NextFin News - Tech credit is selling off because investors are no longer treating AI expansion like a pure growth story. They are treating it like a funding problem. Oracle’s five-year credit-default swaps jumped to about 2.03 percentage points annually on July 20, the highest level on record for data back to the end of 2008, while a broader wave of AI-related borrowing has pushed the market to ask whether the biggest technology companies are building data centers fast enough to justify the debt that is funding them.
The timing matters. Oracle raised $25 billion in debt earlier this year, Meta sold $25 billion of investment-grade bonds in a later deal, and investors placed $96 billion of orders at the peak of that Meta sale. In the background, BofA Securities said five major AI hyperscalers issued $121 billion of U.S. corporate bonds last year, Moody’s said the Big Six hyperscalers were on track to spend $500 billion this year, and Morgan Stanley has projected $400 billion of hyperscaler bond issuance this year and roughly $1.5 trillion over five years. The market is not confronting one issuer. It is confronting a new scale of capital intensity.
That distinction is the heart of the selloff. In a normal credit wobble, spreads widen because rates move, volatility spikes, or investors trim exposure after a strong run. Here, the widening reflects a deeper question: if AI infrastructure takes years to monetize while the funding comes due now, who absorbs the timing mismatch? Bondholders do. That is why the move in Oracle CDS became a reference point for the whole hyperscaler complex. Once one marquee issuer is repriced, the rest of the sector no longer enjoys the benefit of the doubt.
The first-order move is obvious. The second-order move is more important. Higher borrowing costs can slow the pace of future AI build-outs, which can in turn pressure the contractors, data-center developers, power suppliers, and equipment vendors that depend on hyperscaler spending. If debt financing gets pricier, the capex curve can flatten before earnings do. The market is therefore not just pricing the current leverage; it is pricing the growth path that leverage is supposed to buy.
Why The Bond Market Is Repricing AI Spending
The mechanism is straightforward. AI infrastructure needs land, power, cooling, servers, and network equipment up front, but the cash flow from that infrastructure may arrive later. When a company funds that gap with debt, the bond market begins to care less about the narrative and more about the schedule. A project can be strategically sound and still be a credit concern if the financing window is too tight. That is exactly what the market is now testing.
Oracle is the clearest case because the company has become a proxy for the credit risk embedded in AI spending. On July 20, the cost of insuring Oracle debt rose to about 2.03 percentage points annually, a record for the available data series back to 2008. That level does not imply imminent distress. It does imply that lenders and hedgers now require more compensation for standing behind an issuer that is spending aggressively into AI infrastructure. Once protection costs move that far, the story stops being only about Oracle and starts becoming a read-through for the sector’s financing model.
Meta’s bond sale reinforced that message. The company sold $25 billion of investment-grade debt, but the terms did not look identical to its earlier financing and investors demanded more compensation than they had before. The deal still cleared by a wide margin, which shows that the market is not closed to AI issuers. It is simply becoming more selective about price. That distinction matters because the credit market usually reprices risk before equity investors fully internalize it.
The scale of issuance also changes the framing. If five hyperscalers issued $121 billion of U.S. corporate bonds last year and the biggest AI spenders are on track to require far more capital this year, then public credit is no longer just a funding option. It is an integral part of the AI growth model. That makes the bond market a gating mechanism. Equity investors can still tell themselves that the long-term payoff will justify the expense. Bond investors have to decide whether the financing bridge is long enough to reach that payoff without cracking.
“The cost of protecting Oracle Corp.’s debt against default reached a fresh multi-year high on Monday while its existing bonds sold off,” Bloomberg reported on July 20, citing ICE Data Services for the CDS move.
The message in that quote is not about default. It is about confidence. Credit spreads widen when confidence in the path from spending to cash flow weakens, and that is what is happening here. The market is saying that the AI race may still be strategically rational, but it is no longer obviously self-funding at current speed.
Is This A Cyclical Selloff Or A Structural Shift?
The answer is both, but not in equal measure. The cyclical piece is the easiest to see. Risk appetite in credit markets often fades after a long run in growth assets, and spreads can overshoot when investors rush to protect gains. That kind of move is usually mean-reverting. If the only problem were sentiment, calmer rates or better earnings guidance would likely restore demand for tech debt quickly.
The structural piece is more important. The AI build-out is turning some of the largest technology companies into much more capital-intensive businesses than they were in the last cycle. Data centers are not software. Power contracts are not app downloads. Server farms, chips, and interconnection capacity all require durable financing, and that financing can no longer be assumed to come entirely from operating cash flow. Once the market recognizes that the sector’s growth requires a heavier balance-sheet load, the credit premium can persist even if stock prices stabilize.
Three comparisons help explain why this feels different. In the 2020-21 technology boom, demand was rising but the biggest platform companies still generated enough cash to fund expansion internally. In the 2022 rate shock, duration-sensitive assets sold off, but the megacaps themselves were not yet being forced into repeated debt issuance to sustain the build-out. In the current AI cycle, the spending burden is bigger, the monetization horizon is less certain, and the financing is showing up in public debt markets. That combination is why the current move looks more structural than tactical.
There is also a second-order implication that is easy to miss if you only watch the quoted spread. Once the largest AI issuers borrow more, the pricing benchmark moves for the rest of the ecosystem. Private-credit lenders, project-finance vehicles, data-center developers, and infrastructure vendors all have to price off a higher reference cost of capital. The immediate effect is a wider spread on a few marquee bonds. The larger effect is a tighter funding environment across the AI supply chain.
The strongest counter-thesis is that investors are over-reading a normal capex cycle. The hyperscalers still have robust business franchises, durable demand, and access to capital. Their debt may be rising, but so are their earnings and their strategic control over the AI stack. From that perspective, the selloff is just a warning shot, not a regime change. That is a real objection. A market can panic about funding and later decide it was merely adjusting the price of a profitable expansion.
The falsifying signal is concrete. If the largest AI spenders continue to raise debt while spreads stop widening, CDS on Oracle and peer issuers retreat materially from recent highs, and new deals clear with order books that are as deep as earlier ones, then the market will have shown that the current selloff was cyclical rather than structural. If instead each new financing arrives with a higher premium and weaker demand, then the structural case strengthens.
What The Selloff Means For Credit, Equities, And The Next Phase Of AI Spending
In the short term, this favors the parts of the market that do not need to borrow aggressively to grow. High-quality short-duration credit and cash-rich issuers should look better than the most levered AI names because the market is charging more for balance-sheet risk. The exposed names are the companies that need the most funding to stay in the race. Their debt is now being judged not only on coupon and maturity, but on how quickly spending can become cash flow.
Over the medium term, the question is whether AI monetization catches up with capex. If cloud demand, enterprise adoption, and pricing power improve fast enough, then debt can remain a bridge. If they do not, then higher borrowing costs can slow the build-out before the business case is fully proven. That would not kill AI. It would make the rollout slower and more selective, with capital flowing first to the projects that look the most immediately financeable.
Over the long term, the sector is moving toward a more industrial capital structure. AI is increasingly looking like a power-and-infrastructure business layered with software margins, not a pure software model with light fixed costs. That does not make it a bad business. It does make it a more balance-sheet-sensitive one. The winners will be the firms that can secure cheap, long-dated capital and convert that capital into recurring revenue fast enough to keep creditors comfortable.
The next signals to watch are simple: Oracle’s next financing steps, any fresh CDS moves, the size and pricing of upcoming AI-related bond sales, and whether the largest hyperscalers keep raising capex guidance without losing bond-market access. If spreads continue to drift wider even as the deals keep coming, the market will be telling investors that AI is still winning in technology but losing some of its freedom in credit.
NextFin News - The bond market is not refusing AI; it is putting a price on speed. If the earnings bridge arrives on time, this will look like a temporary scare. If it does not, the selloff may mark the point where tech debt stopped being cheap growth fuel and started becoming the market’s first hard test of the AI boom.
Explore more exclusive insights at nextfin.ai.
