AI Data Center Financing Pushes Nordic High-Yield Bond Market to Its Limits
0xBroomberg
A record €800 million AI data-center bond opened the Nordic high-yield market to mega-scale deals — then two follow-on issues failed in quick succession, exposing a capacity ceiling in a market built for mid-cap borrowers.
€800 million went through — why did the next two stall?
In May, PolarDC — a data-center developer under H.I.G. Capital — raised €800 million in the Nordic bond market, the first deal of its kind in Europe.
Days later, U.S. operator Prime Data Centers cut its planned issue from $600 million to $500 million, then shelved it after investors questioned the terms.
U.K. operator Pure Data Centres (backed by Oaktree Capital) abandoned a potentially record-breaking €1 billion bond and turned to bank loans instead. This means → the first deal's success did not open the floodgates; it revealed how little the market could swallow at once.
Why can't the Nordic market absorb deals this large?
The Nordic high-yield market is known for flexibility and higher risk tolerance, but it mainly serves mid-sized companies. Lead arrangers are regional firms like Pareto Securities and Arctic Securities.
In plain terms = this is a market used to tickets in the tens-of-millions-to-low-hundreds range, suddenly asked to digest $500 million to $1 billion single issues.
For comparison, Wall Street can place a $25 billion mega-cap tech bond in a single day — an entirely different scale.
Investors cannot value these bonds — what is the problem?
Pure Data Centres' unsecured bond was marketed using two metrics at once: "contracted EBITDA" (common in infrastructure finance) and "loan-to-value" (a real-estate metric).
This means → high-yield investors could not apply their standard cash-flow-to-debt framework; valuation became guesswork.
Adam Darling, high-yield fund manager at Jupiter Fund Management, was blunt: "I have zero interest in these data-center deals" — a judgment on the entire category, he said, not any single issuer.
How wide is the gap between "AI narrative" and actual revenue?
Jayadev Mishra, portfolio manager at Bank J. Safra Sarasin, noted: "A data center comes to market and your first thought is AI compute. But many issuers are not that — they are just riding the AI-tenant narrative."
Pure Data Centres' main clients are IT service providers, telecom operators, and financial institutions. Its 2025 financials show revenue driven primarily by cloud business.
The company told Bloomberg its current revenue is a mix of cloud and AI-inference income, and that it chose bank financing because the terms were more favorable. This reflects an issuer effectively conceding that a pure-AI revenue story does not hold up.
Does buying this debt feel more like equity risk?
Bonds issued at the holding-company level fund multiple projects under construction and sit structurally subordinate to project-level financing.
In plain terms = bondholders bear risk close to equity — they lose first if a project fails, yet capture none of the upside if it succeeds.
Two hard risks loom during the build phase: tenants may exit leases over disputes, and power-grid access issues can strand a project entirely.
What does this mean for the broader AI infrastructure boom?
Concerns are rising over overheating in AI infrastructure "picks-and-shovels" deals — many issuers must keep borrowing before investment returns materialize, and those returns may be years away.
The Nordic market showed signs of "indigestion" after just a few deals. This means → whether mega-scale data-center financing can find deep enough capital pools will be a deciding variable for how far this AI buildout wave can go.
Put simply = where the money to build AI data centers comes from — and whether markets will keep writing the checks — has shifted from a technology question to a financial one.
Content is for reference only, not financial advice.