The real danger is not hyperscaler debt, but the capital they divert

EUROPEAN CORPORATE and sovereign bond markets face an uncomfortable fact: American tech giants are growing more creditworthy than some of the governments on whose territory they operate. That observation, which sounded like a thought experiment a few years ago, is now plumbing the mechanics of the eurozone's bond market.

Hyperscalers - large technology firms such as Amazon, Alphabet, Microsoft and Oracle - are borrowing at a scale and in a geography that European issuers did not expect. Their debt issuance has reshaped the composition of European bond indexes, competed for investor capital that might otherwise have gone to local companies and sovereigns, and tested the limits of demand. Morgan Stanley has noted that persistent supply may contribute to sector-level spread fatigue. The diagnosis is half right. The danger is not merely that hyperscaler debt floods European markets. It is that the very same investors who buy it must also buy the bonds of European governments and corporations. Capital is not infinite.

The numbers are striking. Six hyperscalers have issued around $244 billion in bonds globally this year, compared with $108 billion in all of last year. They are on pace for a full-year total near $400 billion. Goldman Sachs expects debt to fund roughly one third of hyperscaler capital expenditure this year, rising to roughly 35% in 2027 when total spending approaches $1.14 trillion. Alphabet alone reported its first quarter of negative free cash flow since the company went public as Google in 2004, making the turn to bond markets no longer optional but structural.

The incentive to issue in Europe is obvious. Borrowing costs in euro are competitive with those in dollars, the market has deepened, and diversifying currency exposure reduces the risk of over-tapping the US market, where frequent issuance can weigh on existing bonds. Amazon raised 14.5 billion euros ($16.88 billion) in March from an eight-part deal, the largest ever in the euro corporate bond market, according to LSEG. Morgan Stanley expects hyperscalers to raise some €50 billion in euro debt this year, enough to help push the United States ahead of France as the eurozone's biggest source of corporate borrowing. The share of US companies in the Morningstar Eurozone Corporate Bond Index has risen to 20.12%.

The trouble is not that hyperscalers are borrowing. It is that the pool of investment-grade bond demand is shared. European pension funds, insurance companies and retail investors who track bond indexes are now compelled to buy proportionally more of Amazon's debt and less of, say, Siemens or the Italian treasury. Because bond indexes are market-capitalisation-weighted, every new investment-grade bond that qualifies for inclusion increases the issuer's share and forces passive buyers to follow. This is not a choice made by portfolio managers. It is an accounting mechanic.

Worse, the supply is not finished. Goldman Sachs, UBS, Bank of America and S&P Global all point to more issuance ahead. Hyperscaler capital expenditure in 2026 is on pace to consume close to 100% of operating cash flows, compared with a 10-year average of 40%, according to UBS. The gap between spending and earnings is being bridged by bond markets. And the borrowing that shows on balance sheets - itself approaching $1.35 trillion at the five largest US tech firms - understates the picture. Moody's estimates an additional $1.2 trillion of off-balance-sheet, debt-equivalent liabilities, much of it tied to data centres under construction. Nikkei found that "hidden debt" - long-term chip purchase contracts, data-centre leases and similar obligations - has exploded by 8x in just four years to $1.65 trillion.

Market participants are beginning to push back. The investor-order ratio for hyperscaler bonds - how many dollars of demand arrive for each dollar of issuance - was nearly five times in February, according to Torsten Slok, chief economist at Apollo Global. By July it had fallen below two. Amazon had to add 18 to 21 basis points of extra yield on its longest-dated debt in a recent sale, with orders at just 2.5 times the bonds on offer, down from 3.2 times in March, according to Bank of America. Spreads are widening. Investors are not running for the exits, but they are asking for more compensation.

This matters for Europe because European governments and companies now compete for the same investors who are absorbing hyperscaler supply. The US federal deficit is projected to approach $2 trillion this fiscal year. The Federal Reserve is no longer buying bonds, leaving private investors to absorb both the government surplus and the corporate wave. European sovereigns face their own borrowing needs, with public-debt burdens elevated and growth sluggish. If the global supply of investment-grade bonds grows faster than the appetite of institutional buyers, the cost of capital rises everywhere - but unevenly. European issuers with weaker credit profiles or smaller investor bases will bear the brunt.

To be sure, the borrowing has a purpose. AI infrastructure is not a speculative asset bubble in the classical sense: data centres, chips and power grids are real physical capital. If the investment generates returns that exceed its cost of capital, the debt is justified and the eventual cash flows can service it. Hyperscalers entered this cycle from positions of extraordinary financial strength - enormous cash balances, strong credit ratings, and balance sheets that credit agencies still rate among the most robust in the corporate world.

But three things could go wrong. First, returns on AI infrastructure may disappoint. The hyperscalers are spending hundreds of billions on capacity that the market has not yet demonstrated it needs. Cloud computing in prior cycles saw periods of overcapacity; a similar outcome here would leave debt-servicing costs without corresponding revenue. Second, even a modest repricing of risk could prove disruptive. A rise in spreads of 50 basis points on a $1.2 trillion debt-equivalent exposure is not a headline but it is a real drag on cash flows. Third, the political risk of rent-seeking is rising. The same firms that compete fiercely in the cloud are collaborating in infrastructure, lobbying for subsidies, and leveraging regulatory relationships to secure power, permits and tax incentives. That is not illegal, but it changes the nature of the competition from one based on innovation to one based on access. Investors who funded the former may be less comfortable with the latter.

The deeper question, which Morgan Stanley's warning touches but does not fully articulate, is about the allocation of capital across borders. Europe has long struggled with its low productivity, its shortage of deep technology firms, and its reliance on foreign innovation. The irony of the current cycle is that European investors, through index funds and pension mandates, are now one of the primary funders of the very American technology infrastructure that Europe's own companies cannot replicate. That is not inherently harmful. Cross-border capital flows are a feature of an integrated financial system. But it does raise the uncomfortable question of whether Europe's financial markets are becoming a distribution channel for American corporate expansion rather than an engine of European investment.

The better answer for European policymakers is not to restrict capital flows. That would be counterproductive and unenforceable. It is to address the structural reasons why European corporations cannot match the scale, creditworthiness or investor appeal of American tech firms. A smaller, less dynamic corporate base means fewer high-quality domestic borrowers, which means investors have fewer choices, which means they buy foreign bonds. The fix is not a capital-control disguise; it is a competition policy that fosters European technology firms capable of issuing their own investment-grade bonds at attractive spreads.

The hyperscalers are not doing anything wrong. They are borrowing to build infrastructure, and investors are buying because the credits are strong and the yields are decent. The risk is not in the bonds themselves. It is in the crowding-out effect that occurs when the world's largest borrowers happen to be headquartered on the other side of the Atlantic.

That imbalance will not correct itself.