Meta's guarantee rose 47.7% in nine months while cash and marketable securities increased just 1.05%.
The Federal Reserve's more than $3 trillion figure is the undiscounted maximum nominal value of off-balance-sheet obligations; it is not financing raised. Meta's maximum residual-value guarantee already equals 50.3% of its cash and marketable securities. We would put the guarantor's liquid assets at the center of this story. The Fed's 13 November 2026 Financial Stability Report counted leases not yet commenced above $1 trillion and take-or-pay contracts plus residual-value guarantees above $2 trillion across seven large technology firms. Governor Michael Barr called the total the "potential maximum nominal value" on 16 November in his speech. The report separately put advanced-computing debt issuance above $680 billion by October 2026.
The $3 trillion matters because it maps the companies standing behind weaker-credit projects. The Contractual Backstop is the mechanism by which a strong-credit company makes a weaker project financeable through leases, take-or-pay commitments and residual-value guarantees, while keeping the loss if the project fails. Its test is the maximum guarantee divided by cash and marketable securities. The Fed describes the same chain: long-term contracts, leases and guarantees let strong-credit firms support projects developed by weaker-credit firms without an immediate, direct balance-sheet effect. When tenant cash flows and collateral values weaken together, the project's funding problem can become the hyperscaler's obligation.
Meta supplies the clearest measure of that backstop. Its 30 September 2026 10-Q lists a maximum residual-value guarantee of $41,493 million, against $82,449 million in cash and marketable securities. $41,493 million over $82,449 million is 50.3%. At 31 December 2025, the corresponding figures were $28,091 million and $81,592 million, making the ratio 34.4%. The guarantee rose $13,402 million, or 47.7% ($13,402 million over $28,091 million), in nine months. Liquid assets rose only $857 million, or 1.05% ($857 million over $81,592 million). The liquidity base barely changed while the maximum promise expanded sharply.
Meta disclosed $355.73 billion of leases not yet commenced at 30 September 2026; Oracle disclosed $288 billion of additional lease commitments at 31 August, substantially all for data-center arrangements. Together, $643.73 billion is roughly 64% of the Fed's $1 trillion threshold for seven firms. The $643.73 billion total combines Meta's 30 September and Oracle's 31 August figures, with the Fed comparison dated 2026:Q2, so it is a cross-date approximation. Oracle's leases generally run 15 to 19 years; Meta's range from more than one year to 30 years. Oracle also reported $664 billion of company-wide remaining performance obligations at 31 August, across cloud contracts. The lease commitments are large enough to matter without treating the RPO as evidence of AI demand.
Meta's cash generation gives the resilience argument its strongest footing. In the nine months to 30 September 2026, Meta produced $96,873 million of operating cash flow against $83,055 million of capital expenditures. That left a $13,818 million surplus. The Fed says hyperscalers' "strong cash flows have reduced their need to issue debt" for similar investment levels, at least until this year. The Fed also says the largest technology firms may absorb downturn losses, "mitigating their effects on the rest of the financial system." Meta says "RVG payments are not probable," and has recorded no liability to date. The cash cushion makes the backstop more credible, while leaving its size as a separate risk.
Van Nieuwerburgh's numbers give the buildout its strongest long-range defense. Stijn Van Nieuwerburgh of Columbia Business School writes: "It would be premature to conclude that AI infrastructure already poses systemic risk comparable to earlier credit booms." His Fall 2026 Brookings paper projects $10.3 trillion of AI investment from 2025 through 2032. It says the mature industry would need $3.725 trillion of annual revenue by 2032. That is a case for building ahead of utilization. Early capacity is installed before the revenue base reaches scale, and the forecast assumes a market large enough to fill it. In that world, the guarantees are insurance behind productive infrastructure.
Meta's nine-month figures put a limit on that reassurance. The $13,818 million cushion over capex was nearly matched by the $13,402 million increase in the maximum guarantee. Cash and marketable securities advanced only $857 million. The guarantee-to-liquidity ratio moved from 34.4% to 50.3% while the company generated cash, so resilience and exposure expanded together. Strong cash flow reduces the chance that a sponsor needs to borrow to keep building. It does not prevent its contingent promise from growing faster than the liquid assets behind it. The Contractual Backstop is becoming a larger claim on a largely unchanged liquidity base.
Lucent Technologies gives the mechanism an adverse precedent, with an important difference. Customer-financing commitments fell from $8.1 billion at 30 September 2000 to $5.3 billion at 30 September 2001. The drawn portion rose from $2.0 billion to $3.0 billion. Yet fiscal 2001 provisions for uncollectibles and customer financings reached $2.249 billion, including about $1.3 billion tied to three projects, among them Winstar and One.Tel. Revenue fell from $28.904 billion in fiscal 2000 to $21.294 billion in fiscal 2001. Lucent lent cash to customers, while the hyperscaler structure is largely leases and take-or-pay contracts with customer revenue behind them, so the plumbing differs.
The Fed itself compares this buildout with the telecom cycle that began in 1995, noting that current investment has been accelerating. The shared mechanism is that a stronger supplier can absorb losses when the customer's economics fail. Lucent's 2001 filing records what that outcome looked like.
That support also creates a second-order risk. A lease or take-or-pay commitment can bring marginal data-center capacity into existence before its standalone economics work. If demand shifts to a newer generation, the hyperscaler can remain committed to facilities it no longer needs. That reduces room for the next wave of spending. If several tenants and projects weaken together, the guarantors face correlated calls on the same balance sheets whose credit enabled the buildout. The Fed's warning is about economic exposure larger than on-balance-sheet liabilities. Meta's ratio gives one observable measure of that transfer.
We would demand a margin of safety when underwriting projects whose economics depend on the Contractual Backstop, because the guarantee now equals half the sponsor's liquid assets. Meta's 2026 Form 10-K, due by 1 March 2027 under the SEC's 60-day rule for large accelerated filers, would prove us wrong if it reports a maximum residual-value guarantee below $41,493 million or a guarantee-to-cash-and-securities ratio below 50%.