Architect pushes back on 'AWS sold out through 2027' claims as hype
DavidLinthicum · x · 2026-10-24
Responding to reports that AWS cloud capacity is "largely sold out through 2027," cloud architect David Linthicum calls it worst-case hyperbole in an already confused AI infrastructure market.
His key points:
- Selling out massive, pre-reserved, multi-year hyperscale capacity blocks that large model providers want is not the same as "AWS is sold out"
- Enterprises can still log into the AWS console today and allocate GPU capacity for AI workloads
- The market already has enough confusion around GPU capacity, reserved instances, on-demand services, and committed use — panic-driven headlines make it worse
Takeaway: no need to panic-migrate; the headline conflates tight hyperscale reserved capacity with a general capacity shortage.
More from Infra
- TensorSharp vs llama.cpp: audited benchmarks show up to 5.5x prefill gains on A40s, with caveats — fuzhongkai · 2026-10-25
- New Yorker longread: semiconductors are the most magnificent things humanity has ever made — SuB8u · 2026-10-25
- ik_llama.cpp Fork Update: Matches Strata Speeds on a 6GB RTX 2060 — IceFog72 · 2026-10-25
- Watch live: YC Paper Club on the LLM inference stack — ycombinator · 2026-10-25
- YC Paper Club on inference: every training FLOP costs ~6x per year in serving, a $220B stack — ycombinator · 2026-10-25
- Running DeepSeek at 10 tok/s on an RTX 5090 costs more in electricity than API calls — DeMischi · 2026-10-25