Big Tech built its reputation on mountains of cash. It is building its AI empire on debt, and the bill is starting to land in Europe. The five biggest builders of AI data centres in the US—Alphabet, Amazon, Meta, Microsoft, and Oracle—have doubled their debt over five years, piling on about $350bn, according to data compiled by Bloomberg. They are betting that cutting-edge AI will one day pay it all back.

The scale of the gamble

For now, the interest is almost a rounding error. The five paid a combined $10bn on their debt last year. That is more than double 2019, but tiny beside Google’s cash flow of roughly $64bn last quarter. These are still some of the most profitable firms on Earth. Yet the strain shows at the edges. Amazon’s free cash flow turned negative in the March quarter. Oracle’s debt reached about 2.5 times its sales, and S&P cut its rating on Thursday to one notch above junk, blaming AI spending.

The AI infrastructure build-out is unlike any other in corporate history. Hyperscale data centres—facilities that can house hundreds of thousands of servers—require enormous upfront capital. A single data centre can cost $1bn or more, and the largest firms are building dozens simultaneously. The chips needed to power AI models, primarily Nvidia's H100 and newer Blackwell GPUs, are also in short supply and command premium prices. The trillion-dollar question is whether the demand for AI services will grow fast enough to generate returns that cover these massive investments.

History offers cautionary tales. The dot-com bubble of the late 1990s saw companies borrow heavily to lay fibre optic cable and build internet infrastructure. Much of that capacity went unused when the bubble burst, leading to bankruptcies and a wave of consolidation. Today’s AI boom has drawn comparisons, though proponents argue that AI adoption is following a trajectory similar to the early internet, only faster. The difference now is the unmatched financial firepower of Big Tech.

The market is starting to flinch

Investors have lapped up the bonds so far. That may be changing. Amazon’s $25bn sale this week drew a notably cool reception, the softest hyperscaler launch since Meta’s last October. Traders are now dumping older tech bonds, including Amazon’s, Nvidia’s, and Oracle’s, simply to make room for the new flood. The buyers are running out of space, not confidence. Still, the warning signs are stacking up. “Credit risk is too undervalued right now in the market,” Morgan Stanley’s Vishal Khanduja told Bloomberg TV.

Bond markets are a leading indicator of financial stress. When demand for new issues weakens, it typically means investors are reassessing risk premiums. In the case of Big Tech, the risks are twofold: first, that AI revenue may take longer to materialize than expected, and second, that the sheer volume of debt issuance will overwhelm even the deepest capital markets. Already, the five firms have cumulatively issued more bonds in the last three years than in the prior decade combined.

Credit rating agencies are taking notice. Moody’s and S&P have flagged rising leverage across the sector, though most ratings remain investment grade. Oracle’s downgrade, however, is a warning shot. If other hyperscalers see their credit profiles slip, their borrowing costs would rise, squeezing margins and potentially slowing the pace of data centre construction.

Why this reaches Europe

Here is the part that should matter on this side of the Atlantic. American tech has run short of dollars to borrow, so it has turned to Europe. Hyperscalers issued no non-dollar bonds in 2024. By 2026 it is a core part of their funding. Morgan Stanley expects their euro borrowing to reach €50bn this year, as TechFundingNews reports. That would make US Big Tech the single largest source of corporate debt in the eurozone, ahead of France. Alphabet alone has borrowed in yen, Canadian dollars, Swiss francs and sterling inside a year, and even sold a 100-year bond.

When American giants crowd into the same euro debt that European scale-ups and infrastructure funds rely on, the cost of money shifts here too. A startup in Munich or Paris with no AI exposure at all can end up paying more, simply because Amazon got there first. The mechanism is straightforward: as the supply of euro-denominated bonds from these issuers swells, yields rise to attract enough buyers. That pushes up yields on all similarly-rated corporate bonds, especially those from technology and growth companies. Small and medium-sized European firms, which typically have lower credit ratings than Big Tech, see their borrowing costs increase disproportionately.

The European Central Bank has also begun to wind down its bond-buying programmes, removing a key support for corporate debt markets. This double whammy—rising supply and shrinking official demand—could create a credit squeeze for non-tech sectors just as the European economy is struggling with sluggish growth and high energy prices. Policymakers in Brussels have already expressed concern about the concentration of debt issuance among a handful of US giants, warning that it could amplify financial stability risks.

The Intel warning

Not everyone is worried. Amazon boss Andy Jassy says he has “high confidence this will be monetized.” Mark Zuckerberg insists demand for computing power keeps outstripping supply. Gil Luria of DA Davidson agrees the load looks manageable: “If they were borrowing an order of magnitude more? That would look bad.” Others are blunter. “It seems like a lot of demand hype that is very aspirational at this point,” said Fitch’s Jason Pompeii.

The cautionary tale sits one industry over. Intel spent years loading up on debt, missed the AI chip boom entirely, and needed a US government bailout and an Nvidia investment to survive. Intel’s fall from grace is instructive. Once the world’s leading chipmaker, Intel borrowed aggressively to fund share buybacks and dividends in the 2010s, while Nvidia and AMD invested heavily in AI-focused architectures. By the time AI workloads exploded, Intel had no competitive products, and its debt pile became a millstone. The company’s recent reliance on a $8.5bn CHIPS Act grant and an Nvidia partnership underscores the risks of betting the balance sheet without a clear payback path.

There are parallels with today’s AI debt binge, but also important differences. Big Tech’s debt is used to build physical assets—data centres and the electrical infrastructure to power them—which can be repurposed for cloud computing or other advanced workloads even if AI demand disappoints. Yet the scale is unprecedented. If only half of the projected AI revenue materializes, the debt service could become crippling for leveraged players.

Why it matters

The AI build-out has quietly become one of the largest debt bets in corporate history, and one research shop thinks the AI debt market could reach $7tn by 2029. Only Alphabet’s stock has beaten the market this year, while Microsoft and Oracle have both fallen more than 20%. The share buybacks that once defined these companies have all but stopped. The gamble is no longer just whether AI works. It is whether the revenue arrives before the debt does, echoing the bubble warnings and eye-watering valuations already circling the sector. It is the same wager driving Nvidia’s bond sales and ByteDance’s borrowing.

Europe just became a place where that wager gets settled. The continent’s reliance on imported tech and capital means it will feel the consequences first. European regulators are already scrutinizing the debt concentration, and the European Investment Bank has signaled willingness to provide alternative funding for local digital infrastructure. Yet the immediate reality is that every euro borrowed by a US hyperscaler is a euro that cannot finance a European AI startup or a green data centre project. The next few months will show whether this dynamic accelerates or fades. One thing is certain: the era of easy money for AI is ending, and the reckoning is global.

Source: TNW | Amazon News