Buyer told to wait 8-12 months for an H100: the GPU shortage is about power, cooling and talent, not chip supply
ingliguori · x · 2026-09-29
The author tried to buy an H100 GPU last month and was told to come back in 8-12 months, while NVIDIA's market cap keeps climbing.
He argues the GPU shortage isn't about chip fabrication but about supporting bottlenecks:
- Power infrastructure: GPUs can't run without enough electricity
- Cooling systems: the hardware runs extremely hot
- Talent: few people know how to actually optimize GPU usage
Tech companies are hoarding GPUs like dot-com domains in 1999, but the winners will be those who use them smartest, not those who own the most. He closes by asking whether readers are buying or building their AI infrastructure.
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