Is Meta’s AI Spending The Easiest To Cut?
On Monday, Nov 16, the S&P 500 fell 1.9%, and the big AI spenders fell with it: Alphabet 4.1%, Amazon 3.3%, Meta 2.4% and Microsoft 1.1%. The question investors keep asking is the same one: these four plan about $730 billion of capital spending in calendar 2026 between them, so why not simply spend less?
The short answer: for Microsoft, Amazon and Alphabet, much of that spending is already sold. Their customers have signed contracts worth $1,816 billion that have not yet turned into revenue: Microsoft’s $677 billion of commercial remaining performance obligation, a $578 billion AWS backlog and a $561 billion Google Cloud backlog (each company measures this its own way). Cutting the build, in our view, would mean failing to deliver revenue that customers have already committed to.
The other side of the ledger is signed too. Alphabet ended the third quarter with $924 billion of future purchase commitments and another $152 billion of leases signed but not yet started. Meta is the exception: it has no cloud backlog in these numbers, and much of its build serves its own products, which makes its spending the easiest of the four to slow if investors push.

Contracts On Both Ends
A good part of it is due soon. Microsoft says roughly 30% of its backlog becomes revenue in the next 12 months, and Alphabet expects just over 55% of the Google Cloud backlog to be recognized within 24 months.
Supply is booked well in advance. Amazon CEO Andy Jassy said chunks of capacity of hundreds of megawatts, sometimes up to a gigawatt, are reserved well ahead, and Amazon says it has very good visibility into 2027, 2028 and some of 2029 for customers’ big capacity reservations. Alphabet says it has signed multiyear agreements for infrastructure, inventory, leases and energy.
Demand is not the bottleneck either. Microsoft’s finance chief said customer demand continues to exceed available supply. Alphabet began renting outside compute at the end of the third quarter because supply is tight. Amazon says it installs an AI accelerator about every 14 seconds and customers take it almost as fast as it goes in. Jassy said Amazon will invest to stay the leader.
| Company | 2026 capex guide | Capex as % of operating cash flow, September quarter | Customer backlog |
|---|---|---|---|
| Microsoft | about $170B | 67.9% | $677B |
| Alphabet | $195B-$205B | 121.4% | $561B |
| Amazon | about $220B | 182.1% | $578B |
| Meta | $135B-$145B | 103.5% | none in our data |
Capex guides use each company’s own basis; the cash-flow ratio uses Morningstar data.
Why Meta Is Different
None of the backlog above belongs to Meta. Mark Zuckerberg said Meta has not yet found the limit of how well it can apply compute to its own core business, and it is building capacity to serve more than a billion people over the next few years. That is a choice Meta can revisit. It says it will scale new AI businesses behind demonstrated customer demand, and that it is still early in setting its 2027 budget.
Meta is also paying for the build in a new way. Its free cash flow was negative $2.1 billion in the third quarter, it ended the quarter with $82.4 billion of cash and securities against $83.7 billion of debt, and it priced a $25 billion common stock offering plus mandatory convertible preferred stock with net proceeds of about $12.5 billion. If investors want less spending anywhere, Meta is where management has the most room to say yes.
The Case Against “They Can’t Cut”
The strongest counter is cash. Capex ran ahead of operating cash flow at Alphabet, Amazon and Meta in the September quarter, so free cash flow was negative at all three: Alphabet’s was negative $9.1 billion and Amazon’s negative $34.8 billion. Stocks have been moving hard after reports: Alphabet fell 12.6% in the first session after its results and Meta 7.0%, while Microsoft slipped 0.16% and Amazon 1.4%. Those moves show the mood, not their cause.
Timing can also slip. Microsoft said third-party capacity it expected in its fiscal first quarter was delayed into the second half of its fiscal year, so a contract does not lock in the quarter the money goes out. And efficiency could do some of the work: Meta aims to improve compute cost efficiency by more than 10 times over the next year, and Microsoft says engineering work let it serve more demand from CPU capacity that sat underused. If those gains spread, the same backlog could need less building.
Still, the companies are pointing the other way for now. Alphabet expects capex to rise significantly in 2027, and Amazon expects meaningfully more in 2027 than about $220 billion this year. For what this means for owners, see Has Microsoft Stock Become A Different Bet? and whether Alphabet’s spending can turn back into cash.
What An Owner Can Do This Week
A falling price tells you what other people feel; a changed business tells you something about the company. If you own one of these stocks, write down the one change that would make you sell. For the three cloud companies, a sensible test is the backlog: if it stops growing while capex keeps climbing, the reason the spending can’t be cut starts to weaken. For Meta, the test is whether the build starts producing the demand it says it will wait for. Weighing the two directly? See Why We Prefer Alphabet Stock Over Meta Right Now. This question sits inside our wider AI storyline.
Nobody can schedule the day this spending pays off or the day investors lose patience. That is why it can make more sense to be set up for either outcome than to bet on one.
Stop Guessing. Start Positioning.
No one can predict surprises like this one. The Trefis High Quality (HQ) Portfolio doesn’t try: it holds 30 quality stocks in equal weight, so no single surprise decides your result. Come see how it works.
For questions about this article or underlying research, email help@trefis.com.