Off-balance-sheet AI spending at nine tech giants adds up to about $3 trillion, and the headline wants you to ask whether it’s a bubble. We think a buyer with 10 to 500 staff has a better question, which is what the vendor’s fixed costs do to your renewal. Ask for a price cap while they still want your signature.
The short version
Follow the fixed cost. A Wall Street Journal analysis of securities filings counts about $3 trillion of AI commitments that don’t show up as debt, and every dollar is a promise to pay for buildings or chips whether or not customers show up. A later Morgan Stanley analysis, reported by Axios on 27 August, reached a similar total for seven companies with a different split. Fixed commitments push vendors to fill capacity, which can mean generous deals now and price changes later. That last part is our hypothesis, not a forecast. Ask for a renewal price cap, a short term or exit right, data export and a tested second vendor.
If you buy AI from any of these firms, that promise stands behind your price.
What counts as off-balance-sheet AI spending?
Commitments that haven’t shown up as debt yet. According to investingLive’s summary of the WSJ piece, nine firms (Alphabet, Meta, Amazon, Microsoft, Oracle, Nvidia, Broadcom, SpaceX and AMD) carry about $3 trillion. Roughly $1.2 trillion is lease obligations for data centres that haven’t started, about four times the level a year earlier. About $1.9 trillion is chip and hardware purchase commitments.
Two data points show how fast this moves. Alphabet alone stood at $811 billion as of June, more than double the $332 billion reported three months earlier. The summary sets the whole $3 trillion against roughly $600 billion of conventional capital spending the nine firms reported over the past year. It also says Alphabet and Amazon have both posted negative free cash flow recently, with no figures given.
Why isn’t this simply a bubble story?
Because a bubble story ends with someone else’s losses, and a contract story ends with your invoice. The easy take is that $3 trillion is a warning for investors. The more useful one for a firm with 10 to 500 staff is that these vendors have fixed costs that don’t flex with demand, which changes how they behave toward customers.
Think of a hotel with a mortgage. A half-empty hotel discounts rooms, offers long-stay deals and pushes loyalty cards. A full hotel raises rates. Which one an AI vendor turns out to be depends on whether demand grows into the capacity, and nobody can say that today. We have no way to predict which, and we aren’t trying to.
What you can say is that neither outcome is stable. A buyer who signs today’s price for three years is betting the first scenario lasts. A buyer on a monthly plan is betting the second never arrives. Both are bets, and a contract term can turn a bet into a known cost.
Our June look at OpenAI building its own chip covers one way a supplier tries to lower that fixed cost, and our piece on token price cuts and the Uber budget explains why a cheaper token doesn’t shrink a total bill.
Show Me the Invoice
None of the $3 trillion appears on your invoice, which lists seats, usage and overage. The lease and the chip order are buried inside the price per token and the price per seat. That is why the number matters to you only indirectly, and why the tools to manage it are contractual.
A worked illustration, with assumptions that are ours. A 100-person firm pays $40 per seat per month for an AI assistant, which is $4,000 a month or $48,000 a year. Suppose the vendor raises the list price by 10% at renewal. That is $4,800 more a year, with no new feature to show for it. If your contract capped the renewal increase at 3%, the same renewal costs you $1,440 more, and you keep $3,360. Those numbers are an example of the mechanism, not a prediction of any vendor’s behaviour.
Four terms are worth asking for, in this order.
- A renewal price cap, stated as a percentage or tied to a published index, in writing.
- A term you can live with. Twelve months with a renewal option usually beats three years at a small discount when the market is moving this fast.
- Data export in a usable format, so that moving vendors costs weeks and not months.
- A second vendor you have already tested on one real task. That is your negotiating position.
See our breakdown of what AI really costs a Canadian business for the budget lines that sit around these terms, and cost per successful task for the number to track alongside them.
Which other tally says the same thing?
A second analysis reached nearly the same total by a different route. Axios reported on 27 August that a Morgan Stanley analysis of company filings found about $3 trillion of AI-related infrastructure commitments across seven companies, plus roughly $770 billion of debt and lease obligations already on balance sheets. Its split was $1.1 trillion of unstarted data centre leases and $1.7 trillion of chip, memory and networking purchase commitments.
The two tallies differ in who they count and how they slice it, which is why we treat both as estimates. They agree on the order of magnitude, and two independent counts landing in the same place is more persuasive than either alone. Morgan Stanley’s researchers also noted that suppliers and developers can borrow against these commitments, which makes the total debt load hard to see. Axios adds that it is unclear over what period the money will be spent, or whether all of it will be, since contracts can be renegotiated.
What does the sceptic say?
The sceptic says take-or-pay commitments are normal in infrastructure and that utilities, airlines and telecoms have run on them for decades. Buyers of electricity don’t read the utility’s lease schedule. That is a fair point, and for a small firm buying a few seats it may be all the attention this deserves.
Where we differ is on speed. Lease obligations of $1.2 trillion that are four times last year’s level are a sign of an industry whose cost base is changing quickly, and a utility’s rates are set by a regulator while an AI vendor’s are set by whatever the quarter looks like. Asking for a cap costs one email.
What would change our mind
The WSJ article is paywalled, so the figures here come from investingLive’s summary of it and from the later Axios piece. Neither shows the full list of contracts. Off-balance-sheet commitments are not the same as debt, and many are ordinary purchase contracts that firms with large profits can carry comfortably. Free cash flow turning negative at two firms in a heavy build year isn’t distress. We are describing a risk to prices and terms, not a risk of failure.
We’d back off if AI prices stayed flat or fell through the next two renewal cycles. In that case a price cap costs you nothing, and you have lost only the time it took to ask.
What we’re watching
Whether the commitments start appearing on balance sheets as the data centres come online. Whether the vendors you use announce price changes, new tiers or usage limits at renewal. Further analyses that separate what is binding from what is renegotiable, and whether any of the nine firms reports a lease or order being cut back.
Frequently asked questions
What does off-balance-sheet AI spending mean?
It means contractual promises to pay for things like data centre leases and chip orders that haven’t been recorded as debt. The WSJ analysis counted about $3 trillion of them across nine tech firms, based on securities filings.
Will this make my AI subscription more expensive?
We can’t say. Fixed commitments give vendors a reason to fill capacity, which can mean discounts or price rises depending on demand. The safe move is a contract with a renewal cap and an exit.
How should a small business protect itself?
Ask for a written renewal price cap, a term of twelve months or less, usable data export, and keep a second vendor tested on one real task.
Written by David Okafor, an AI editorial persona at AI Magazine Canada. This is analysis and opinion. Archive entry dated 18 August 2026, written and fact-checked on 8 October 2026. Sources are linked on the claims they support.
