Morgan Stanley has become the leading Wall Street bank arranging and structuring debt and bond financing for large-scale artificial intelligence infrastructure projects. According to the Financial Times, the bank is channeling tens of billions of dollars into data-center construction and related hardware purchases through new credit and capital structures.
Key transactions that propelled the bank
Industry sources identify several high-profile deals led by Morgan Stanley:
- a $3.2 billion bond issuance for TeraWulf that is guaranteed by Google;
- a $27 billion loan package for Meta’s Hyperion project, arranged jointly with Blue Owl;
- a roughly $35 billion chip-financing agreement for Broadcom.
In addition, Morgan Stanley and MUFG structured a $3.1 billion loan in May for CoreWeave to acquire Nvidia GPUs — reported as the first widely syndicated GPU-financing deal, which attracted nearly $20 billion of investor interest.
Business impact: fees and ranking
The surge in AI-related transactions helped Morgan Stanley overtake Goldman Sachs in the first half of the year: the bank’s investment-banking fee revenue rose from $1.4 billion to $2.3 billion, moving it from fourth to second place globally, behind only JPMorgan Chase.
The financing model: securitisation and hyperscalers
Rather than relying on traditional project finance or plain corporate loans, Morgan Stanley’s bankers design hybrid instruments that securitise long-term compute capacity leases together with the balance-sheet strength of major technology firms. The strategy hinges on involving hyperscaler guarantors — Google, Amazon, Meta, Microsoft — because if one of these companies guarantees a data-center’s lease payments, financing costs fall substantially, in some cases by nearly half.
The TeraWulf structure became an early template. William Graham, co-head of Morgan Stanley’s leveraged finance business, and his team developed a tradeable bond that incorporates certain project-finance protections and sits behind a Google guarantee. That enabled TeraWulf to raise $3.2 billion at a 7.75 percent yield.
Investor protections are reinforced by project-finance mechanisms such as lockbox accounts that segregate lease payments and route them directly to lenders.
Scale and geographic rollout
Since the TeraWulf deal, Morgan Stanley has sold more than $40 billion of similar securities and has begun applying the model in Asian and European markets. William Graham expects that bonds linked to AI infrastructure could make up a large portion of future high-yield issuance.
Risks: widening spreads and credit quality
Rapid growth brings risks. Widening credit spreads highlight a central concern: the farther financing moves away from the balance sheets of well-capitalised hyperscalers and toward AI development labs that actually consume compute capacity, the more uncertain the underlying credit quality becomes.
Raj Joshi, a senior vice president at Moody’s, noted that the financial positions of companies such as Anthropic and OpenAI are risk factors to monitor closely. He described the situation as an unusually large investment cycle without close historical precedent.
Conclusion
Morgan Stanley’s structured financing approach has quickly become influential, boosting the bank’s capital markets revenue and opening the data-center financing market to institutional investors. Nonetheless, rising spreads and a shift of credit exposure toward less-proven AI developers introduce meaningful credit risks that market participants will need to track.
