NextFin News - Cantor Fitzgerald Co-CEO Christian Wall warned on Friday that the artificial-intelligence infrastructure build-out will drive "trillions of dollars" of new debt issuance, a call that lands as the bond market is already straining to absorb a record flood of technology borrowing. The warning arrives with the 10-year Treasury yield near 4.84% and the 30-year yield above 5.29%, levels that make the cost of funding an AI arms race more expensive by the quarter.
The stakes are straightforward but easily understated: the five largest builders of AI data centers have added roughly $350 billion to their debt obligations over the past five years, and they issued $121 billion of U.S. corporate bonds in 2025 alone - more than four times the $28 billion annual average from 2020 through 2024. If Wall is right that the next leg of the build-out is measured in trillions rather than hundreds of billions, the question is no longer whether tech companies can raise the money. It is what the bond market will demand in return, and who else gets crowded out of the way.
The Funding Regime Has Flipped
The most important fact about the AI debt boom is not how much has been borrowed. It is who is doing the borrowing, and how. For years, the largest technology companies defended their AI spending as equity risk: speculative, funded out of cash flow, and therefore not a credit concern. That compact is gone.
Through July 7, Amazon, Alphabet, Meta and Oracle had issued about $194 billion of bonds in 2026, up 79% from the roughly $108 billion they sold in all of 2025. The median spread on their two- to four-year bonds widened to 40 basis points from 30 basis points a year earlier, and 78 of 91 hyperscaler bonds issued this year with comparable pricing data were trading at higher yields in late July than at issuance. The market is not refusing to lend. It is repricing the loan.
Wall's "trillions" framing should be read against that backdrop. The Big Five's borrowing is accelerating from a roughly $140 billion annual run rate that Bank of America analysts expect to hold over the next three years - with a ceiling that could exceed $300 billion a year if capex plans keep climbing. Overall U.S. corporate bond issuance is forecast to reach $2.46 trillion in 2026, up 11.8% from 2025, according to a January note from Barclays, while Morgan Stanley's global co-head of investment-grade debt has said issuance could push above $2 trillion for the first time this year. Debt at that scale is not a rounding error in the credit market. It is a new source of duration supply that did not exist five years ago.
"For years, we've been told this AI spend would be funded by generated cash flow - that it is equity risk, it is speculative, and not to worry about it from a credit point of view. There now seems to be a change in the unspoken contract."
That assessment came from a UBS credit strategist in February, but it describes the mechanism Wall is betting on: the moment AI capex moves onto the balance sheet as debt, credit investors become the gatekeepers of the build-out's pace.
The Transmission Channel: Duration Supply Meets a Crowded Rates Market
Why does more tech borrowing matter to anyone who does not own a tech bond? Because the AI build-out is not borrowing in a vacuum. It is borrowing into a market that is already clearing against a $40 trillion national debt, a federal deficit on track for $2 trillion this fiscal year, and roughly $1 trillion a year in debt-servicing costs. Every dollar of private capital that flows into a hyperscaler bond is a dollar not flowing into a Treasury.
Ed Yardeni, the Wall Street veteran who has tracked the phenomenon most closely, put the mechanism plainly: U.S. investment-grade corporate issuance totaled about $1.7 trillion in the year to date through July, roughly 27% above last year's pace and on track to exceed $2 trillion for the first time. Because AI-related demand has been so strong that corporate spreads stayed compressed, "the market has adjusted not through higher corporate borrowing costs relative to Treasuries but through higher Treasury yields themselves," he wrote. "In short, the AI revolution is producing a classic crowding-out effect, causing Treasury yields to rise."
That is the second-order effect most investors are still underweighting. The conventional read of the AI debt story is first-order and company-specific: more borrowing means higher interest expense, which pressures margins if AI revenue does not arrive on schedule. The more consequential read is cross-market. If corporate issuance absorbs risk capital that would otherwise sit in government bonds, the clearing price for duration rises across the curve - and the 10-year Treasury yield, the benchmark for everything from mortgages to corporate discount rates, becomes the transmission belt.
The 10-year note has already climbed 61 basis points this year, according to Goldman Sachs, which noted that AI-related issuers now account for about 24% of U.S. dollar investment-grade gross supply. A 61-basis-point move in the risk-free rate does more to destroy equity value through a higher discount rate than a 10-basis-point widening in a tech credit spread does to hurt a bondholder. The bond market is not just financing the AI build-out. It is setting the price at which the build-out must eventually earn its keep.
The discount-rate channel cuts both ways, and it cuts the hyperscalers first. A higher 10-year yield lowers the present value of the very AI cash flows that justify today's capex. So the companies racing to borrow are, through the market's pricing of that borrowing, raising the hurdle rate their own projects must clear. That feedback loop is the quiet asymmetry in the build-out: the more debt the sector issues, the harder its investments become to justify on a discounted-cash-flow basis.
The pressure is also global. Hyperscalers have begun issuing bonds in euros, Canadian dollars and Asian currencies to tap a wider investor pool and avoid saturating the U.S. market with colossal volumes of supply. Alphabet's finance chief has said the company has accumulated $100 billion in outstanding debt across six major currencies, and Morgan Stanley's global co-head of investment-grade debt has pointed to Alphabet and Amazon diversifying into European, Canadian and Asian markets. The globalization of AI funding is not a footnote. It is evidence that the U.S. market alone cannot absorb the issuance at the price borrowers want.
Is This Cyclical or Structural? The Regime-Shift Call
Here is the judgment that separates a credit-cycle story from a regime-change story. A cyclical debt boom is mean-reverting: spreads widen, issuance slows, balance sheets repair, and the cycle resets. A structural shift does not revert on its own because the underlying funding model has changed.
Three pieces of evidence point to structural. First, the funding source has permanently shifted from cash flow and equity to debt - a balance-sheet decision, not a sentiment swing. Second, the borrower base has broadened beyond the hyperscalers: data-center developers, private-credit funds, and even chipmakers are layering debt onto AI projects, creating a financing ecosystem that will persist as long as the build-out does. Third, the demand side is yield-agnostic by design. Treasury Secretary Scott Bessent observed recently that a lot of corporate issuance is "almost yield-agnostic, because the build-out for AI, the returns on that, the companies believe they're going to be so high. They don't really care what they're paying." When borrowers do not care about price, the market clears through quantity and through the price of everything else - namely, Treasury yields.
Federal Reserve Chairman Kevin Warsh nodded to the same dynamic at Jackson Hole, noting that "ever-expanding pools of capital are pouring into AI-related infrastructure of all sorts." A central banker flagging private capital pools as a rates-market factor is itself evidence that the phenomenon has moved from a sector story to a macro one.
But the structural call has a boundary. If AI revenue begins to cover capex and debt service within the next two to three years, the borrowing need peaks and the debt stock stabilizes - a long cycle, but a cycle nonetheless. If returns lag, the debt remains on the books and rolls over at whatever the market demands. The structural case rests on the assumption that the build-out is a multi-trillion-dollar, multi-decade infrastructure project, not a five-year capex spike.
The Strongest Case Against the Trillions Call
The counter-thesis is not that AI borrowing is small. It is that it is not the dominant force in the bond market, and that blaming Treasury yields on AI debt confuses correlation with causation. Industry data compiled by SIFMA show the high-technology sector accounted for only 12.8% of issuance in 2026 so far, well behind financials, which held a 45.2% share. On that read, the bond market's indigestion has more to do with federal deficits, oil-price shocks, and inflation than with data-center financing.
That objection is serious and should not be waved away. A 12.8% share does not, by itself, justify "trillions" of incremental debt driving rates. But it also understates the marginal effect. The relevant question is not what share of total issuance AI represents today; it is what share of the increment it represents, and whether that increment lands in a market with limited risk-capital capacity. Goldman's 24% figure for AI-related share of USD IG gross supply points to the marginal reality: nearly a quarter of new investment-grade supply this year is AI-linked. In a market where foreign private buyers are purchasing more corporate bonds than Treasuries, the marginal dollar sets the price.
The falsifying signal is specific. If the share of AI-related supply in U.S. dollar investment-grade issuance falls back below 15% while total hyperscaler borrowing stays above $150 billion a year - meaning the market absorbs the debt without the AI share rising - the structural-crowding thesis weakens materially. Equally, if the 10-year Treasury yield trades back below 4% while issuance accelerates, the "AI drives rates" channel is not operating as described.
What to Watch and Who Is Exposed
The near-term path is a function of three signals. First, the hyperscaler spread: a move back toward the 30-basis-point median of 2025 would signal that investor appetite has recovered; a push toward 50 basis points would signal fatigue setting in. Second, the 10-year Treasury yield: a break above 5% would raise the discount rate on every AI project's future cash flows and could force a capex reassessment. Third, the refinancing wall: bonds issued in 2025 and early 2026 at compressed spreads will need to roll, and the price at which they roll is the market's verdict on whether the "yield-agnostic" era is ending.
By time horizon, the implications split. In the short term, sentiment and liquidity dominate: as long as spreads stay compressed and the Federal Reserve holds rates, issuance can continue at a blistering pace. In the medium term, fundamentals take over - AI revenue must begin to cover debt service, or credit investors will demand a larger premium and slow the build-out through price. In the long term, the structural question is whether the bond market has become a permanent co-financier of the AI economy, which would embed a higher term premium into the rates complex for years.
The base case is that issuance stays elevated - somewhere between $200 billion and $300 billion a year for the Big Five - while spreads grind wider in fits and starts. The upside case for the bulls is that AI revenue inflects faster than expected, debt service stays covered, and the market absorbs the supply with only modest premium demands. The downside case is a reflexive loop: higher Treasury yields raise debt-servicing costs, deficits widen, yields rise further, and the bond market forces a capex slowdown before the AI payoff arrives.
The beneficiaries are the intermediaries who underwrite and distribute the debt - investment banks, private-credit funds, and the insurers and asset managers who buy it. The exposed are everyone holding duration: Treasury investors facing a higher clearing yield, corporate borrowers outside tech who will pay more to refinance, and equity holders whose valuations depend on a low discount rate. The irony is that the companies racing to build AI infrastructure are, in the process, making the cost of capital higher for the very projects they are trying to finance.
Wall's trillions warning is not a prediction that the bond market will break. It is a prediction that the bond market will be the arena where the AI build-out is decided - and that the price of admission will be paid in higher yields, wider spreads, and a permanently larger role for debt in funding the future of technology. The market has already begun to collect that price. The only question is how much more it will charge.
Explore more exclusive insights at nextfin.ai.
