Investment bank UBS is betting that the next big wave in artificial intelligence won’t just be about faster GPUs. It will also spark a major surge in CPU demand, driven by the rise of “agentic AI” software—AI systems that can plan, take actions, and manage multi-step tasks with less hands-on direction.
In a recent investment note, UBS argues that agentic AI increases processor workload in a way that makes CPU performance matter more across the AI stack. Instead of CPUs playing a secondary role, they increasingly act as the coordination engine—handling orchestration, scheduling, control functions, and the many smaller tasks that keep AI systems running smoothly alongside accelerators.
This idea gained fresh attention after Intel’s latest earnings report, where the company posted a major profit beat and its CEO, Lip-Bu Tan, highlighted what he described as clear signs that the CPU is becoming indispensable again in the AI era. Tan said customers are viewing the CPU as the orchestration layer and critical control plane for AI deployments—an argument that aligns with UBS’s broader thesis that CPU relevance is rising, even in a GPU-dominated market.
Where UBS goes further is in explaining which CPU designs stand to benefit most from agentic AI workloads. The bank says these workloads tend to favor processors with higher core counts and strong power efficiency. In other words, as AI agents increase the number of concurrent tasks and background processes, having more cores to spread work across—and doing it without excessive power draw—becomes a key advantage for data centers and enterprise buyers.
UBS expects this shift to significantly expand the server opportunity over the next several years. The bank estimates total server total addressable market (TAM) could grow roughly fivefold by 2030, climbing to about $170 billion from around $30 billion in 2025. In that expanding market, UBS believes Arm-based CPUs are positioned to benefit the most, potentially capturing 40% to 45% of total share.
To support its outlook, UBS points to expert feedback it gathered that highlights three major themes behind the coming jump in CPU demand. First, agentic AI workloads are tilting more heavily toward CPU cores, with expectations that core counts per user—and per GPU—may need to rise by three to five times. Second, servers built around standalone CPUs are projected to require more chips as compute demand scales. Third, agentic AI isn’t limited to the cloud: it can push more work onto local PCs as well, similar to how coding-focused agent tools are beginning to run workloads closer to the user rather than relying exclusively on remote infrastructure.
Put together, UBS sees a clear winner list emerging from these requirements. Arm is viewed as the top beneficiary thanks to its efficiency-driven ecosystem and growing presence in servers. AMD is seen as next in line, also well-positioned to gain from higher-core server demand. Intel is expected to benefit too, mainly because the overall market is growing; UBS also notes Intel could address this demand through its Coral Rapids platform.
For anyone tracking the future of data center hardware, the takeaway is straightforward: as agentic AI adoption grows, CPUs may become one of the most important battlegrounds in AI infrastructure again—especially for chipmakers that can deliver more cores with better performance per watt.
