Hyper 5 AI Spending: Why $5.3 Trillion in Capex Is Testing Cash Flow

The AI infrastructure race has become a test of cash-flow timing rather than a test of whether the largest cloud companies can build. S&P Global’s analysis describes an exceptionally concentrated capital-spending cycle led by Amazon, Alphabet, Microsoft, Meta and Oracle—the “Hyper 5.” They have substantial profitable businesses behind them, yet the pace of investment is pushing free cash flow lower and increasing reliance on financing beyond internally generated cash. 12

The scale: $1.1 trillion spent, $5.3 trillion more projected

S&P Global Market Intelligence says the Hyper 5 collectively spent $1.1 trillion in capital expenditures over the past five years. Visible Alpha consensus estimates cited by S&P Global point to a further $5.3 trillion through 2030. Those are projections, not committed spending, but they convey the extraordinary scale of the expected AI and digital-infrastructure build-out. 1

The annual pace has accelerated sharply. Major technology companies’ capital expenditure rose from about $80 billion in 2019 to roughly $700 billion in 2026, with about $850 billion projected for 2027, according to S&P Global reporting from a webinar. 5

This spending supports the physical layer required for AI services: data centers, computing capacity and related digital infrastructure. Alphabet, Amazon and Microsoft alone indicated about $495 billion of 2026 capex in their fourth-quarter 2025 earnings calls, much of it tied to technical infrastructure and AI data centers. 6

Why free cash flow is the pressure point

A company can remain profitable while producing weak or negative free cash flow. Profit measures operating performance over an accounting period; free cash flow is reduced when a business makes very large up-front investments in long-lived assets.

That distinction is central to S&P’s assessment. S&P Global Ratings expects negative free operating cash flow across the six largest hyperscalers in both 2026 and 2027. Its group includes the Hyper 5 plus SpaceX. 2

The implication is not that the companies lack earnings power. Rather, outlays for AI capacity are arriving before the full revenue, utilization and returns from that capacity are certain. S&P Global Ratings has said that rising capex, increasingly complicated financing and long payback periods are gradually weakening hyperscalers’ credit quality. 37

The funding mix is shifting beyond internal cash

In the earlier stages of the cycle, the companies could primarily draw on operating cash flow. As investment requirements rise, S&P Global Ratings says debt, leases, guarantees and other financing structures are playing an increasingly important role. 2

That shift matters because it changes where the financial burden sits and when it is recognized. Leasing or partner-led structures may spread investment costs over time, but they do not make the underlying infrastructure obligation disappear. For investors and creditors, the key issue is whether future AI-related cash generation can cover both the infrastructure base and the financing commitments used to create it.

More intense than earlier tech cycles

S&P Global says AI investment intensity, measured as capital expenditure relative to depreciation, has exceeded peaks seen in the dot-com era and around the Great Recession. The build-out is also unusually concentrated in a small group of companies. 1

That does not automatically make the current cycle a repeat of the dot-com bubble. The leading hyperscalers are mature companies with large existing businesses in cloud services, advertising, software, commerce and subscriptions. S&P’s point is that the sector has not yet developed the financing characteristics that made the dot-com build-out fragile. 1

Still, maturity does not eliminate risk. A profitable company can become financially stretched if capital commitments rise faster than operating cash flow for long enough. The most important variables are future utilization of AI capacity, pricing, revenue growth and eventual returns on invested capital.

Oracle shows how capex can become a credit issue

Oracle has become a prominent example of the credit-market stakes. S&P Global Ratings downgraded Oracle to BBB- from BBB in 2026, according to the ratings action’s title and S&P’s related materials. 1720

The broader lesson is not that every hyperscaler will face the same outcome. Their business mixes, balance sheets and funding options differ. It is that an aggressive AI infrastructure strategy can affect credit metrics even for an established technology company with meaningful operating businesses.

The question that determines whether the build-out pays off

S&P Global’s analysis is conditional, not a declaration that AI investment will fail. The outlays may prove justified if AI services generate enough additional demand, revenue and margins—and do so soon enough.

The decisive question is whether returns from AI infrastructure can catch up with the capital being deployed before debt and other external financing become a structurally dominant source of funding. If monetization accelerates, today’s capacity could become a durable advantage. If it arrives slowly, the Hyper 5 could face a longer period of negative free cash flow, more complex financing and greater pressure on credit quality. 12