Why Big Tech Spends $1.15 Per $1 Of Cash
Follow the whole story: every piece in this series is on our AI storyline page.
What Does Everyone Already Know?
The Nasdaq 100 closed Monday more than 10% below its Oct 6 record, and was still about 9% below it at Wednesday’s close. The story everyone repeats: investors are worried that roughly $730 billion of capital spending by the four biggest AI spenders this year won’t earn its money back, so the spending will have to slow.
Here’s the problem with that story. The companies writing the checks just told us the opposite.
| Company | Q3 2026 capex | Q3 operating cash flow | Capex per $1 of cash | What they said next |
|---|---|---|---|---|
| Microsoft | $39.4B | $58.0B | $0.68 | More than $50B next quarter |
| Alphabet | $51.4B | $42.3B | $1.21 | $195–205B in 2026; up “significantly” in 2027 |
| Amazon | $77.3B | $42.4B | $1.82 | About $220B in 2026; “meaningfully more” in 2027 |
| Meta | $33.9B | $32.8B | $1.04 | $135–145B in 2026, range narrowed, ceiling kept |
Three of the four spent more than their businesses brought in last quarter, and all four said they’ll spend more.
Added together, the four spent about $202 billion in the quarter against about $175 billion of operating cash flow: roughly $1.15 for every dollar. Over the twelve months to June, the same four were spending 77 cents per dollar. In one quarter, the build-out went from expensive to more than self-funding can cover.
See more lines: how we computed the ratios
Capex is cash spent on property and equipment as each company reported it for the quarter ended Sept 30, 2026. Microsoft’s figure excludes finance leases; with them, spending was about $45.2 billion, or 78 cents per dollar. Alphabet’s cash flow was held down by a $5.2 billion EU fine and a build-up of chip inventory. Amazon’s quarterly operating cash flow comes from its cash-flow statement (the nine-month total less the first half). On Amazon’s own trailing twelve-month figures, operating cash flow was $168.3 billion and free cash flow −$42.4 billion, which puts capex at about $1.25 per dollar. Meta’s figure is its purchases of property and equipment; with principal payments on finance leases, its headline capex was $34.9 billion. Ratios are ours, from reported figures.
Why Won’t Anyone Cut First?
Put yourself in one of those four CEO chairs. If all four slow down together, everyone saves money and nobody loses ground. That’s the best outcome for the group. But if you slow down alone, and the other three don’t, you lose the customers who need computing power next year. Those customers are not a side business. For Microsoft and Amazon, they are Azure and AWS, which grew 48% and 39% last quarter. For Alphabet, the risk is search itself. For Meta, it’s the ad engine that pays for everything.
Economists call this a prisoner’s dilemma. Each player’s best move on its own is to keep spending, even though all four would be better off if everyone spent less. And there’s no way for them to agree to slow down together. So nobody blinks.
That’s why the market’s punishment hasn’t worked. Alphabet fell after it reported, and it’s now 28% below its 52-week closing high; Meta is 19% below. Neither cut its plan.
What Does Nike Have To Do With It?
The same logic now runs far outside Big Tech. In an Oct 20 piece on inflation and the risk of a crash, we made the point that every company wants to be an AI leader, and few CEOs or boards are volunteering to spend less. Nike is the clearest case. It cut about 775 jobs in January, mostly in U.S. distribution centers, as it automated, and then about 1,400 more in April, most of them in technology, while concentrating its tech work in two hubs. Nike’s stock fell 39% from March 2 to Oct 15. Even so, the cuts look aimed at paying for the shift, not abandoning it.
That’s the unexpected part. When money gets tight, companies like Nike cut people and keep funding AI. In those cases, the spending is the last line to go, not the first.
What Should You Do This Week?
Use this as a thesis test, not a market call. If you own AI stocks because you believe the build-out continues, the thesis hasn’t changed this quarter. The price has. That’s the difference between information about other people (the chart) and information about the business (the spending plans).
1. Look at the suppliers, not the cutters. The money the four giants spend lands somewhere: chips, memory, servers and power. Several of those suppliers are down sharply from their highs: Broadcom 32%, Dell 37%, Micron 34%, Constellation Energy 35%, Vistra 37% and Nvidia 16%. Companies cutting staff to fund their own AI projects, by contrast, are paying the bill, not collecting it.
2. Commit at a price instead of guessing the bottom. One example, for education only: Nvidia closed Wednesday at $200.13. A cash-secured $170 put expiring May 21, 2027, about six and a half months out, paid about $9.75 a share at Wednesday’s close (midpoint of the bid and ask), an effective entry near $160, about 20% below Wednesday’s price, or a 5.7% premium kept if the stock stays above $170. “Cash-secured” means $17,000 set aside per contract. The risk: if Nvidia falls 40%, you own it at $170 while it trades near $120.
3. Keep the size inside a band. If a theme position drifts more than 5 percentage points above your target, trim it back; if it drifts 5 below, add. That rule does the buying and selling for you, at prices you set in advance.
4. Write down what would make you sell. The thesis checklist:
- Any of the four cuts next year’s capex guidance for real, not a timing or accounting change. Microsoft’s lower calendar-year figure this quarter came from delays in third-party capacity deliveries, which it said shift into the second half of its fiscal year.
- Cloud growth falls below spending growth for two quarters running. Today Azure is at 48% and AWS at 39%.
- The giants lean on borrowing to pay for it. Today it’s mostly operating cash, though that no longer covers it all, and Amazon and Alphabet borrowed a net $49 billion and $25 billion in the quarter. The companies already leaning on debt, Oracle and CoreWeave, are down 50% and 58% in a year.
- Supplier order books shrink, not just their stock prices.
Only the third has started, at Amazon and Alphabet, so watch it most closely. Until more of them turn, a falling price is a buying range, not a sell signal.
What’s The Bigger Idea?
Sell the thesis, not the price. A 30% drop tells you how other investors feel this month. A capex cut would tell you something about the business. The four giants have just shown that they’d rather outspend the cash their businesses bring in than fall behind, which is the strongest evidence yet that the spending thesis is intact, and also the clearest warning of how it ends: when one of them finally can’t afford it, all of them will be tested at once.
AI spending can’t be stopped by a falling stock price, only by a broken business case. The four giants spent about $1.15 for every dollar of cash they brought in last quarter, and promised more. Hold the thesis, buy the drop in the suppliers at prices you set, and sell only if more of the items on the checklist turn.
What Should You Ask Next?
- Is this 2000 or 2018? Three numbers that tell an AI bubble from a correction
- Dell fell 34% in a month. If you still believe in AI spending, what’s the plan?
- How much AI do you really own? Count it in three numbers
We’re following the AI spending story every week, with the numbers as they land. Follow along, and tell us the next question you want answered.
Is Your Portfolio Betting On The Spenders Or The Suppliers?
Corrections test concentration, not conviction. If you have $1 million or more invested, see how exposed your portfolio is to the AI trade in a free 15-minute call with a Trefis partner firm.
For questions about this article or underlying research, email help@trefis.com.