Tokens get cheaper while GPUs get pricier: H100 rental up ~30% since March
julsimon · x · 2026-09-28
julsimon runs the numbers: buying tokens is getting cheaper, renting GPUs is getting more expensive.
Per million output tokens (falling):
- Anthropic Claude Sonnet 5: $15 → $10 (Aug)
- xAI Grok 4.20: $6 → $2.50 (Jul)
Per GPU-hour since March (rising):
- RunPod H100 SXM: $2.69 → $3.49
- Lambda H100 SXM: $3.44 → $3.99
- AWS and Google Cloud: flat
Takeaway: model vendors are cutting per-token prices while the GPU rental market keeps raising rates on demand.
Related event: H100 Rental Prices Surge Up to 30% in Six Months(3 posts)→
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