Moody's: $662B In Data Center Leases Hidden From Big Tech Balance Sheets Pose Looming Risk

The world’s largest tech companies are vastly understating the economic risk tied to their data center leasing deals, according to a new report from Moody’s Ratings.

At least half a trillion dollars in data center lease commitments are invisible in tech giants’ financial reporting, and the actual volume of this phantom spending is likely far higher. 

The vast sums being spent by Amazon, Microsoft, Google, Meta and Oracle on building and leasing data centers and other infrastructure to support artificial intelligence are some of the most scrutinized metrics in global finance. But according to Moody’s, hyperscalers are using accounting practices that allow them to keep the majority of their data center leasing commitments off their balance sheets and invisible to investors. 

While this spending is hidden today, it represents added liabilities due to materialize in Big Tech financials in the near term — a major looming risk in a sector still awaiting a verdict on whether its unprecedented AI bet will pay off. 

“Under the current rules it is possible to defer the reporting of this liability until the last minute, so the fear is that this indicator of the health of a project or even AI in general will be lagging,” David Gonzales, vice president and senior accounting specialist at Moody’s Ratings, said in an email to Bisnow.

Of the nearly $1T in future lease commitments from the five major hyperscalers, more than two-thirds doesn’t appear on these firms’ balance sheets due to accounting rules allowing the omission of leases signed but not yet commenced. 

That means $662B in hidden liabilities is set to appear suddenly in Big Tech financial reporting in the coming quarters — a figure greater than the total debt obligations of the five big tech firms today, according to Moody’s.

“To put that into perspective, the amounts related to leases not yet on the balance sheet is 113% of the hyperscalers' most recent adjusted debt,” the authors wrote. “Even so, these numbers do not tell the entire story.”

A shift in how hyperscale leases are structured has created even more opacity, allowing tech firms to use a separate accounting loophole to conceal billions more in committed data center spending. 

Hyperscale leases are getting shorter. While their data center lease terms have historically been between 10 and 15 years, newer leases span just four to six years to the first renewal. The change is meant to align leases with the lifespan of costly graphics processing units and other AI infrastructure. 

While these shorter leases look better on tech firms’ balance sheets, they don't provide sufficient guaranteed revenue for lenders to provide project financing for these facilities. So, to facilitate financing, hyperscale tenants “backstop”  these deals — guaranteeing payments to the data center landlord in the event the lease isn't renewed. 

This means tech firms are promising to either renew their leases or pay landlords a roughly equivalent sum. Either way, this represents a massive capital commitment. 

Yet because of vague language in U.S. accounting standards, hyperscalers often don’t have to report either the potential lease renewal or the backstop payment in their financial disclosures. It is a loophole that means tech giants are vastly underreporting their future lease-related liabilities, Moody's says. 

These are potentially hundreds of billions of dollars in additional hidden capital commitments that will rear their heads on Big Tech balance sheets when the leases expire. However, tech firms’ lack of transparency around these deals makes placing an exact dollar figure on these looming liabilities impossible, according to Gonzales.  

“When major leases are entered we may get this disclosure on a lease by lease basis but [there’s] no way to do this for a company at once,” Gonzales said. 

The takeaway, the authors say, is that hyperscale data center spending may represent a greater economic risk than tech giants’ financial reporting suggests. 

The report warns that major tech companies’ debt and lease obligations tied to AI infrastructure are poised to surge in the coming years, potentially hundreds of billions of dollars beyond what is reflected on their balance sheets. Whether intentionally or not, the full scale of Big Tech’s AI wager is being obscured from investors’ view. 

While tech firms are betting that revenue from AI products and services will ultimately justify the massive outlays for data centers and other infrastructure, hundreds of billions of dollars in off-balance-sheet data center lease commitments are already locked in. Even if demand shifts or companies slow their expansion, those contractual payments to landlords will remain.

The liabilities may be hidden today, but the cash obligations are real — and unavoidable.

“Moody’s Ratings anticipates a material increase in adjusted debt and lease-related cash outflows,” the authors wrote. “While higher earnings may mitigate this, there remains significant uncertainty regarding the long-term growth and profitability of the AI market.”