Big Tech Didn’t Hide the AI Bill. It Scheduled It.


The nearly $900 billion lease pipeline is real. But is it a debt bomb, or tomorrow’s rent for tomorrow’s AI revenue?

While investors obsess over Nvidia’s latest chip sales, another number deserves attention: the enormous data-center lease commitments sitting in Big Tech’s financial-statement footnotes.

Microsoft, Meta, Amazon and Alphabet disclosed approximately $830.5 billion in uncommenced leases as of June 30, 2026. Meta subsequently signed another approximately $68 billion in July. That takes the combined figure to roughly $898.5 billion.

Eye-watering? Absolutely. Secret debt? No. An automatic financial disaster? Also no.

The real question is whether AI revenue will arrive quickly enough to justify the infrastructure bills coming due.

1. ASC 842: The accounting rule investors should understand

Under US accounting standard ASC 842, companies generally recognize a right-of-use (ROU) asset and corresponding lease liability when a leased asset is available for use—the lease commencement date.

A signed lease for a data center still under construction may therefore appear in the commitments footnote before it becomes a recognized lease liability.

This is not an accounting loophole. It is the accounting framework.

But investors should understand the economic distinction: a company can commit to substantial future fixed costs before those costs appear as recognized lease liabilities.

And remember: disclosed future lease payments are generally undiscounted. The eventual recognized liability is based on their present value—not the full headline figure.

2. The numbers: Who has the biggest commitments?

Figures are based on different disclosures and dates; Meta’s July addition is subsequent to its June 30 reporting date. They are not directly comparable measures of debt.

Microsoft’s $329.1 billion commitment is particularly striking: its filing says these leases will commence between fiscal 2027 and 2033, with terms of one to 20 years. Amazon separately disclosed $130.1 billion of unconditional purchase obligations, illustrating why lease and purchase commitments must not be casually combined.

Important: Do not add every purchase order, construction commitment, energy contract and lease together and call the result debt. Some commitments relate to goods and services received over time; some have conditions; and some may overlap with other disclosures.

3. The real risk: Fixed costs arrive before guaranteed returns

Here is the strongest bear case:

  • Utilization risk: Data centers still require payments if AI demand disappoints.
  • Pricing risk: AI services may grow rapidly while competition pushes prices and margins lower.
  • Power risk: A signed lease does not guarantee grid connections, transformers or electricity will be ready on time.
  • Technology risk: Accelerators and infrastructure may become economically obsolete faster than expected.
  • ROIC risk: Capital invested in infrastructure must earn more than its full cost of financing, operating and maintaining it.

There is another important wrinkle: some projects use joint ventures and special-purpose entities (SPVs). These structures can move construction financing outside the parent company's direct debt. But guarantees, residual-value commitments and minimum-purchase obligations can still leave the parent exposed.

Meta's filing, for example, describes residual-value guarantees connected with a data-center venture, including an aggregate threshold of approximately $28 billion that declines over time. Alphabet also discloses interests in unconsolidated entities involving data centers and energy infrastructure. These arrangements deserve more attention than a simple debt-to-equity ratio.

4. The depreciation debate: Earnings are not cash flow

Suppose a company buys $10 billion of AI equipment.

That is a $1.33 billion annual difference before tax, assuming no residual value.

A longer useful life increases near-term reported earnings. But it is not automatically aggressive accounting: the company must estimate how long the equipment will generate economic benefits.

The real question is whether accounting useful lives reflect economic usefulness. A GPU may still function after five years but be far less competitive.

Net income can look healthy while free cash flow (FCF) takes the punch today. Investors should examine capital expenditure, operating cash flow, FCF conversion and incremental return on invested capital (ROIC)—not just earnings per share.

5. Valuation: Don't turn footnotes into fake precision

Some draft estimates suggested that lease-adjusted enterprise-value-to-EBITDA multiples could rise substantially. The direction is reasonable, but precise company-by-company multiples are not defensible without consistent assumptions about discount rates, commencement schedules, rent adjustments and existing lease liabilities.

A better framework:

  • Start with market capitalization, debt and cash.
  • Include recognized lease liabilities, checking whether the valuation provider already includes them.
  • Estimate the present value of material uncommenced lease payments.
  • Adjust EBITDA consistently for rent if using a lease-capitalized comparison.
  • Stress-test FCF and ROIC under lower utilization and weaker pricing.

Do not add the full undiscounted commitment figure to enterprise value. That is how an analysis becomes impressive-looking—and wrong.

6. The investor watchlist

The bull case is credible too: these companies are not ordinary retailers with thin margins. Their existing businesses generate substantial cash, and securing scarce power and data-center capacity may prove strategically valuable. Uncommenced projects may also have more flexibility than facilities already operating—but cancellation rights must be checked contract by contract.

The 2018 retail lease-accounting comparison is therefore imperfect. ASC 842 changed balance-sheet visibility; it did not automatically cause a sector-wide valuation collapse.

7. The Wealth Builder checklist

Before investing in an AI infrastructure winner, ask:

  • How much is already recognized versus still uncommenced?
  • What is the estimated present value of future lease payments?
  • Are purchase, energy and construction commitments being counted separately?
  • Is the capacity contracted to customers or speculative?
  • Is power and grid interconnection actually secured?
  • Is FCF growing alongside AI revenue?
  • Does incremental ROIC exceed the full cost of the infrastructure?

Bottom line: The AI infrastructure boom may create extraordinary winners. It may also create expensive buildings, long leases and underutilized GPUs. The accounting footnotes are not proof of a bubble—but ignoring them is a great way to overestimate a company's financial flexibility.

Educational research, not investment advice. These companies are research starting points, not buy recommendations. Verify the latest filings and conduct your own due diligence.

Sources & notes

  1. Microsoft FY2026 Form 10-K, Note 13—Leases and additional uncommenced lease commitments.
  2. Meta Q2 2026 Form 10-Q, Note 9—Leases and contractual commitments; July 2026 subsequent lease commitments.
  3. Amazon Q2 2026 Form 10-Q, Notes 3 and 4—Leases and commitments.
  4. Alphabet Q2 2026 Form 10-Q, Note 4—Leases; Note 5—Variable Interest Entities.
  5. ASC 842: US GAAP lease-accounting standard. Consult the FASB Codification and company disclosures for detailed accounting treatment.

#AIInvesting #BigTech #DataCenters #FinancialAnalysis #WealthBuilding #PassiveIncome

Read the footnotes. Follow the cash. Stay curious.

Wealth Builder

Read more from Wealth Builder

Everyone is chasing the next Nvidia. But what if the bigger opportunity is the company selling the one component Nvidia's customers cannot get enough of? #FollowTheMoney #ChokepointInvesting #AIStocks #WealthBuilder The Boring Earnings Calls That Could Reveal the Next Big Opportunity Everyone knows AI is booming. Congratulations, Sherlock. 🔍 But knowing where the money is going is not the same as knowing who will make the money. During October 2026, Amazon, Microsoft, Alphabet and Meta...

How Pump-and-Dump Syndicates Turn Retail Investors into Exit Liquidity — And the 5-Candle Test That Could Save You Wealth Builder | Achieve Passive Income, Sustain Financial Freedom Imagine joining a WhatsApp investment group. The professor looks professional. Members post winning trades. Everyone is making money. Then the administrator recommends a little-known stock and asks for a screenshot of your purchase. You think it's proof of membership. Wrong. You may have just shown the people...

Wall Street's AI advantage is moving into retail investors' hands. Here's how to turn it into a smarter investing machine without letting the machine run away with your money. Wealth Builder | Achieve Passive Income, Sustain Financial Freedom 1. Meet your new investment assistant. It has ambition. And questionable boundaries. In June 2026, an internal OpenAI research model was tasked with updating a crypto-trading prototype and deploying smart contracts on Ethereum's Sepolia test network. One...