Goldman Sachs is structuring a $1.15 billion high-yield bond offering to finance a data center in Virginia that will be leased to CoreWeave, the AI cloud infrastructure company that has become Wall Street’s favorite ticket into the GPU economy.
CoreWeave’s name on a lease has become something close to a credit enhancement in its own right. Investors keep showing up, and the yields they’re willing to accept keep falling.
How a GPU company became Wall Street’s favorite tenant
CoreWeave sits at an unusual intersection: it is simultaneously one of the largest renters of AI computing capacity and one of the most active borrowers in the high-yield market. By April 2026, the company had issued multiple tranches of notes totaling at least $2.75 billion, one of which carried a 9.75% coupon. By August 2026, its combined debt and equity raises had crossed $30 billion.
The $2.6 billion delayed-draw term loan CoreWeave closed around August 10, rated Ba2 by Moody’s and BB+ by S&P, gave the company more flexibility to finance shorter-duration contracts.
Goldman’s Virginia deal fits neatly into this pattern. A real estate entity or data center developer builds the facility, locks in CoreWeave as a long-term tenant, and then finances the construction through the high-yield bond market using that lease as collateral.
A crowded trade with compressing yields
Goldman is not alone in seeing the opportunity. Morgan Stanley is co-leading a roughly $3.5 billion high-yield offering for Galaxy Digital, announced July 22, 2026, to finance a data center in Texas that will also be predominantly leased to CoreWeave.
Earlier in 2026, Elk Grove Village Property LLC raised between $850 million and $900 million in a high-yield note deal to fund a Chicago-area facility leased to CoreWeave, pricing at a 7.5% yield. That number matters because it represents where investor appetite has driven spreads: down from double-digit territory to a range more commonly associated with investment-grade-adjacent credits.
When investors first started buying CoreWeave-backed paper, they demanded yields north of 10% to compensate for the novelty and the concentration risk. Now they are accepting yields in the 7% to 7.5% range.
What the AI infrastructure debt wave actually means
The underlying logic of these deals is straightforward. AI model training and inference require enormous amounts of GPU compute. Building and running GPU clusters requires physical data centers. Data centers require capital. CoreWeave provides the long-term lease commitments that make the capital accessible through debt markets. Goldman, Morgan Stanley, and their peers collect fees for arranging it all.
There is also a concentration concern that does not require much squinting to see. Multiple billion-dollar bond deals, all with CoreWeave as the anchor tenant, all issued within months of each other, create a web of interconnected credit exposure.