Big Tech's defining financial advantage — abundant cash flow — is colliding with the physical demands of generative AI. Alphabet Inc. (GOOG), Amazon.com, Inc. (AMZN), Meta Platforms, Inc. (META) and Oracle Corp. (ORCL) continue to post strong revenue growth and operating cash flow. Yet their forward free cash flow is set to fall toward zero, or even turn negative, as data-center spending accelerates.
Capital expenditures across the sector could reach $785 billion in 2026 and approach $1 trillion next year, according to Moody's Ratings. The result is a fundamental change in financing. Hyperscalers could issue about $175 billion in debt this year, compared with a five-year average of around $30 billion.
"Investors are underestimating how fundamentally AI is changing the Big Tech business model," said Shay Boloor, chief market strategist at Futurum Equities. "These companies were historically valued as asset-light platforms because revenue could scale much faster than capital requirements, but AI is pushing them toward a hybrid model where software, advertising and cloud economics increasingly depend on enormous physical infrastructure spending."
Bond Market Raises The Ante
Bondholders are already signaling they want more proof. Credit spreads on leading hyperscaler debt have widened to about 78 basis points over Treasuries from 50 basis points two months earlier — the sharpest move since tariff-related volatility in April 2025. The widening might be modest by historical standards, but it marks a break from the near-automatic confidence that accompanied tech mega-cap issuance.
Oracle has become the clearest stress case. The company's credit-default swap spread widened above 125 basis points, a level not seen since the 2009 financial crisis, after its AI data-center expansion pushed free cash flow negative. Its 5.20% notes due 2035, issued in 2025 at a 105-basis-point spread, later widened as much as 175 basis points, pushing the bonds' price lower.
Amazon also encountered more demanding buyers in a $25 billion bond sale in July. The company offered an additional 18 to 21 basis points of yield on its longest maturities, while orders totaled about 2.5 times the amount offered, down from 3.2 times in March.
Apollo's Chief Economist Torsten Slok warned that coverage of new hyperscaler deals slipped to twice the size of offerings by July, down from roughly five times in February, signaling waning interest.
Revenues Up, Share Prices Down
Equity investors are delivering a similar message. Alphabet reported second-quarter revenue of $119.8 billion, up 24% from a year earlier, while cloud sales rose 82%. But shares fell after the company raised its 2026 capital expenditure guidance to $195 billion to $205 billion and warned of elevated spending in 2027.
"Earnings growth may not be enough to justify investment if capex is depleting cash," said David Russell, global head of market strategy at TradeStation. "Companies exist to make money, not spend money."
From Software To Industrial Infrastructure
The deeper concern is structural. Moody's noted Silicon Valley is shifting from "asset-light to asset-heavy models," requiring data centers, power-hungry servers and long-lived infrastructure rather than easily scalable software and cloud platforms.
Direct debt across six companies tracked by Moody's — including Microsoft Corp. (MSFT), Amazon, Alphabet, Meta, Oracle and CoreWeave, Inc. (CRWV) — has reached about $460 billion.
The larger shadow sits off-balance-sheet. Data-center lease commitments total about $1.2 trillion, including more than $820 billion tied to projects not yet operating. Rating agencies treat such obligations as debt-equivalent because they bind companies to future rent payments regardless of the pace of AI monetization.
Moody's warned that the move "from asset-light to asset-heavy models requires unprecedented levels of investment and capital raising" and "threatens credit quality."
Microsoft and Alphabet retain substantial balance-sheet flexibility, while Oracle, rated Baa2 with a negative outlook, and Ba3-rated CoreWeave face greater vulnerability.
Still, as lease obligations grow and AI labs such as OpenAI and Anthropic become both investment recipients and major cloud customers, creditors may increasingly value Big Tech less like unconstrained software platforms and more like industrial infrastructure utilities.
