A number that's already broken last year's record, with months to go

Goldman Sachs strategist Amanda Lynam put a figure on the technology sector's borrowing spree this week that stopped a lot of market watchers mid-scroll: $489 billion in AI-related debt has been issued so far in 2026, according to her research note published Wednesday, July 22. That's already well above the $322 billion Goldman estimated for the entirety of 2025 — meaning this year's AI debt issuance surpassed last year's full-year total with more than five months still remaining on the calendar.

Roughly 40% of that 2026 figure has come directly from the hyperscalers themselves — Amazon, Microsoft, Alphabet, and Meta, the four companies whose cloud computing platforms and AI infrastructure investments have driven much of this cycle's spending. Amazon alone has raised approximately $53 billion in debt this year, a total that includes a $37 billion U.S. bond offering plus more than $16 billion in euro-denominated bonds. Alphabet has issued roughly $20 billion in bonds, while Oracle has raised about $25 billion specifically to fund what the company has described as its aggressive AI expansion.

Why companies that used to sit on cash piles are suddenly borrowing this heavily

For most of the past decade, Amazon, Alphabet, and Microsoft were defined in financial circles by conservative balance sheets and enormous cash reserves — companies that rarely needed to tap debt markets at scale because their core businesses generated so much free cash flow internally. That era has ended abruptly. Goldman Sachs data shows hyperscaler capital intensity — the share of revenue being plowed directly into capital expenditures — has reached between 45% and 57% of total revenue, a level one analysis described as "historically unthinkable" for companies of this size and maturity.

The scale of planned spending explains why internal cash flow alone can no longer cover it. Amazon is projecting roughly $200 billion in capital expenditures for 2026, while Microsoft is tracking toward approximately $190 billion, driven heavily by scaling Azure's AI capacity and the company's Stargate supercomputer project. Alphabet has guided to $175 billion to $185 billion, aimed substantially at expanding its custom TPU chip production and Gemini model infrastructure, while Meta has guided toward $115 billion to $135 billion. Combined, Goldman now projects the four largest hyperscalers will spend a staggering $5.3 trillion on capital expenditures between fiscal 2025 and fiscal 2030 — a figure that has itself climbed from an earlier Goldman estimate of $4.5 trillion made just months prior, before this year's first-quarter earnings reports came in.

The market is starting to show real signs of indigestion

What's changed most significantly over just the past several weeks isn't simply the size of the borrowing — it's how bond markets are responding to absorbing it. Multiple reports this month have described Goldman Sachs's own internal trading desk raising alarm about the pace of issuance specifically, with one Goldman investment-grade bond trader describing $75 billion in AI-related bond issuance over a single recent month as leaving the market "struggling to digest." That's a notably blunt characterization coming from inside the same bank whose strategists have spent months documenting the broader trend somewhat more clinically.

The market's absorption capacity appears to be shrinking in a measurable way. According to reporting that cited Goldman's internal assessment, where roughly $75 billion in new supply was previously needed to visibly stress the bond market, it now takes just $25 billion to put the market on the defensive — a threefold reduction in how much new debt the market can absorb comfortably before showing strain. Leverage ratios among these companies have reportedly surged from 0.9 times to 1.8 times in just two fiscal quarters, a pace of balance-sheet change that has reportedly already surpassed leverage levels in the traditionally capital-intensive energy sector.

Widening spreads and a market that's starting to demand a premium

Concrete signs of that strain have shown up directly in how specific bonds are pricing. Reporting has noted that spreads on new bond offerings from both Amazon and SpaceX have widened significantly in the secondary market, while subscription ratios — a measure of how enthusiastically investors are bidding for a new bond offering relative to the amount being sold — have declined noticeably compared to earlier in the AI borrowing cycle. Widening spreads and falling subscription ratios together typically signal that investors are demanding higher compensation to keep absorbing this volume of debt, rather than accepting the same favorable terms hyperscalers could command when this borrowing cycle first began.

That's a meaningfully different market environment than existed even six months earlier, when AI-related bonds from blue-chip technology names were generally treated as premium, high-demand issuance regardless of size. The shift suggests bond investors are beginning to price in genuine credit risk associated with the sheer scale and pace of this borrowing, rather than treating it as a routine extension of already-strong corporate balance sheets.

Not just bonds: a broader financing ecosystem straining at every level

The debt story extends well beyond direct corporate bond issuance from the hyperscalers themselves. Since early 2025, infrastructure funds, real estate investors, private equity firms, and private credit vehicles have collectively provided more than $140 billion in financing specifically for data center transactions — a figure that excludes chip-related investment entirely and captures only the physical infrastructure financing layer of this broader buildout. Goldman Sachs projects that global digital infrastructure assets broadly will surpass $3 trillion by 2030, giving some sense of how much additional capital is expected to flow into this space beyond what shows up in headline corporate bond figures.

Multi-currency issuance has also become a notable feature of this cycle, with major hyperscalers issuing roughly $194 billion in bonds across various currencies during 2026, including substantial euro, British pound, Japanese yen, Swiss franc, and Canadian dollar-denominated debt. That international diversification likely reflects, at least in part, an effort to tap additional pools of global capital as the domestic U.S. dollar investment-grade market shows signs of struggling to absorb the full scale of hyperscaler borrowing on its own.

The risk scenario analysts are now taking seriously

Bank of America Chief Investment Officer Michael Hartnett has specifically flagged "AI capex cuts" as what he considers the single biggest tail risk currently facing broader markets, laying out a specific chain of events that could trigger it: bond market investors — sometimes described as "bond vigilantes" — cut off liquidity supply to hyperscalers by refusing to absorb further debt on favorable terms, forcing those companies to turn instead to equity financing or workforce reductions to fund their continued AI infrastructure commitments. Notably, that scenario isn't purely hypothetical at this point — Meta, Microsoft, and Amazon have already conducted workforce reductions of 13%, 10%, and 9% respectively, according to figures cited in coverage of Hartnett's warning, even as their capital expenditure commitments have continued climbing.

That combination — rising capex alongside falling headcount — is precisely the kind of internal contradiction that makes credit analysts nervous. A company simultaneously cutting staff while dramatically increasing capital spending is, in effect, signaling that its AI infrastructure investments are being prioritized above nearly every other cost center, a stance that only remains sustainable if the underlying AI business eventually generates returns proportional to the capital being deployed against it.

Why this week's Big Tech earnings matter more than usual

This debt story is landing at a particularly consequential moment because several of the companies at its center are reporting earnings this week, with Alphabet and Tesla both scheduled Wednesday and Amazon set to report July 30. Markets are watching these reports for a specific signal beyond the usual revenue and profit figures: whether AI-related spending is translating into revenue growth that can plausibly justify the scale of debt being taken on to fund it. Goldman Sachs data indicates AI infrastructure-related stocks now account for roughly 42% of the S&P 500's total market capitalization and are expected to contribute about half of the index's earnings growth in 2026 — meaning the financial health of this specific borrowing cycle carries stakes that extend well beyond the individual companies involved, into the performance of the broader stock market these companies now disproportionately represent.

Whether this AI debt boom ultimately resolves as a historically large but manageable investment cycle, or as the kind of overextended borrowing spree that eventually forces a painful correction across both credit and equity markets, remains genuinely unresolved. What's no longer in question, based on Goldman's own internal trading desk commentary and the visible widening in bond spreads over recent weeks, is that the market's capacity to absorb this debt at favorable terms is being tested more directly right now than at any prior point in this particular AI investment cycle.

*This article was researched using publicly available reporting from Goldman Sachs research notes, Yahoo Finance, BigGo Finance, 24/7 Wall St., and Introl's coverage of hyperscaler AI infrastructure debt issuance in 2026. It is intended for informational purposes and does not constitute financial advice.*