"This will happen to inference": Suhail bets LLM inference costs will fall 1000x like lighting did
npew · x · 2026-09-15
Resurfacing Paul Graham's data point that the cost of lighting fell over 1000x through technological progress, Suhail argued the same trajectory awaits AI inference: "This will happen to Inference."
OpenAI researcher nat (npew) re-shared the take, signaling agreement. The implication: today's expensive tokens are on a steep long-term cost-decline curve as infrastructure improves.
Related event: Lighting Cost Fell 1000x—Suhail Says AI Inference Is Next(2 posts)→
More from Infra
- RTX 5090 doubles to $6,899 as RAM prices eclipse GPUs in worst-ever PC build market — Yuchenj_UW · 2026-09-15
- Software Engineer vs LLM Inference Engineer: Three Key Differences Explained — ashishllm · 2026-09-15
- GTX 1080 Ti + MI50 Vulkan llama.cpp benchmarks: small MoE hits 640 t/s prefill — tabletuser_blogspot · 2026-09-15
- DDRop: dropping DDR5 writes breaks freshness in confidential VMs on Intel TDX and AMD SEV-SNP — jedisct1 · 2026-09-15
- Video of New Albany, Ohio data centers stuns: scale is 'staggering' — natesiggard · 2026-09-15
- HTTP 200 is a weak success metric for AI endpoints, engineer warns — gethackteam · 2026-09-15