The numbers are staggering. Analysts report that hyperscaler spending on AI-related capital expenditures surged from roughly $250 billion in 2024 to more than $430 billion in 2025, with spending expected to hit $600-$700 billion in 2026. Amazon alone has committed to $200 billion in capex for 2026, Alphabet $175–185 billion, Microsoft to roughly $120 billion, Meta to $115–135 billion, and Oracle to $50 billion. For some of these names, this represents approximately 40%-50% of annual revenues, a level of spending that is unprecedented coming from traditionally investment-light businesses.
Further analysis shows that cash simply isn't keeping up. Bank of America estimates that these hyperscalers are positioned to consume roughly 90% of their operating cash flow on capex in 2026, up from 65% in 2025. Analysts have estimated that Amazon is projected to post negative free cash flow of $17–$28 billion this year, a stark reversal of recent trends.
The consequences are that the shareholder return model is now straining under the weight of these new investments. Market research reports outline that combined share buybacks across these five companies plunged to $12.6 billion in Q4 2025, the lowest level since Q1 2018, compared to a peak of $149 billion in 2021.
The natural question becomes, where is all the money coming from? The response has been a turn to debt markets.
In 2025, hyperscaler-linked corporate bond issuances totaled $121 billion in the U.S., versus an average of just $28 billion per year from 2020 to 2024. The individual transactions paint an even clearer picture. Meta issued a record-breaking $30 billion in bonds, Alphabet raised $25 billion, Oracle $18 billion, and Amazon $15 billion. This pace shows no signs of slowing down in 2026 either. This group of companies has already issued over $100 billion in U.S. bonds as of March 2026, nearly their total for all of 2025.
A few are getting creative with how they are structuring these debt issuances as well, with Meta using an off-balance-sheet special purpose vehicle to fund a data center in Louisiana. This could be a sign of things to come, with total hyperscaler borrowing showing no signs of slowing down.
This is a fundamental identity shift for technology companies that have historically commanded premium valuations built on the assumption that software scales without proportional capital investment, leaving room for ever-increasing profits. These reserves of cash have funded buybacks, research expenses and “moonshot” investments like Alphabet’s now successful autonomous vehicle project called Waymo.
The bond market has now taken the place of retained earnings to finance the future of these technology giants. The question isn't necessarily whether these businesses can pay the debt, they very well could. The larger question is whether markets will continue to price them as high-multiple software businesses once balance sheets and cash flow profiles look increasingly different from what investors are used to seeing. The answer largely depends on what the returns are to these massive investments, which remains an interesting question the markets have yet to see definitive evidence for.
Given these potential uncertainties, investors would be well-served to maintain exposure to the possible upside while ensuring they are diversified in their portfolios against downside risk if returns on these significant capital expenditures do not materialize.