S&P Global Ratings projects hyperscaler AI capital expenditure will exceed $1.3 trillion while their combined free cash flow turns negative.

Combined capital expenditure at the world’s largest hyperscalers is projected to exceed $1.3 trillion by 2027. According to new research published August 27, 2026, all six companies will run negative free operating cash flow through 2026 and 2027, with no recovery projected before 2029. The findings come from S&P Global Ratings’ analysis of how the artificial intelligence buildout is being financed.

Titled “S&P Global Ratings’ View On Artificial Intelligence And Hyperscalers,” the report examines the funding models at Alphabet (GOOG), Amazon (AMZN), Microsoft (MSFT), Meta, Oracle (ORCL), and SpaceX. Its central conclusion focuses less on the sheer scale of spending and more on what is paying for it: operating cash flow no longer covers the bills. Consequently, the six firms are increasingly relying on debt, equity issuance, lease commitments, and other financing arrangements to sustain construction.

“As AI infrastructure investment accelerates, the focus is expanding beyond the scale of spending to the funding models, financial commitments and long-term implications that accompany it,” said Naveen Sarma, Managing Director and Sector Lead at S&P Global Ratings. “Understanding how these investments are financed and managed will be increasingly important in assessing credit quality across the sector.”

The report’s detailed examination of the financing mix reveals a growing reliance on joint ventures, special purpose vehicles, residual value guarantees, and similar structures, all of which increase the complexity of credit analysis. In this context, a residual value guarantee is a commitment by the hyperscaler to compensate a lessor or financing partner if servers, GPUs, or data-center assets trade below their expected value at the end of a lease term. While such obligations sit off the headline debt figure, they function like debt whenever hardware values decline.

This structure carries direct implications for holders of hyperscaler bonds and firms leasing capacity to them. S&P notes it is now monitoring contractual commitments and other debt-like obligations alongside the monetization of AI investments, demand durability, and overcapacity risk. As more of the buildout is funded through off-balance-sheet vehicles and guarantees rather than retained cash, credit exposure disperses across lessors, lenders, and joint-venture partners instead of concentrating on a single corporate balance sheet.

S&P’s financial models generally anticipate a 2028 inflection point, assuming revenues will accelerate and capital expenditure growth will moderate as monetization improves. Until then, the agency projects a multi-year window in which the six largest buyers of compute act as cash consumers rather than generators on a free operating basis—a period during which equipment vendors, landlords, and power providers recognize the corresponding revenue. Securities.io has tracked this capital flow-through at the site level, noting Digital Realty’s 50-megawatt allocation in Singapore and Bitdeer’s $400 million of contracted AI capacity within a 350-megawatt Malaysian facility, as well as at the equipment layer through Vertiv’s cooling and power exposure to the same order cycle.

While the negative-cash-flow projection remains a forecast, its trajectory is already visible in recent corporate filings. Microsoft, one of the six firms analyzed, reported $115.9 billion in additions to property and equipment for its fiscal year ended June 30, 2026, against $182.9 billion in net cash from operations. That represents an 80% year-over-year increase from the $64.6 billion spent in the prior year, according to the company’s July 29, 2026 earnings release. Although that surge still left Microsoft cash-flow positive after capital expenditures, S&P’s projections indicate that even companies of this scale will cease self-funding as the 2026–2027 spending tranches materialize, with smaller balance sheets reaching that threshold sooner.

Microsoft’s disclosures also underscore the demand-side foundation of S&P’s 2028 thesis. Commercial remaining performance obligations—contracted business not yet recognized as revenue—rose 84% year over year to $678 billion as of June 30, 2026. Additionally, Azure revenue surpassed $100 billion for the fiscal year. These contracted backlogs form the revenue base that S&P’s models rely upon to project monetization eventually catching pace with construction.

S&P explicitly stated that the report does not constitute a rating action. No hyperscaler’s credit rating was altered upon publication, and none of the six companies received a new outlook. The projections reflect the agency’s internal model assumptions, including the 2028 inflection and 2029 cash-flow recovery, and the report outlines specific conditions that would stress those forecasts: slower-than-modeled AI monetization, less durable demand, or overcapacity emerging if infrastructure buildout outpaces actual usage.

For bondholders and financing counterparties currently holding residual value guarantees and lease commitments, the analysis hinges on three key milestones: 2027, when cumulative capital expenditure crosses $1.3 trillion; 2028, the modeled revenue inflection; and 2029, the first year S&P projects positive free operating cash flow returns across the group.