Hundreds of billions of dollars a year are being committed to AI infrastructure against revenues that remain a fraction of that figure, and this page keeps the indicators that will settle the argument — reported capex, reported revenue, utilization, prices — alongside every dated bubble call and boom call on the record. In the last week of July 2026 the gap stopped being an abstraction and became a cash-flow event. Alphabet posted the first negative free-cash-flow quarter in its history as a public company, spending $44.9 billion of capital against $39.1 billion of operating cash, and raised its full-year guidance to $195–205 billion anyway; its shares then had their worst day in over a year despite a 24% revenue rise, the clearest sign yet that the market has inverted its verdict on AI spending. The credit side had already moved: S&P cut Oracle to one notch above speculative grade over its AI capex, and Moody's counted roughly $460 billion of hyperscaler debt against $1.2 trillion of lease commitments, most of it on data centres not yet built. Meanwhile the financing of demand changed instrument, with Nvidia reported — unconfirmed — to be in early talks to guarantee as much as $250 billion of OpenAI's leases.