AI Agent UpdateBig Tech's AI IOUs top $3 trillion — and some run for 30 years
Alphabet, Meta, Microsoft, Amazon and Oracle have disclosed more than $3 trillion in future AI spending commitments — up from about $2.4 trillion in July. The bill is no longer a forecast. It is a contract.
The News
Bloomberg reported on October 30 that the five largest players in the data-center race — Alphabet, Meta, Microsoft, Amazon and Oracle — have disclosed more than $3 trillion in future spending commitments tied to the AI buildout. That is up from about $2.4 trillion counted for four of the five companies in July, a jump of roughly a quarter in a single quarter. The figures come from the companies' own quarterly filings, not from analyst estimates: leases, buildings, energy contracts and equipment orders, added up.
Alphabet alone accounts for more than $1 trillion of the total — nearly eight times its level a year ago — and Meta for more than $800 billion. About $1.3 trillion of the total is future leases, some running as long as 30 years, and about $1.9 trillion is purchase and construction orders. In other words, roughly two-fifths of the bill is rent on buildings that, in some cases, have not been built yet, and the rest is hardware and construction already ordered.
The tally landed in the same week the market started pricing the bill instead of the dream. Alphabet lost about $500 billion of market value in a single session on October 29 after its own spending surge, and the Nasdaq fell sharply on October 30, its worst single day since June. The cash side already shows strain: Alphabet posted negative quarterly free cash flow, and Amazon negative trailing-twelve-month free cash flow. Free cash flow is the cash left after running the business and paying for equipment — the money that actually funds dividends, buybacks and debt repayment.
Three trillion dollars is not a plan anymore. It is a mortgage — and the payments start whether the AI boom pays off or not.
Why It Matters
The number matters because it converts a debate into a liability. For months, the argument about AI spending was whether the companies were investing too much or too little — a matter of opinion. A disclosed commitment is a matter of record. When a company signs a 30-year lease or orders a data center, it is promising to pay regardless of whether AI demand in 2035 looks anything like AI demand in 2026. The risk is no longer 'will they build it' but 'what if they built too much.'
The speed of the increase is the second shock. Going from about $2.4 trillion to more than $3 trillion in one quarter means the commitments are growing faster than the revenues meant to pay for them. Alphabet's share rising nearly eightfold in a year is not a gradual ramp; it is a step change in how much of the company's future is already spoken for. Investors who bought these stocks for growth are discovering how much of that growth is now pre-spent.
The third reason is the cash-flow squeeze hiding underneath. Building data centers consumes cash years before the customers arrive to pay rent on them. Alphabet's negative quarterly free cash flow and Amazon's negative trailing-twelve-month figure show the gap between spending and self-funding. As long as the companies could cover the bill from operations, the commitments looked like strength. The moment they need outside money — new shares, new debt — the market starts charging for the risk, and the whole edifice gets more expensive to hold up.
Analyst's Take
The verdict is Bearish. A $3 trillion commitment pile changes what every future earnings report means. Until now, investors graded these companies on how fast they could build. From here, they will be graded on whether the buildings earn their keep — and the leases do not wait for the answer. The companies with the biggest commitments have the least room to slow down without paying for it.
What to watch next is the financing. Meta has announced a new non-voting Class C share plan and a financing plan to raise money without adding debt, and a credit agency has just cut its outlook on the company (see our companion story). If more agencies follow, the cost of funding the buildout rises — and the $3 trillion gets more expensive to carry. The AI boom is no longer being judged on ambition. It is being judged on arithmetic.
By The Numbers
FAQ
What is a spending commitment?
A spending commitment is money a company has already promised to pay in the future — signed leases for buildings, orders for equipment, and contracts for power. It is different from a budget or a wish list: breaking these deals usually costs money, and the amounts show up in the company's official filings.
How is this different from the capital spending figures companies announce?
Capital spending (capex) is what a company actually spends in a given quarter or year. Commitments are the spending already lined up beyond that — the future bill. A company can report $50 billion of quarterly capex and still have hundreds of billions more already promised.
Why do 30-year leases matter?
Because the technology inside the buildings changes far faster than the lease. A data center signed today could be outdated several times over before its lease ends — but the rent still has to be paid. Long leases turn a bet on AI demand into a fixed cost that lasts decades.
Does $3 trillion mean they will definitely spend $3 trillion?
Not exactly. Some commitments can be renegotiated, delayed or cancelled at a price. But the total is a disclosed obligation, not a marketing number — and it only grows if the companies keep signing. The direction of travel is what rattled the market.
What are NVIDIA and Broadcom doing in this story?
They are named here as read-through tickers: they sell the chips and networking gear that fill these data centers, so a $3 trillion buildout is their order book — and a pullback in commitments would be their warning sign.
👁
Source: Bloomberg