Updated
Updated · Frontier Enterprise · Aug 5
AMD, Cisco Unveil 1 Control Plane for AI Agents as Inference Tops 60% of Compute
Updated
Updated · Frontier Enterprise · Aug 5

AMD, Cisco Unveil 1 Control Plane for AI Agents as Inference Tops 60% of Compute

3 articles · Updated · Frontier Enterprise · Aug 5

Summary

  • AMD and Cisco said enterprises need a unified control plane to manage AI agents across cloud, private data centers and desk-side systems, with runtime monitoring, policy enforcement and resource control.
  • 35 quadrillion tokens are now consumed each month—up nearly 160-fold in two years—as always-on agents reason, call tools and access data repeatedly, driving steadier and heavier inference demand.
  • 60% of global AI compute capacity will go to inference in 2026, Lisa Su said, marking the first year model running exceeds training and increasing demand for both GPUs and CPUs.
  • Helios racks and Ryzen AI Halo PCs are AMD’s answer to that shift, while Cisco adds observability, security guardrails and unified management for distributed inference near employees as well as in the cloud.
  • AMD tied the strategy to a broader open-platform push and forecast the AI accelerator market will reach $1.4 trillion by 2030, with server CPUs growing from $25 billion to more than $200 billion.

Insights

Will agentic AI make enterprise networks and CPUs as critical as GPUs in the new inference-first era?
If AI agents swarm across cloud, data centres, and PCs, who really controls them when they act on their own?