GPU rentals up 27% in six months while fixed-intelligence prices fell 63%
jon_durbin · x · 2026-10-12
jondurbin discusses the inference economy's paradox: LLM inference is a race to zero while GPU prices race to infinity. He shares charts from his Parallax project:
- Apr–Oct 2026: top GPU (H100/H200/B200/B300) rental rates rose 27% (+3.9%/month), B300 up 28%
- Over the same period, the cheapest price for a fixed intelligence level (GPQA Diamond 80/85/90%) fell 63% (-15.5%/month)
- Trend projection: GPU rental index hits 141 by Jan '27 while the intelligence index drops to 22
He says Parallax tries to bridge that gap, and that rising costs forced his business away from tokenmaxxing toward more careful usage. He also notes his precision-trading side bets (TSLA, SPCX) and getting burned on small $MU test trades.
Related event: LLM Inference Prices Plummet While GPU Rental Costs Soar(3 posts)→
More from Infra
- NVIDIA's ELR removes the one-loop tax: Koi-4B scores 18 at one loop, 29 at four loops — nvidia · 2026-10-12
- Turn your dusty old laptop into a home server: Proxmox + NetBird starter guide — aestheticedwar1 · 2026-10-12
- Mac Studio M5 Ultra 512GB vs same-price RTX 5090 rig for local AI video generation — This-Gene1631 · 2026-10-12
- NVIDIA opens Vera Rubin rack designs at OCP as Cisco, Google, Meta adopt — nordicinst · 2026-10-12
- TSMC and ASML explore 12-inch photomasks to unlock High-NA EUV full-field potential — pstAsiatech · 2026-10-12
- The top two US counties for data centers are also the two richest — is there a flywheel? — gregmushen · 2026-10-12