// infrastructure-costs

Source: Chartbook

For the first time in two consecutive quarters, global AI vendors are generating enough revenue ($25B) to cover the actual wear-and-tear costs of their capital-intensive infrastructure. The industry has moved beyond pure cash burn. This crossing point doesn't mean profitability—operating costs, R&D, and other expenses still dwarf gross margins—but it marks when the AI buildout stops being a pure sinkhole and becomes a functioning business model that can theoretically self-fund its own infrastructure replacement. The exclusion of China from these figures reveals a bifurcating AI economy where Western vendors are moving toward sustainability while China's AI sector operates under different capital dynamics and ROI timelines.