Alphabet, Amazon, Meta Platforms, Microsoft and Oracle are projected to spend more than $1 trillion a year on capital expenditures from 2027 through 2030, more than 10 times their combined spending in 2020. The scale of the buildout is increasing the financial stakes for shareholders as the companies expand data-center capacity and rely more heavily on debt financing.
Meta’s second-quarter results illustrate the trade-off.
Some data-center financing is also moving through private-credit and off-balance-sheet structures, making the full distribution of risk harder to observe.
The spending must eventually produce enough revenue to support the infrastructure.
Earlier investment booms provide context but do not determine the outcome. Railroad expansion in the 1870s, electrification in the 1920s and telecommunications investment in the 1990s created overbuilt capacity and financial setbacks, while the infrastructure later supported new businesses and services. The comparison leaves open whether AI spending will follow the same pattern.
For investors, the central issue is whether AI revenue and free cash flow will grow quickly enough to justify higher debt, weaker cash generation and capital redirected from measures such as share buybacks. If cloud capacity becomes less differentiated, some eventual value could shift toward companies that build distinctive products and services on top of that infrastructure.