Arm exec: AI datacenters need a third 'agentic CPU' class as DPUs become security table stakes
BenBajarin · x · 2026-10-31
- Ben Bajarin shares highlights from a paid interview with Mohamed Awad, EVP of Arm's Cloud AI business unit.
- CPUs are strategic again: GPUs do the thinking, CPUs do the actions — tool calls, code execution, database queries and next-step decisions all land on CPUs. Arm entered the merchant CPU market this year with a part built for agentic workloads.
- Three CPU classes in AI datacenters: accelerator head nodes, cloud-native CPUs hyperscalers have run for two decades, and a new "agentic CPU" class acting as virtual users — matching Bajarin's general-purpose / accelerator-host / agent-execution forecast framework.
- DPUs rise: once agents become users, operators can't assume workloads are trusted; security compute becomes table stakes, pushing DPUs and hardware isolation deeper into the AI control plane.
More from Infra
- Anthropic S-1 math: subscriptions are ~18% of revenue but burn over half the compute — teortaxesTex · 2026-10-31
- Reddit: 4-bit quantization could cut memory 75%, but incentives block it — Excellent_Place4977 · 2026-10-31
- NVIDIA discloses Brev platform breach: SSH keys exposed Oct 4–20 — HowDevelop · 2026-10-31
- Northflank hacked: users told to rotate all secrets as AI keys confirmed abused — basedjensen · 2026-10-31
- Seroter's 900th reading list: hill-climbing agents and prefix caching costs — rseroter · 2026-10-31
- Post-training API ships end-to-end custom losses: define locally, run on hundreds of GPUs — ibab · 2026-10-31