Chamath breaks down AI compute: prefill is compute-bound, decode is bandwidth-bound
rohanpaul_ai · x · 2026-09-28
Chamath explains the two key phases of AI compute: Prefill processes the input context and is compute-bound, favoring massively parallel GPUs — which is why Nvidia dominates as context lengths grow. Decode generates tokens one at a time, with each next token requiring a scan of everything generated so far, making it memory-bandwidth bound rather than compute-bound. The takeaway: inference optimization's real battleground is bandwidth, not FLOPs.
More from Infra
- Gemini 3.5 Flash now runs on air-gapped Google Distributed Cloud, offline up to 12 months — shashib · 2026-11-16
- Gemma.c: full Gemma 1/2/3 multimodal inference in a single C file, ~72 tok/s on CPU at INT8 — This-Purchase-3325 · 2026-11-16
- Sarvam AI Launches Token Factory: India-Hosted Inference for Open Models — itsOmSarraf_ · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16