Approximately $628 billion of Meta's off-balance-sheet liabilities have long been buried in financial statement footnotes. Once this figure is incorporated into enterprise value calculations, the valuation framework for the entire hyperscaler tech sector faces reconstruction.

Investment bank Needham analyst Laura Martin published a research note on October 2 pointing out that if Meta's off-balance-sheet commitments are included in enterprise value (EV), its valuation multiple would be adjusted upward by 35%, making the stock far more expensive for shareholders than traditional metrics suggest.

Martin subsequently stated publicly on CNBC: "Meta has about $100 billion in on-balance-sheet debt and leases combined, with off-balance-sheet liabilities reaching $600 billion. If you don't include that $600 billion in EV, your valuation is 35% too low." This judgment directly challenges the market's pricing logic for Meta.

More noteworthy is that when Needham's methodology is consistently applied to Alphabet, the conclusion undergoes a dramatic reversal: Google's adjusted EV/FY27E revenue multiple reaches 9.0x, higher than Meta's 8.3x—meaning that after full disclosure of off-balance-sheet liabilities, Google actually becomes the more expensive company on valuation. Meanwhile, Meta added approximately $68 billion in new data center leases in July alone, with the pace of off-balance-sheet obligation expansion still accelerating. Whether the market has adequately priced this hidden leverage has become a core question that hyperscaler tech investors cannot avoid.

The 5.6x Gap Between Meta's On- and Off-Balance-Sheet Liabilities

Needham's analytical framework starts from fundamental accounting principles. The firm argues that "equity value is the residual value attributable to shareholders after repaying debt and other contractual obligations," and therefore regardless of accounting treatment, rising obligations and commitments constitute headwinds to stock prices and should be included in EV calculations.

Specifically for Meta, as of the end of June 2026, its on-balance-sheet debt and leases totaled approximately $112 billion, while the two categories of off-balance-sheet liabilities combined reached approximately $628 billion—a ratio of 5.6x. In other words, for every $1 of leverage disclosed on Meta's books, another $5.60 is hidden in the footnotes.

The composition of these two off-balance-sheet liabilities is clearly disclosed in footnote 9 of Meta's Q2 10-Q report: first, operating and financing lease obligations not yet commenced total approximately $279 billion, covering data centers, hosting facilities, and some network infrastructure, with lease terms extending from 2026 to 2036, the longest reaching 30 years; second, irrevocable contractual commitments total approximately $349.3 billion, primarily involving third-party cloud capacity arrangements, server and network infrastructure investments, data center construction, and Reality Labs consumer hardware products, with approximately $53.5 billion and $81.7 billion maturing in 2026 and 2027 respectively. Additionally, Meta has up to $14.7 billion in contingent cloud capacity purchase obligations.

After incorporating the above off-balance-sheet liabilities into EV, Meta's EV/FY27E revenue multiple rises from 6.15x to 8.32x, an increase of 35%.

The Accelerating Pace of AI Infrastructure Commitments

The above figures do not yet reflect the latest developments. Meta disclosed in its Q2 10-Q that in July 2026—after the quarter ended—the company signed approximately $68 billion in new data center lease obligations.

This figure means that in just one month, Meta's newly added lease commitments equaled approximately 80% of its total on-balance-sheet debt as of the end of June (approximately $84 billion), while expanding the "not yet commenced" lease obligation bucket by more than 10% before the third quarter even ended. Needham concludes based on this that Meta's off-balance-sheet obligations are showing a "rapid growth" trend in Q3.

According to research published by Goldman Sachs credit strategist Amanda Lynam in August, as of the Q2 earnings season, hyperscaler tech companies had tracked a combined $1.5 trillion in lease commitments, of which $1.0 trillion came from leases not yet commenced. Due to the special treatment under US GAAP lease accounting standards, these "not yet commenced" lease commitments do not appear on financial statements but will form actual payment obligations in future years. Lynam noted that this treatment "may understate leverage levels and future liquidity needs," and expects that "AI-related issuers will continue to bear additional spread risk premiums."

Morgan Stanley's accounting team (Todd Castagno and others) expanded the scope in late August to include Nvidia, Broadcom, and various guarantee and backstop arrangements, arriving at a total off-balance-sheet commitment figure exceeding $3.1 trillion. According to ZeroHedge tracking, as of September 30, hyperscaler tech companies' off-balance-sheet liabilities had reached $3.8 trillion, an increase of approximately $700 billion from the end of June.

Google Is Actually the "More Expensive" One: Consistency Correction to Needham's Methodology

Needham's original research note contains a notable methodological flaw: its table counts "irrevocable contractual commitments" for Amazon and Google parent Alphabet as zero, but the two companies' financial disclosures are not consistent with this.

Alphabet's Q2 10-Q disclosed $707 billion in long-term purchase commitments—consistent with the figure used by Morgan Stanley; Amazon disclosed $130 billion in unconditional purchase obligations. These commitments are of the same nature as those Needham included for Meta.

After consistently applying Needham's methodology to all three companies, the conclusions change significantly:

Alphabet: EV adjustment increases from 3% in Needham's original report to 21%, with the adjusted EV/FY27E revenue multiple rising to 9.0x, higher than Meta's 8.3x.

Amazon: EV adjustment increases from 9% to 14%.

Meta: If further incorporating the $68 billion in new July leases, guarantee exposures for the Hyperion and El Paso data center projects (approximately $46 billion and $13 billion in maximum loss exposure respectively), and $14.7 billion in contingent cloud purchase obligations, the actual adjustment would reach approximately 42%.

In other words, Meta's relative off-balance-sheet leverage scale remains the highest among the three, and Needham's core judgment holds; however, the original "35% versus 3%" comparison severely understates Google's off-balance-sheet liability scale. After full adjustment, Google actually becomes the more highly valued target on EV/revenue multiples.

Has the Market Already Priced This Hidden Leverage?

Supporters argue that purchase commitments are not debt in the traditional sense—they are server, cloud capacity, and chip contracts that Meta expects to convert into revenue. Credit rating agencies typically treat lease liabilities as debt-equivalent but generally do not afford the same treatment to purchase obligations. Additionally, Meta's disclosed $628 billion in off-balance-sheet liabilities is an undiscounted figure; if calculated on a present value basis, the EV impact would narrow somewhat.

However, counterarguments are equally compelling. Morgan Stanley explicitly warns: "These contracts have strategic value when supply is tight, but if supply and demand normalize ahead of schedule, companies may be forced to pay for excess capacity or attempt to renegotiate terms." In other words, the premise that purchase commitments are "not debt" is that the company will always need what it has committed to buy.

Free cash flow is already showing signs of stress. According to Morgan Stanley, Amazon and Google's free cash flow turned negative in Q2, and Meta is expected to enter negative territory in the next quarter.

In the bond market, Goldman Sachs data shows that hyperscaler tech companies have issued $250 billion in bonds this year, a sharp jump from less than $20 billion in 2024; the forecast issuance for 2027 is $420 billion, and this forecast does not yet include data center financing issued through special purpose vehicles (SPVs)—which is precisely the core component of current off-balance-sheet structures identified by Morgan Stanley.

Needham maintains a "Hold" rating on Meta, and its concerns extend beyond accounting. Laura Martin noted that Meta is simultaneously pursuing custom chips, data centers, enterprise AI, commercial agents, model APIs, compute sales, smart glasses, and multiple other fronts, with "management attention, engineering talent, and shareholder capital spread across too many directions," and that the margin for error has narrowed substantially against the backdrop of $628 billion in commitments.

Meta's Q3 10-Q is expected to be released by the end of October. With the $68 billion in new July leases already a fait accompli, the market already knows full well where that footnote is heading.