The AI Demand Bubble: When Hyperscalers Built a Trillion-Dollar House of Cards

Microsoft has spent over $260 billion on AI infrastructure since 2022. Yet 70% or more of its AI revenues come from a single customer: OpenAI. That same company just raised $122 billion in March and delayed its IPO because advisors didn't think it would get a trillion-dollar valuation. This is not a sign of a healthy, growing industry. It is the central mechanism of what Ed Zitron calls "the AI demand bubble."

The 70% Problem

Zitron's newsletter, published hours before Bloomberg independently confirmed his findings, pulled together financial analyst notes from Wells Fargo, Barclays, and UBS. All three estimated that 70% or more of the AI revenues for Microsoft, Google, and Amazon come from either OpenAI or Anthropic. UBS went further, estimating that Anthropic and OpenAI's compute spend alone will account for 48% of all Google Cloud revenues in the coming year.

The implications are stark. Microsoft's fiscal year 2026 year-over-year growth was about 17.7%. Remove OpenAI's $24.1 billion contribution—which comes from OpenAI's compute spending and revenue share—and Microsoft's growth drops to roughly 9.24%, its lowest since fiscal year 2023. Meanwhile, Microsoft spent $115 billion in capital expenditures in fiscal year 2026 alone, and plans to spend another $175 billion in fiscal year 2027.