Goldman says CPU will top GPU in near-term capex as agent workloads explode
bookwormengr · x · 2026-09-21
Reacting to a claim that labs see personal agents as the next "Claude Code moment" and that Goldman now calculates CPU > GPU in short-dated capex, bookwormengr shares a practitioner's view:
- Only agentic environments actually keep GPUs busy for extended periods; most agent steps—web search, file search, Python execution, browser use, computer use, and waiting on MCP tool responses—are CPU-bound
- AWS AgentCore runtime can share CPUs across multiple agents since they don't run continuously, but the bottom line stands: we need billions of CPU cores
The thread frames agent workloads as a shift where CPU capacity, not GPU, becomes the near-term infrastructure bottleneck.
More from coding & agent
- e2b to demo agents forking live VMs to explore solutions in parallel — badphilosopher · 2026-09-21
- User drops Astra after bad week, says Codex needs a real orchestrator — koltregaskes · 2026-09-21
- Agent evals' cold start is smaller than you think: start ugly, iterate — sarahcat21 · 2026-09-21
- Models Don't Have Agency. Systems Do.: Structured Output Is What Makes Agents Work — sethjuarez · 2026-09-21
- LukeW: Using the humble ALT tag to fix irrelevant image retrieval in AI replies — LukeW · 2026-09-21
- Dev shares AGENTS.md rule: skip unit tests after coding, rely on E2E only — alexcovo_eth · 2026-09-21