NextFin News - For a decade, the risk hierarchy of global credit was simple: investment-grade technology debt sat near the top, and emerging-market bonds sat near the bottom. In 2026, that order is being inverted. The artificial-intelligence buildout is forcing the largest borrowing wave in corporate history through the bond market, and the market is starting to price the borrowers funding it as the riskier corner of the credit universe — while developing-nation governments, long shunned after a decade of crises, are drawing debt inflows at a more than two-decade high. The clearest signal that the AI boom has entered a new phase is no longer a chip launch or a model release. It is the price of risk itself, and it is moving in the wrong direction for the companies that promised to be the safest borrowers in the world.
The Two Sides of the Same Trade
On one side of the trade is the scale of the AI financing wave. Morgan Stanley expects global AI-related debt issuance to nearly double to about $570 billion in 2026, with roughly $236 billion already priced by the end of May — four times the pace of a year earlier. The four largest hyperscalers — Alphabet, Amazon, Microsoft and Meta Platforms — are on track to spend $700 billion on capital outlays this year, a jump of more than 75% that ratings agency Fitch says has become a dominant force in credit markets. U.S. corporate bond issuance surged 26% in the first half of 2026, driven largely by AI fundraising, and Fitch estimates booming IT investment directly added 1.4 percentage points to first-quarter U.S. gross domestic product growth.
The borrowing is concentrated enough to change the shape of the market. Alphabet, Amazon, Meta, Microsoft and Oracle sold roughly $121 billion of U.S. corporate bonds in 2025, more than four times their average of about $28 billion a year between 2020 and 2024, according to a Bank of America Securities analysis. Alphabet alone raised $17.5 billion and Amazon $15 billion in November, and hyperscalers accounted for four of the five biggest U.S. investment-grade bond deals of the year. Taken together, the group borrowed about $93 billion in the U.S. investment-grade market in 2025, roughly 6% of all debt issued.
On the other side of the trade are the borrowers that spent the last decade in the doghouse. War, tariffs and AI gyrations have done little to dent the flow of cash into emerging markets this year, with developing-nation debt inflows running at their strongest level in more than 20 years and governments issuing record amounts of bonds. Portfolio flows into emerging-market assets rebounded to $58.3 billion in April after a $66.2 billion exodus in March, and emerging-market debt alone drew $51.9 billion in April. "Roughly from 2015 to 2025 was like the valley of tears for emerging markets: strong dollar, U.S. exceptionalism, lots of crises, defaults, COVID," said David Hauner, head of emerging-markets fixed-income strategy at Bank of America. "There was so much outflow, a few months of inflows are not going to compensate for that... it's still only scratching the surface of the under-investment that has occurred over the past decade."
Put the two sides together and the inversion is stark. The assets that credit committees once treated as the anchor of safety are being asked to absorb the largest, fastest and most opaque borrowing wave in the market's memory. The assets that were written off as too risky are being rediscovered as the diversified, transparent alternative. The risk hierarchy is not just shifting at the margin — it is being rewritten.
Why the Market Is Starting to Push Back
The first sign of fatigue is in demand. Apollo reported that orders for hyperscaler bonds covered nearly five times the amount offered in February; by July, coverage had slipped below two times. That does not mean the market is breaking, but it does mean issuers are likely to pay more. Credit spreads on Google, Amazon and Meta have widened as fixed-income investors demand greater reward for lending to the companies, and Oracle's five-year credit-default swaps have become the market's go-to proxy for AI debt fears, trading near an 18-year high.
Oracle is the canary. S&P Global cut the company's credit rating to BBB-, one notch above junk, after its fiscal 2026 results showed negative free cash flow of $23.7 billion and roughly $260 billion of signed data-center leases. S&P analyst Andrew Chang projects Oracle's debt-to-EBITDA, which he estimated at 3.6 in May, will peak at 4.4 over the coming two fiscal years — just below the 4.5 threshold that would trigger another downgrade. The company plans to raise $40 billion through debt and equity financing, including a $20 billion share sale, to fund the buildout, a move that Srini Ramaswamy, a senior financial economist at the Federal Reserve Bank of Dallas, described as turning Oracle into one of the largest suppliers of duration risk in the investment-grade universe.
"The scale of AI investment is such that the exposure of the economy and overall capital market to such a correction is significant," Fitch said in its third-quarter Global Risk Outlook, the bluntest warning yet from a major ratings firm.
The concern has spread beyond a single name. Fidelity told clients in its midyear outlook, published June 3, that it is steering its core bond portfolios away from the wave of new debt that Amazon, Alphabet, Meta, Microsoft and Oracle are selling to pay for data centers, chips and the power to run them. S&P Global warned in July that hyperscalers are paying a higher premium relative to risk-free yields after absorbing the flood of debt in a short time. And analysts note the hyperscalers are on track to collectively spend more on capital expenditure than they generate in free cash flow by next year — a first for companies long regarded as cash machines.
The Risk Has Moved Off the Balance Sheet
The deeper reason the risk hierarchy is being rewritten is that the risk is no longer where investors are looking. Public bonds are transparent and easy to price; much of the AI buildout is migrating into structures that are neither. Morgan Stanley projects roughly an $800 billion private-credit opportunity in data-center financing through 2028, and much of that sits in off-balance-sheet vehicles that Bank for International Settlements officials call "shadow borrowing." Meta's $27 billion Hyperion joint venture illustrates the architecture: funds managed by Blue Owl own 80%, Meta holds 20%, and some of Blue Owl's funding comes from debt sold to PIMCO and other investors. The structure keeps much of the project debt off Meta's balance sheet.
Similar financing is spreading through securitized deals and project-finance loans, and utilities are borrowing more to serve data-center demand — U.S. investment-grade utility issuance was around $135 billion in 2025 and is projected to reach $145 billion in 2026. The consequence is that the public balance sheets investors watch most closely now tell only part of the story. "Alongside traditional bonds, hyperscalers have turned to off-balance-sheet arrangements to finance infrastructure expansions, often in partnership with private credit firms," BIS officials Egemen Eren, Ingomar Krohn and Karamfil Todorov wrote in a note. The combination of opaque structures and thin creditor protections makes losses harder to spot if data-center utilization or pricing falls short of what the leases assume.
This is the mechanism behind the reordering. For decades, the risk premium on emerging-market debt compensated investors for political instability, currency volatility and weak institutions — risks that are visible, priced and, in many cases, diversified away. The risk now building in AI credit is different: it is leverage layered on top of leverage, hidden in special-purpose vehicles, and tied to revenue assumptions that have never been tested at this scale. A developing-nation government bond comes with a prospectus and a sovereign balance sheet. A slice of a data-center financing vehicle often comes with neither.
The Second-Order Question Nobody Is Asking
The market's first-order question is whether AI will earn enough to justify the spending. The second-order question — the one that matters for the risk hierarchy — is who finances the wait, and where in the capital stack the losses land if returns arrive late or fall short.
The answer is that the losses, if they come, will not land where most portfolios assume. A diversified investment-grade corporate bond fund that was bought for stability now carries concentrated AI exposure through hyperscalers, utilities, equipment suppliers and the private-credit vehicles that lend to all of them. Mariya Entina, a portfolio manager at DoubleLine Capital, has argued that the AI boom has spread well beyond chipmakers and Big Tech into utilities and industrial suppliers, meaning bonds that do not look like AI plays may not provide the insulation investors expect. Meanwhile, the Treasury market is competing for the same pool of long-duration capital: U.S. debt stands at $40 trillion, the federal deficit is on track to reach $2 trillion this fiscal year, and debt-servicing costs are running at about $1 trillion a year. "The reverse crowding out in the corporate bond market has even gotten the Treasury Secretary's attention," Jurrien Timmer, director of global macro at Fidelity Investments, wrote on X.
Emerging-market debt sits on the other side of that competition. The International Monetary Fund noted in April that the share of emerging-market debt financing coming from portfolio investors — hedge funds, pension funds and insurers — has doubled over the past 20 years to 80%. That makes the flows faster and more reactive, but it also means the risk is transparent, marked daily and diversified across dozens of sovereigns and currencies. In a world where the safest-sounding credit is being funded through opaque structures, the old "risky" bucket is starting to look like the more honest one.
The Counter-Thesis: This Is Manageable, Not a Crisis
The strongest case against the reordering thesis is that it confuses scale with danger. The hyperscalers are not distressed borrowers; they are the most profitable companies in history, and they are borrowing by choice, not necessity. Goldman Sachs analysts have noted that consensus estimates put hyperscaler capex near 100% of operating cash flow, which is heavy spending but not insolvency. Morgan Stanley describes the group's starting position as one of "exceptional financial strength." Tai Hui of JPMorgan Asset Management said in July that while investors are watching AI companies' borrowing more closely, "debt levels remain manageable relative to the size of these businesses."
There is also a bullish reading of the bond market's move. Rising long-term yields may reflect investor expectations of an AI-driven productivity boom rather than fear. "The more positive interpretation of what the bond market is doing, which I think is getting a little bit less airtime among market participants, is this idea that the bond market might be sniffing out a little bit of a better productivity cycle from all the AI investment that's taking place," said Jacob Manoukian, JPMorgan Private Bank's U.S. head of investment strategy, in August. And JPMorgan has noted that credit spreads remain near cycle lows — the widening is real but starting from historically tight levels.
This counter-thesis is serious, and it is right about one thing: a broad credit crisis is not the base case. The hyperscalers' balance sheets can absorb a great deal of stress. But the reordering thesis does not require a crisis. It only requires that the market stop treating AI-linked credit as the risk-free anchor of the portfolio. That shift is already underway, and it is being driven by the structure of the financing as much as by the size of it.
The Signal That Would Prove This Wrong
The judgment that the risk hierarchy has structurally shifted rests on one observable condition: that AI borrowers remain large, leveraged and increasingly opaque. The falsifying signal is specific. If Oracle's debt-to-EBITDA stays below S&P's 4.5 downgrade trigger and hyperscaler free-cash-flow coverage returns to historical norms within four quarters, the widening is cyclical — a sentiment wobble that will revert — and the old risk hierarchy holds. If, instead, debt-to-EBITDA breaches 4.5, coverage ratios keep deteriorating, and off-balance-sheet financing keeps growing faster than disclosed debt, the reordering is structural and durable.
There is also a falsifying signal on the emerging-market side. If EM debt inflows reverse sharply — a return to monthly outflows approaching the March level of $66.2 billion — while AI credit spreads tighten back toward historic tights, then the inversion was a brief rotation, not a regime change.
What Comes Next: Three Time Horizons
In the short term, the trade is driven by sentiment and supply. Every new capex guidance raise from a hyperscaler, every data-center financing announcement and every ratings action on Oracle will move spreads. Investors should expect volatility to stay elevated and demand to remain two-sided.
Over the medium term, the fundamental test is cash flow. The hyperscalers must convert capex into revenue at a pace that restores free-cash-flow coverage. Base case: capex growth moderates toward cash flow, spreads stabilize at a wider but orderly level, and the risk premium between AI credit and EM debt narrows but does not fully close. Upside case: AI monetization accelerates, coverage ratios recover, and the old hierarchy reasserts itself. Downside case: utilization disappoints, leverage peaks above forecasts, and the opaque financing structures force a broader repricing of investment-grade tech credit.
In the long term, the structural question is whether the financing architecture of the AI buildout remains opaque. If data-center debt migrates permanently into private-credit and off-balance-sheet vehicles, then a portion of credit risk has left the transparent, mark-to-market part of the market for good. That is a durable change in where risk sits — and it is the change that makes emerging-market bonds, of all things, the relatively cleaner way to own credit risk.
The irony is worth sitting with. The borrowers that spent a decade being punished for weak institutions and hidden risks are now the transparent option. The borrowers that promised to be the safest in the world are the ones whose risk has gone underground. The AI debt binge did not just add leverage to the system. It reordered the map of where risk lives.
Explore more exclusive insights at nextfin.ai.
