# As AI moves from single-task generation to autonomous, multi-step "agentic" systems, the economic value is migrating to the broader infrastructure stack.

*genai · news · 2026-04-22 · Livemint*

## Key points

- Morgan Stanley reports CPUs and memory are emerging as new bottlenecks for agentic AI systems.
- Agentic AI can make CPU-side orchestration account for 50-90% of workload latency.
- Agentic workloads could drive 15-45 exabytes of extra DRAM demand by 2030.
- CPU vendors and memory suppliers may gain as much revenue as GPU leaders by 2030.
- AI infrastructure investment is broadening to include foundries, interconnects, and advanced packaging providers.

**Companies:** Nvidia
**Countries:** India

[Read the full story on Livemint](https://www.livemint.com/technology/morgan-stanley-agentic-ai-shifts-value-from-gpus-to-cpus-and-memory-creating-up-to-60bn-incremental-cpu-tam-by-2030-11776844417171.html)

---

Canonical: https://newsio.io/n/f90604bf-728a-43e1-8fa4-9f87efe20232/as-ai-moves-from-single-task-generation-to-autonomous-multi-step-agentic-systems
Summarized by Newsio from Livemint. https://newsio.io/how-it-works