AMD gains edge in agentic AI data center race

Agentic AI is expected to reshape data center hardware demand as autonomous systems rely more heavily on CPUs for sequential reasoning and action execution. Nvidia projects the market could reach $200 billion in the coming years, while industry forecasts suggest the GPU-to-CPU ratio could move from 8-to-1 in training to 4-to-1 for inference and eventually 1-to-1 for agentic AI workloads.

AMD is positioned as the strongest candidate among AMD, Intel and Arm because of its data center CPU momentum and upcoming Venice architecture, which packs up to 256 cores onto a single chip. The company is also building a second growth engine in inference through its Instinct GPUs, the acquisition of memory optimization specialist MEXT, and major deals with OpenAI and Meta Platforms.

Intel has benefited from strong data center CPU demand and a stock surge of approximately 323% over the past year, but faces pressure from a stagnant PC business, higher component costs and continuing losses in its foundry unit. Arm is pursuing a bigger server opportunity by moving beyond licensing into its own CPUs, targeting a $100 billion data center CPU market over the next five years and a 15% share, but the shift risks straining its licensee ecosystem and manufacturing capacity.