AMP Grid says bad scheduling and low MFU waste most AI compute
Kyrannio · x · 2026-07-24
The quoted AMP Grid remarks frame wasted AI compute as an infrastructure problem, not just an accounting one.
- Long-term leases can lose about 30–40% of FLOPs to bad scheduling.
- Inside the chip, only about 15% of FLOPs may be utilized because work is waiting on storage, memory, networking, or another chip.
- Combined, the speaker argues that roughly 85% of FLOPs can be wasted, which he describes as a national security issue.
Related event: AMP Founder Says Inefficient Scheduling Wastes Most AI Compute(2 posts)→
More from Infra
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11