the complex sequences of actions that enable AI agents to operate autonomously, steps that a GPU simply can’t do.
These activities include planning tasks, scheduling subtasks, calling external tools, managing interactions between multiple agents and evaluating results. Agents also rely on conventional computing resources to execute Python code, search databases, crawl websites and process information, according to TrendForce.
As a result, the CPU is enjoying a new lease on life and becoming a more significant component of AI infrastructure rather than simply serving as a supporting processor for GPUs, according to TrendForce.
TrendForce notes that traditional AI data centers have typically deployed approximately one CPU for every four to eight GPUs, reflecting the industry’s emphasis on accelerating computationally intensive AI workloads.
But agentic AI could change all that. TrendForce cited estimates from Arm indicating that CPU core requirements could increase from approximately 30 million cores per gigawatt of data center capacity to 120 million cores per gigawatt as agentic AI workloads become more prevalent.
That represents a potential fourfold increase in CPU core demand. TrendForce expects the CPU-to-GPU ratio in agentic AI infrastructure to move toward one CPU for every one to two GPUs, compared with significantly lower CPU deployment in many existing AI deployments.
The shift could also affect system performance, energy consumption and infrastructure costs. Research cited by TrendForce indicates that CPU-based tool processing can account for as much as 90.6% of total latency in some agentic workloads. CPU dynamic energy consumption can also represent a substantial portion of total system energy use under certain operating conditions.
These figures underscore the importance of balancing CPU and GPU resources. Increasing GPU capacity alone may not improve overall performance if processors responsible for coordinating AI agents and executing their associated tasks become bottlenecks.
A clear sign of this growing need for CPUs is coming from, of all places, Nvidia. For the longest time, Nvidia downplayed and even disparaged CPUs, positioning the GPU as the be all end all for everything.
But it changed its tune a few years ago with the introduction of the Grace processor, now on the Vera generation. Designed to work closely with Nvidia’s GPU platforms, Vera features 88 CPU cores and 176 threads, along with a high-bandwidth NVLink-C2C interconnect for communication and memory sharing with Nvidia accelerators.
Nvidia is also developing dedicated Vera CPU racks, allowing customers to deploy CPU systems independent of GPU-heavy or GPU-only systems. So it’s clear that Nvidia sees the CPU is needed to do things that the GPU cannot.
TrendForce’s analysis points to a broader change in AI computing: as systems move from generating individual responses toward executing multistep tasks autonomously, the processors coordinating those tasks become increasingly important.