Morgan Stanley estimates that from 2025 to 2028, global data center spending driven by artificial intelligence will reach $2.9 trillion. However, their judgment is that the largest technology companies will only be able to cover about half of this amount with their own cash flows. This means that there is expected to be a funding gap of approximately $1.5 trillion, and private credit is expected to provide a significant portion of the external funding for this need.
The data center expenses driven by AI are divided into two parts: hardware and construction.
Morgan Stanley allocates approximately $1.6 trillion to IT hardware, including chips and servers; another $1.3 trillion is used for physical infrastructure and construction. This forecast covers the computing devices required for artificial intelligence, as well as the facilities that house these devices.
Amazon, Microsoft, Alphabet, Meta, and Oracle are expected to contribute approximately $1.4 trillion through their own cash flows. In other words, these hyperscale cloud service providers will bear about half of the expected investment amount themselves, with the remaining portion requiring external financing.
Private credit plays a more significant role in construction financing.
Research by Morgan Stanley indicates that the private credit market could provide funding for about $800 billion of the $1.5 trillion gap. This type of financing will adopt an asset-backed structure: loans are secured by specific facilities or equipment, rather than by the borrower's overall balance sheet.
Similar arrangements have appeared in the financing of independent computing service providers, where lenders use graphics processors and customer software contracts as collateral. Banks also provide financing for physical construction, arranging and underwriting project loans.
The study indicates that, compared to traditional corporate bonds, asset-based private credit financing may offer higher returns. In addition to these loans, hyperscale cloud service providers have committed a total of over $1.2 trillion under leasing agreements, with such substantial future payments calculated on an undiscounted basis.
These lease payments constitute a significant portion of the financing for construction. Therefore, this expenditure forecast also exists in parallel with the substantial contractual commitments made to the facility developers.
The report also stated that NVIDIA participated in discussions regarding a maximum standby guarantee of $250 billion to support data center debt financing. These discussions make supplier guarantees, along with private loans and leases, part of the financing landscape for AI-driven data center expenditures.
Another constraint is the composition of the electric power supply in the United States.
It is estimated that by 2028, the power gap in data centers in the United States will expand to 32 to 33 gigawatts. This projected gap means that in addition to the financing shortfall, the physical power supply is another constraint faced by the advancement of construction efforts.
Morgan Stanley's investment forecast covers the period from 2025 to 2028, which is consistent with its revised forecast for the electricity shortfall in the United States.
This article was generated with the assistance of artificial intelligence and has been reviewed by an editorial team.
