AFFN disaggregation is the holy grail of inference networking, says Gavin Baker
GavinSBaker · x · 2026-08-24
Investor Gavin Baker argues that AFFN disaggregation — splitting attention/FFN inference stages across different hardware — is the holy grail, but an extremely hard networking problem.
The post he replies to describes the new "Pareto optimal" scheduling: agent CPU control & tool-call scheduling, GPU prefill (KV cache), and ASIC FFN decode (e.g. Cerebras). It further claims AWS is technically prepared for this era, other hyperscalers are lagging, and no NeoCloud or inference leader today comes close.
Related event: Investor Predicts 100x Growth in AI Spending and New Computing Paradigm(4 posts)→
More from Infra
- Wells Fargo sees Broadcom AI chip revenue at $205B by FY28, far above consensus — Beth_Kindig · 2026-08-24
- Disaggregated LPDDR memory: A new path for AI infrastructure — jwt0625 · 2026-08-24
- Study notes: how speculative decoding accelerates LLM inference without quality loss — helloiamleonie · 2026-08-24
- Global AI Power Demand Projected to Surge 1,100% by 2033 — KyeGomezB · 2026-08-24
- RTX 3080 Memory OC to +1200 Boosts Flux Generation Efficiency — MakionGarvinus · 2026-08-24
- xyOps: Open Source Platform for Scheduling, Automation, and Monitoring — tom_doerr · 2026-08-24