Morgan Stanley has become the leading investment bank structuring some of the largest financing deals fueling the rapid expansion of artificial intelligence infrastructure, as demand for computing capacity drives unprecedented investment in data centers and advanced chips.
According to industry executives, the bank has established itself as the dominant adviser behind many of the most significant AI-related debt and equity financings since last year. Its transactions include a $3.2 billion bond for data center developer TeraWulf backed by Google, a $27 billion debt package supporting Meta’s Hyperion data center partnership with Blue Owl, and advisory work for Broadcom on a $35 billion chip financing deal.
The surge in AI-related transactions has strengthened Morgan Stanley’s position on Wall Street. Data from LSEG shows the bank generated $2.3 billion in debt and equity capital markets fees during the first half of the year, up from $1.4 billion during the same period a year earlier. The performance lifted Morgan Stanley above longtime rival Goldman Sachs to become the world’s second-largest bank for capital markets fees behind JPMorgan Chase, improving from fourth place a year earlier.
The financing structures being developed illustrate how artificial intelligence is transforming not only the technology sector but also the mechanisms used to fund large-scale infrastructure projects. Instead of relying solely on conventional project finance or traditional corporate borrowing, investment banks are increasingly packaging long-term computing contracts together with the financial strength of major technology companies into securities that can be sold to a broad range of institutional investors.
These arrangements have significantly expanded the amount of capital available for AI infrastructure while increasing the financial system’s exposure to continued demand for artificial intelligence computing capacity.
Morgan Stanley’s Co-Head of Investment Banking, Mo Assomull, said the scale of transactions has increased dramatically as AI investment accelerates. Deal sizes that once ranged between $1 billion and $5 billion are now regularly reaching $10 billion, $20 billion or more.
The investment surge comes as technology companies including Google, Amazon, Meta and Microsoft continue to increase spending after indicating they remain unable to fully meet customer demand for AI services. Morgan Stanley estimates the broader AI infrastructure build-out could require approximately $10 trillion in investment over the coming years.
One of the key developments lowering financing costs has been the involvement of these major technology companies, often referred to as hyperscalers. Their strong balance sheets allow projects backed by long-term lease commitments to secure financing at substantially lower borrowing costs than would otherwise be available.
Morgan Stanley’s Co-Head of Leveraged Finance, William Graham, played a central role in developing what has become a template for AI infrastructure financing. His team structured a bond for TeraWulf that combines characteristics of a broadly distributed bond with protections commonly found in project finance loans. The security was effectively supported by Google through long-term lease commitments, while most of the facility’s computing capacity is expected to serve Anthropic, according to people familiar with the matter.
The hybrid structure attracted insurers, pension funds and asset managers that had not traditionally participated in financing data center construction. TeraWulf ultimately raised $3.2 billion through bonds yielding 7.75%.
TeraWulf Chief Financial Officer Patrick Fleury said the financing enabled the company to avoid the lengthy approval process associated with conventional project finance loans while securing borrowing costs that made the economics of the project viable. Additional safeguards included lockbox mechanisms directing lease payments directly to bondholders and other forms of collateral designed to strengthen investor confidence.
Since completing the TeraWulf transaction, Morgan Stanley has arranged more than $40 billion of similar construction bond financings and is expanding the model into markets across Europe and Asia. Graham said he expects AI infrastructure bonds to become the largest source of new non-investment-grade debt issuance, describing the sector as the fastest-growing segment of the credit market in two decades.
Not every financial institution has embraced the trend with equal enthusiasm. Some competing bankers have expressed reservations about becoming heavily associated with data center financing, citing local opposition to new facilities in parts of the United States. JPMorgan Chief Financial Officer Jeremy Barnum also said the bank had reviewed certain data center lending proposals that it ultimately declined because of their financing terms.
Morgan Stanley has also expanded the financing model beyond physical data centers. In May, the bank partnered with MUFG to arrange a $3.1 billion syndicated loan for cloud computing company CoreWeave to purchase Nvidia graphics processing units, marking the first broadly syndicated term loan specifically designed to finance AI chips. The transaction reportedly attracted nearly $20 billion in investor demand.
Under this structure, the chips and the data centers housing them are financed separately. Buildings are supported by lease agreements, while the graphics processors are financed against long-term take-or-pay contracts with customers. Graham compared the arrangement to financing a high-performance vehicle separately from the garage needed to house it, emphasizing that investors remain primarily focused on the strength of the counterparties signing long-term usage agreements.
That distinction also highlights one of the emerging risks in the AI financing market. As transactions rely less on the balance sheets of technology giants and increasingly on AI developers consuming computing capacity, the underlying credit quality becomes less certain. Morgan Stanley helped arrange an $8.5 billion chip-backed loan for CoreWeave earlier this year priced at a narrower spread because it was supported by a hyperscaler contract, while a later financing backed by two AI laboratories carried significantly higher borrowing costs.
Raj Joshi, Senior Vice-President at Moody’s Ratings, said the financial health of companies such as Anthropic and OpenAI remains an important factor to monitor as AI investment continues to expand. He described the current spending cycle as unprecedented in scale, noting that there are few historical comparisons for the pace and magnitude of capital being committed to artificial intelligence infrastructure.
