The European Central Bank has flagged growing concentration in long-dated euro corporate bonds as large US technology companies borrow to fund artificial intelligence infrastructure. These issuers account for close to 8% of investment-grade euro corporate bonds with more than ten years remaining until repayment, despite having a much smaller presence across the wider market.
The financial-stability assessment was published on 17 November. Its focus is the exposure building up in a part of the bond market used by long-term investors, rather than a finding that an AI-related market crash has occurred or that other borrowers have already been squeezed out.
Concentration at the long end
The companies involved are often described as hyperscalers: large technology businesses operating extensive cloud and computing infrastructure. Their investment programmes require substantial funding for data centres and other AI capacity. During 2026, they have issued more than $200 billion of investment-grade corporate bonds denominated in US dollars and have also increased their borrowing in euros.
The near-8% figure has a specific scope. It covers corporate bonds denominated in euros, rated investment grade and with a remaining maturity above ten years. It does not describe the entire euro bond market. The technology issuers’ share of the broader corporate benchmark was slightly above 1% in August.
New hyperscaler debt has maturities averaging two to six years longer than that of comparable companies over the period examined. That makes their presence particularly noticeable among longer-dated bonds. Even if their need for new external finance slows, debt already issued can remain in investor portfolios for years.
There are benefits to a broader choice of securities. New issuers can give investors access to different sectors, higher credit ratings and longer maturities. All US companies combined represent close to one-fifth of the euro investment-grade corporate benchmark in the market examined, however, increasing European investors’ exposure to developments in the US economy.
Long debt and fast-changing technology
A central vulnerability is the mismatch between long-term borrowing and AI assets that can depreciate quickly. Computing equipment may lose economic value much sooner than the associated bonds reach maturity. Sustained revenues therefore matter to the resilience of the companies undertaking these investment programmes.
Long-dated bonds also carry interest-rate risk. Their prices are particularly sensitive to changes in the yields investors require. A reassessment of technology companies’ future earnings could affect both their shares and their borrowing costs, reducing the protection investors might expect from holding different types of assets.
The growing weight of hyperscalers in bond benchmarks can strengthen the connection between US equity performance and European credit markets. The measured relationship has weakened recently, so it should not be treated as a fixed rule. The concern is that a substantial change in confidence about AI revenues could spread across portfolios rather than remaining confined to technology shares.
Crowding out remains a risk, not a finding
Insurance companies and pension funds, important buyers of long-dated debt, have shown a modest tilt towards hyperscaler bonds beyond changes implied by their benchmarks. The portfolio comparison covers the first quarter of 2025 to the second quarter of 2026. These investors have retained a preference for sovereign bonds while underweighting other corporate debt and some public-sector issuers.
Those allocation changes are marginal and cannot be attributed solely to the arrival of technology issuers. They do not establish that European companies have already lost access to finance because of AI borrowing. They do, however, identify a possible route through which continued heavy issuance could eventually displace other borrowers.
If future funding needs grow and investor demand cannot expand sufficiently, competition for capital could increase borrowing costs for other companies and place upward pressure on long-term government bond yields. That remains a conditional scenario. The immediate development is greater scrutiny of issuer concentration and connected market risks as the AI investment cycle expands, not confirmation that those risks have already materialised.