Goldman Sachs Underwrites $5.4 Billion Debt for Microsoft-Linked Data Center
Miles Bennett
Goldman Sachs is sounding out a $5.4 billion debt package for Blackstone's QTS data centers, backing a Georgia facility leased to Microsoft — but with AI infrastructure bonds selling off, whether this deal prices at all has become a market-sentiment test in itself.
How is the $5.4 billion deal structured?
The package splits into roughly $4.9 billion in secured bonds (five-year) and about $500 million in term loans (seven-year), backed by a QTS data center leased to Microsoft in Georgia.
This means → Microsoft's long-term lease is the core credit anchor — stable rental cash flow is what makes the bonds marketable.
Sources say timing remains under discussion and terms may shift, given the recent sell-off in AI-linked bonds. Blackstone and Goldman declined to comment; QTS and Microsoft did not respond.
How cold has the AI infrastructure bond market turned?
This week BlackRock's $12.5 billion investment-grade bond issue for a Meta-linked data center drew tepid demand; CoreWeave had to sweeten terms to close a $2.6 billion loan.
QTS's own outstanding debt is sliding: a $3.25 billion leveraged loan completed this month has dropped from an issue price of 99.5 cents to 97.75 cents. Its $4.6 billion investment-grade bonds issued in April have seen spreads — the gap between a bond's yield and Treasury yields, where a wider gap means buyers see more risk — widen from 138 basis points at issue to roughly 239 basis points now.
In plain terms = the same company, the same asset class, could sell near par weeks ago. Now buyers demand significantly more compensation before they will take the paper — appetite for AI infrastructure debt is contracting fast.
Why has sentiment shifted so abruptly?
The immediate trigger: Alphabet last week raised its 2026 capex forecast to as much as $205 billion and posted its first quarter of negative free cash flow since its 2004 IPO.
This means → Big Tech's AI spending pace has outrun its own cash generation — investors are asking: when do the returns from all these data centers actually arrive?
This reflects a deeper pivot in market concern — from "can AI work?" to "can AI infrastructure pay for itself." Spreads on outstanding bonds keep widening, and new issuers must pay steeper premiums. Whether this $5.4 billion deal prices successfully is itself the next sentiment checkpoint.
Content is for reference only, not financial advice.