QTS, the data center operator Blackstone bought for $10 billion in 2021, went to the private placement market looking for about $800 million and closed with roughly $300 million, a shortfall of some 60 percent. The deal was launched on Oct. 26 through a special purpose vehicle called QTS Thunder Managing Issuer LLC, with Mizuho, Natixis, Societe Generale and Wells Fargo arranging notes of seven to fifteen years. Pricing was expected within days. It took three weeks and ended smaller.
What actually got done, according to people familiar with the matter cited by Bloomberg, was a $100 million seven-year private placement at 245 basis points over benchmarks, plus a $200 million five-year bank loan that still has to be priced in the coming weeks. The original $800 million was meant to refinance existing five-year bank debt. With the smaller deal, that bank financing stays where it is, on bank balance sheets rather than moved to insurers and long-dated asset managers.
The pricing is the part credit investors will study. With the 10-year Treasury at 5.15 percent, a 245 basis point spread puts an investment grade data center landlord above 7 percent. For comparison, QTS sold its first green project bond in April, a 10-year note rated Baa2 by Moody's, with initial price talk around 162.5 basis points, and Goldman Sachs had the dollar investment grade index near 89 basis points earlier this month. On that yardstick the new paper priced about 2.7 times the index, for a shorter maturity, though different structures make the comparison approximate.
Blackstone declined to pay up. People familiar with the discussions said the firm was reluctant to offer meaningfully higher yields because it borrows frequently across multiple markets and through numerous issuers, so a wider print on a QTS deal would reset the reference point for investors pricing other Blackstone-related debt. The choice was to borrow less rather than reprice the whole complex.
Demand was also structurally capped. Private placements are bought mainly by life insurers and long-only managers, a narrow pool. QTS has been a repeat issuer: a $1.65 billion investment grade bond sale last year and a separate $800 million financing, with roughly $11 billion raised over about 18 months by one count. Some investors balked at the spread, while others had simply reached internal limits on exposure to a single issuer, which is what a market looks like when concentration rules bite before credit quality does.
The backdrop is a broader shift in AI infrastructure finance. Goldman Sachs says hyperscaler bond issuance broke its $250 billion full-year forecast two months early, and Barclays puts AI-related bond spreads about 44 basis points wider than the investment grade index, against roughly 10 basis points tighter a year ago. The direction of travel is toward paying more, or borrowing less, or moving to private credit and project finance at higher cost.
QTS still has an expensive build ahead. Bernstein puts its planned U.S. capacity at about 14.3 gigawatts, of which roughly 3 gigawatts is operating, 4.2 gigawatts is under construction and 7.1 gigawatts has not started, with only about 1.6 gigawatts of the construction matched to named tenants in Bernstein's database, though QTS says it builds against signed contracts and some leases may not be public.
This is not a credit event. It is a small deal from a strong borrower, and the company walked away from the table rather than accept worse terms, which is what investment grade issuers are supposed to be able to do. But it is the clearest evidence yet that AI data center debt has moved from a seller's market to one where price discovery happens at the margin, and where the marginal buyer has a limit. With Nvidia reporting earnings on Tuesday, the market now has a concrete data point to weigh against the assumption that AI capex can always be financed.
Comments (0)
Log in to join the discussion
Log InNo comments yet