Key thing to consider when talking about margin in aI. Yes, when that compute is being used for inference, margin is "high". 85% really isn't high by s/w industry standard.

What happened: A post argues that AI inference margins described as high, at 85%, are not high by software industry standards, and points to capex sitting unused, including GPUs in warehouses for American hyperscalers and domestic chips installed in China in 2023-2025. Key numbers: 85% margin cited for inference compute; China chip installations dated 2023-2025. Why it matters: The post questions how strong AI margins really are once idle capex is counted, and says the point comes from people involved in AI data center build out rather than from Ed Zitron.

A post argues that AI inference margins described as high, at 85%, are not high by software industry standards, and points to capex sitting unused, including GPUs in warehouses for American hyperscalers and domestic chips installed in China in 2023-2025.

The post questions how strong AI margins really are once idle capex is counted, and says the point comes from people involved in AI data center build out rather than from Ed Zitron.