Hyperscale cloud providers are currently reinvesting over 40% of their sales revenue into AI capital expenditures. Morgan Stanley points out that this level of investment can no longer be sustained solely by existing operating cash flow. Amazon and Google's free cash flow turned negative in the second quarter of 2026, and Meta is expected to turn negative in the next quarter.
By 2027, these hyperscale cloud providers' cash capital expenditures are projected to exceed $1.2 trillion, while total operating cash flow is estimated at around $1 trillion. This means that new infrastructure development will increasingly rely on leasing, public debt issuance, reduced stock buybacks, and new share issuances.
The market narrative for AI capital expenditures in the past was based on the premise that "tech giants have extremely strong cash flow and can self-finance." However, self-financing capabilities are now reaching their peak, and new capital expenditures increasingly require external capital. This will change how the market measures the return on AI investments. Previously, the main comparison was between AI revenue growth and capital expenditure growth; now, financing costs, credit spreads, and equity dilution will all need to be included in the rate of return calculation.
For the market, as more CapEx is financed by debt, project economics will become increasingly dependent on the cost of external capital. If financing costs continue to rise, the return on investment in new AI infrastructure must cover the higher cost of capital.
