Off-balance sheet lease commitments have quickly become a major factor in the artificial intelligence build-out, and in some respects is even dwarfing the on-balance sheet debt raised by hyperscalers.
With the exception of SpaceX, off-balance sheet exposure is now larger than on-balance sheet obligations for all the hyperscalers including Alphabet, Amazon, Meta, Microsoft, Nvidia, and Oracle. In some cases its 4x-6x larger.
Across this group of hyperscalers, chip makers and satellite operators, off-balance sheet commitments through the latest earnings results totaled $2.7trn — 77% of the group's $3.5tn of aggregate obligations.
For example, Microsoft's off-balance sheet lease commitments have jumped from $92.7bn to $329.1bn in one fiscal year, underscoring the scale of AI infrastructure spending that’s now being funded outside headline debt metrics. In fact, a new deal from data center developer QTS is slated for next week to fund Microsoft’s latest site.
These off-balance sheet figures capture not-yet-commenced leases and unconditional purchase and supply commitments, predominantly related to AI and cloud infrastructure.
Not all of these lease obligations are funded in public debt markets, but when they are, it’s usually a much smaller data center development company that allows the debt...
