GPU scarcity's real cost: compute sold as 3-year blocks starting months out
kevinakwok · x · 2026-09-10
Stefan Bernhardsson argues the biggest downside of GPU scarcity isn't price — it's that nearly all capacity is sold as 3-year-plus "rectangle" commitments starting 3-6 months in the future, which doesn't match how most users' demand actually looks.
More from Infra
- Apple A20 Pro Neural Engine projected to hit 140+ TFLOPS, a third of an A100 for local inference — AIFlow_ML · 2026-09-10
- V4.1 Flash KV cache is 437x smaller than V1, easing China's memory bottleneck — ChrisGPT · 2026-09-10
- DeepSeek-V4.1-Flash Hits the API: 1/8 the SSD Cache, V4-Pro Being Phased Out — deepseek_ai · 2026-09-10
- Aspen Digital webinar reimagines data centers as public compute for communities — lfschiavo · 2026-09-10
- New technique claims another 4x per-token KV cache size reduction — zephyr_z9 · 2026-09-10
- Running a 2.78T-parameter Kimi K3 inference on a single CPU in 8.24 GB of RAM — gekobraa · 2026-09-10