The Scale of the Spending

If Oracle and CoreWeave are included, Moody's puts total 2026 hyperscaler capex at $785 billion, heading toward ~$1 trillion in 2027 . Goldman Sachs projects cumulative capex of $5.3 trillion for the four largest hyperscalers from FY2025 to FY2030 .

How It's Being Funded: Debt and Off-Balance-Sheet Financing

Even the world's most cash-rich companies are shifting toward borrowing and off-balance-sheet structures:

• Direct bond issuance is accelerating. Amazon raised $25 billion in new bonds in mid-2026, its pledged final issuance for the year, to fund contracted AI capacity for OpenAI and Anthropic . Moody's reports that direct debt across the six tracked hyperscalers has reached roughly $460 billion .

• Lease commitments have ballooned. Moody's estimates total lease obligations have swelled to $1.2 trillion, with more than $820 billion for data centers not yet operating—effectively creating a massive off-balance-sheet financing channel that converts capital expenditures into operating expenses but concentrates long-term payment risk .

• Asset-heavy structural shift. Moody's noted that the industry is undergoing a fundamental transition from "asset-light software businesses" to "capital-intensive AI infrastructure"—a shift that inherently weakens credit profiles even for companies with pristine balance sheets .

Moody's Credit Warning

In a July 22, 2026, research note, Moody's Ratings issued a stark warning :

• The AI buildout is "eroding free cash flow and raising balance-sheet risk" at Microsoft, Amazon, Alphabet, Meta, Oracle, and CoreWeave.
• Moody's highlighted stock sales and off-balance-sheet moves that "threaten credit quality" .
• The ratings agency projects that five tech giants could outspend operating cash flow entirely by 2027—Oracle's fiscal 2026 capex already reached 174% of operating cash flow .
• The systemic concern is concentration risk: nearly all new data center capacity is pre-leased to the same small group of hyperscaler tenants, creating a self-reinforcing cycle where the builders and the customers are largely the same entities .

The Conflicting Signals: Strong Revenue vs. Collapsing Free Cash Flow

Strong top-line growth:

• Meta recorded 28% revenue growth in the June quarter .
• AWS demand is contractually locked in: Anthropic and OpenAI signed Trainium capacity commitments, with AWS sold out through 2027 .
• Analysts broadly agree that enterprise AI demand across industries justifies the long-term thesis .

But free cash flow is collapsing:

• The FT reported that Big Tech's combined AI spending spree sent free cash flow to a decade low in 2026 .
• Alphabet's FCF turned negative in Q2 2026 for the first time since its IPO, after $44.9 billion in quarterly capex—shares plunged 7%+ on the news .
• Amazon's FCF went to -$7.6 billion as capex consumed 138% of first-half operating cash flow .
• Meta's cash generation fell 91% in Q2 .
• A Reuters analysis using LSEG data warns that at current trajectories, five hyperscalers could outspend their operating cash flow entirely by 2027 .
• Bank of America forecasts that aggregate free cash flow across eight major companies will swing from an estimated $180 billion in 2025 to roughly -$64 billion in 2026 .

Microsoft's capex-to-revenue efficiency gap is emblematic: capex growth averaged ~62% over 2024–2026, while Azure revenue growth was only ~37%—a widening divergence that markets are starting to penalize .

Market Reactions

• Amazon shares fell 11.5% in after-hours trading in February 2026 when it first guided to $200 billion in capex .
• Alphabet shares plunged 7%+ in July after its FCF turned negative and capex guidance was raised .
• The broader market has revolted: Fortune characterized Big Tech's July 2026 earnings season as a "market in revolt over AI spending" . CNBC reported that Amazon, Alphabet, and Tesla all posted negative cash flow in the same quarter, while Meta's cash generation collapsed .
• Investor focus has shifted from "profitability boost" to "revenue growth trajectory and cash-flow visibility" .

The Central Tension: Will the Returns Justify the Investment?

This is the defining question of the current cycle, and the evidence points to an unresolved gamble:

The bullish case:

• AI demand is real and contractually locked in. AWS is sold out through 2027 on Trainium commitments from OpenAI and Anthropic .
• Analysts interviewed by Valor say strong AI demand across industries "justifies the jump in capex" and that returns should begin to emerge by 2029 .
• The companies see this as a once-in-a-generation infrastructure buildout—analogous to the early cloud or railroad eras—where late entrants risk permanent exclusion.

The bearish case:

• Moody's explicitly questions whether "expenditures will ever be matched by the spending that companies and individuals will do to use generative AI" .
• The capex-to-revenue efficiency ratio is deteriorating—Microsoft's 62% capex growth producing only 37% Azure revenue growth is a structural warning .
• The self-referential risk: the biggest customers of this infrastructure (OpenAI, Anthropic) are themselves backed by the same hyperscalers building the capacity, raising questions about whether end-user demand is as strong as contracted demand suggests.
• The baseline aggregate capex estimate stands at $7.6 trillion between 2026 and 2031 across compute, data centers, and power—a level of spending that would require AI revenue to grow at multiples of current cloud revenue trajectories to generate adequate returns .

The central tension is between first-mover necessity (you cannot win the AI era without building the infrastructure now) and financial sustainability (borrowing hundreds of billions on balance sheets and off them, with free cash flow turning negative, while the monetization timeline remains uncertain and pushed out to 2029 or later). Moody's warning that this threatens credit quality at the world's most cash-rich companies is the most authoritative summary of the risk: the industry is in uncharted territory where capital intensity has structurally altered the risk profile of what were once pristine software-and-services businesses .

The Scale of the Spending

If Oracle and CoreWeave are included, Moody's puts total 2026 hyperscaler capex at $785 billion, heading toward ~$1 trillion in 2027 . Goldman Sachs projects cumulative capex of $5.3 trillion for the four largest hyperscalers from FY2025 to FY2030 .

How It's Being Funded: Debt and Off-Balance-Sheet Financing

Even the world's most cash-rich companies are shifting toward borrowing and off-balance-sheet structures:

• Direct bond issuance is accelerating. Amazon raised $25 billion in new bonds in mid-2026, its pledged final issuance for the year, to fund contracted AI capacity for OpenAI and Anthropic . Moody's reports that direct debt across the six tracked hyperscalers has reached roughly $460 billion .

• Lease commitments have ballooned. Moody's estimates total lease obligations have swelled to $1.2 trillion, with more than $820 billion for data centers not yet operating—effectively creating a massive off-balance-sheet financing channel that converts capital expenditures into operating expenses but concentrates long-term payment risk .

• Asset-heavy structural shift. Moody's noted that the industry is undergoing a fundamental transition from "asset-light software businesses" to "capital-intensive AI infrastructure"—a shift that inherently weakens credit profiles even for companies with pristine balance sheets .

Moody's Credit Warning

In a July 22, 2026, research note, Moody's Ratings issued a stark warning :

• The AI buildout is "eroding free cash flow and raising balance-sheet risk" at Microsoft, Amazon, Alphabet, Meta, Oracle, and CoreWeave.
• Moody's highlighted stock sales and off-balance-sheet moves that "threaten credit quality" .
• The ratings agency projects that five tech giants could outspend operating cash flow entirely by 2027—Oracle's fiscal 2026 capex already reached 174% of operating cash flow .
• The systemic concern is concentration risk: nearly all new data center capacity is pre-leased to the same small group of hyperscaler tenants, creating a self-reinforcing cycle where the builders and the customers are largely the same entities .

The Conflicting Signals: Strong Revenue vs. Collapsing Free Cash Flow

Strong top-line growth:

• Meta recorded 28% revenue growth in the June quarter .
• AWS demand is contractually locked in: Anthropic and OpenAI signed Trainium capacity commitments, with AWS sold out through 2027 .
• Analysts broadly agree that enterprise AI demand across industries justifies the long-term thesis .

But free cash flow is collapsing:

• The FT reported that Big Tech's combined AI spending spree sent free cash flow to a decade low in 2026 .
• Alphabet's FCF turned negative in Q2 2026 for the first time since its IPO, after $44.9 billion in quarterly capex—shares plunged 7%+ on the news .
• Amazon's FCF went to -$7.6 billion as capex consumed 138% of first-half operating cash flow .
• Meta's cash generation fell 91% in Q2 .
• A Reuters analysis using LSEG data warns that at current trajectories, five hyperscalers could outspend their operating cash flow entirely by 2027 .
• Bank of America forecasts that aggregate free cash flow across eight major companies will swing from an estimated $180 billion in 2025 to roughly -$64 billion in 2026 .

Microsoft's capex-to-revenue efficiency gap is emblematic: capex growth averaged ~62% over 2024–2026, while Azure revenue growth was only ~37%—a widening divergence that markets are starting to penalize .

Market Reactions

• Amazon shares fell 11.5% in after-hours trading in February 2026 when it first guided to $200 billion in capex .
• Alphabet shares plunged 7%+ in July after its FCF turned negative and capex guidance was raised .
• The broader market has revolted: Fortune characterized Big Tech's July 2026 earnings season as a "market in revolt over AI spending" . CNBC reported that Amazon, Alphabet, and Tesla all posted negative cash flow in the same quarter, while Meta's cash generation collapsed .
• Investor focus has shifted from "profitability boost" to "revenue growth trajectory and cash-flow visibility" .

The Central Tension: Will the Returns Justify the Investment?

This is the defining question of the current cycle, and the evidence points to an unresolved gamble:

The bullish case:

• AI demand is real and contractually locked in. AWS is sold out through 2027 on Trainium commitments from OpenAI and Anthropic .
• Analysts interviewed by Valor say strong AI demand across industries "justifies the jump in capex" and that returns should begin to emerge by 2029 .
• The companies see this as a once-in-a-generation infrastructure buildout—analogous to the early cloud or railroad eras—where late entrants risk permanent exclusion.

The bearish case:

• Moody's explicitly questions whether "expenditures will ever be matched by the spending that companies and individuals will do to use generative AI" .
• The capex-to-revenue efficiency ratio is deteriorating—Microsoft's 62% capex growth producing only 37% Azure revenue growth is a structural warning .
• The self-referential risk: the biggest customers of this infrastructure (OpenAI, Anthropic) are themselves backed by the same hyperscalers building the capacity, raising questions about whether end-user demand is as strong as contracted demand suggests.
• The baseline aggregate capex estimate stands at $7.6 trillion between 2026 and 2031 across compute, data centers, and power—a level of spending that would require AI revenue to grow at multiples of current cloud revenue trajectories to generate adequate returns .

The central tension is between first-mover necessity (you cannot win the AI era without building the infrastructure now) and financial sustainability (borrowing hundreds of billions on balance sheets and off them, with free cash flow turning negative, while the monetization timeline remains uncertain and pushed out to 2029 or later). Moody's warning that this threatens credit quality at the world's most cash-rich companies is the most authoritative summary of the risk: the industry is in uncharted territory where capital intensity has structurally altered the risk profile of what were once pristine software-and-services businesses .