Oct 22, 2026 · 34 min
Amazon’s stranded GPUs expose the AI buildout’s weak points
Leaked Amazon document reveal they have too many GPUs | Ed Zitron
Unused NVIDIA hardware, delayed data centers, and ambiguous demand raise questions about whether AI infrastructure spending is outrunning reality.
- 1Amazon’s GPU inventory may reflect construction and power delays rather than healthy demand for AI compute.
- 2Billions in planned data-center spending can conceal uncertain capacity, unclear deployment paths, and hardware that risks obsolescence.
- 3Backlog figures and corporate explanations may obscure falling compute prices and a slowly weakening AI investment narrative.
Don't miss
Ed Zitron reframes Amazon’s unused GPUs as evidence of a broader infrastructure problem, not merely a temporary staging delay.
The brief
Ed Zitron and Isaac Pound examine reports that Amazon has billions of dollars’ worth of NVIDIA GPUs sitting unused as data-center construction and power infrastructure fall behind.
Zitron says the inventory buildup was predictable: companies bought hardware faster than they could build facilities, leaving GPUs exposed to delays and technological obsolescence.
Amazon’s staging-area explanation does not settle the issue, he argues, because the reported documents suggest no clear deployment path despite the company’s logistics capabilities.
The discussion widens from Amazon to the AI economy, questioning data-center capacity estimates, massive compute backlogs, and whether contractual commitments represent real demand.
Zitron’s central warning is gradual rather than explosive: stranded GPUs and falling compute prices may mark a slowly breaking AI bubble.
Books & mentions
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.