Advanced Micro Devices vs. Arm Holdings: Which AI CPU Stock Is the Better Buy?

Artificial quality (AI) infrastructure has been dominated by graphics processing units (GPUs). But the adjacent signifier of AI adoption, particularly inference (the deployment of AI models successful accumulation environments) and agentic AI, is besides driving accrued request for cardinal processing units (CPUs).

In ample AI systems, CPUs assistance coordinate information movement, networking, and orchestration crossed aggregate AI chips. That makes Advanced Micro Devices (NASDAQ: AMD) and Arm Holdings (NASDAQ: ARM) 2 precise antithetic ways to put successful the AI CPU opportunity.

Missed Nvidia successful 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" awesome flashed for a little-known chipmaker called Nvidia. For the archetypal clip successful years, that aforesaid "Total Conviction" awesome is flashing for a institution 1/100th the size of Nvidia. Continue »

Let's measure which AI CPU banal is simply a amended bargain now.

AMD's CPU strategy

AMD is progressively valued arsenic a information halfway infrastructure company, not conscionable a PC oregon gaming spot player. In the archetypal 4th , AMD's gross roseate 38% twelvemonth implicit twelvemonth to $10.3 billion. The company's information halfway conception gross jumped 57% to $5.8 billion, driven by beardown request for EPYC server CPUs and the continued ramp of Instinct AI GPUs.

The institution present expects the server CPU full addressable marketplace (TAM) to turn astatine much than 35% annually and transcend $120 cardinal by 2030. This is simply a important upward revision from the expected 18% yearly maturation for the adjacent 3 to 5 years. Management besides expects server CPU gross to turn by much than 70% twelvemonth implicit twelvemonth successful the 2nd quarter.

AMD is seeing robust CPU request from regular server computing, caput nodes that assistance negociate GPUs and different AI accelerators, and agentic AI workloads. As agentic AI workloads grow, each AI cause tin make much CPU tasks for orchestration, information processing, and parallel execution. In older systems, 1 CPU often supported 4 oregon 8 GPUs.

But arsenic inference and agentic AI workloads grow, AMD believes immoderate systems whitethorn necessitate 1 CPU for each GPU. In highly agentic workloads, determination could adjacent beryllium much CPUs than GPUs. Hence, EPYC CPUs are proving to beryllium a nonstop beneficiary of the agentic AI build-out.

AMD's 6th Gen EPYC processor, besides called Venice, has begun ramping connected Taiwan Semiconductor Manufacturing's precocious 2-nanometer process. Advanced manufacturing nodes tin assistance chips present much computing show portion utilizing little powerfulness per task. Since powerfulness and cooling are becoming captious bottlenecks successful AI information centers, Venice whitethorn spot coagulated request successful the coming years.