According to Chao Xiang Research, a Morgan Stanley report points out that the total off-balance-sheet commitments disclosed by hyperscale vendors, Nvidia, and Broadcom have exceeded $3.1 trillion. Among them, lease commitments amount to $1.1 trillion, procurement commitments to $1.7 trillion, and the total on-balance-sheet debt and lease liabilities of hyperscale vendors is $770 billion.

Amazon and Google’s free cash flow turned negative in the second quarter of 2026, and Meta is expected to follow suit next quarter. Financing tools are restructuring the capital structure of AI computing power from six aspects: off-balance-sheet lease and procurement commitments constitute the first layer of financing, the proportion of debt issuance rises from 2% in 2025 to 19% in 2026, Google frees up $110 billion by cutting buybacks and issuing equity, Oracle’s customer prepayments in the second quarter amount to $4.6 billion, and Broadcom and Nvidia launched a chip leasing SPV to support unrated AI labs.

Morgan Stanley believes that as the financing structure itself becomes a core variable of the AI supply chain, the importance of tracking off-balance-sheet commitments and changes in accounting judgments is approaching that of tracking chip shipments themselves.