📌 AI Compute Investment Begins Shifting Toward External Financing
The funding sources for AI data centers may no longer rely primarily on the on-balance-sheet capital expenditures of cloud providers such as Microsoft, Google, Amazon, and Meta. Futurum Equities has raised its forecast for cumulative data center capital expenditures through 2030 to USD 12.5 trillion, and expects that, starting in 2027, investment from Tier 2 operators and third-party financing will exceed that of hyperscale cloud providers.
What makes this forecast noteworthy is not only the higher total investment, but also the changing entities bearing construction costs and financing risks. If funding is provided by independent platforms, alternative asset managers, and other third parties, GPUs, data center assets, and long-term compute contracts could become the basis for financing. Cloud providers may still be the demand side for compute, but need not keep all assets and debt on their own balance sheets alone.
Nvidia’s signing of memoranda of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, with plans to gradually leverage more than USD 500 billion in third-party capital through multiple independent compute financing platforms, provides a possible implementation path. However, memoranda do not mean that funds have already been deployed, nor do they mean that project risks have already been absorbed by the market.
For relevant companies, data center capital expenditure figures alone should not be the only focus going forward. More critical are whether platforms can secure sufficiently long-term contracts, whether financing costs can be controlled, whether asset utilization can cover fixed costs, and who ultimately bears default losses when demand weakens. These variables will determine whether external funding expands effective demand or merely shifts risks from cloud providers to new capital providers.
▌ Sources