Here is a headline that would have sounded absurd a year ago: two of the biggest spenders in the AI buildout came in under what Wall Street expected them to spend. Alphabet reported capital expenditures of $51.36 billion against an analyst consensus of $55.09 billion. Microsoft reported $45.2 billion in capex and finance leases against a consensus of $50.7 billion, according to 24/7 Wall St. If you were worried the AI spending boom was a bubble about to hiss, you might read that as the first sign of air coming out. I think that reading misses what the companies were actually saying.
Spending less because they could not build faster
The reason for the shortfall matters more than the shortfall itself. Microsoft attributed its lighter number to delayed third-party capacity, which pushes some deliveries into the second half of its fiscal year. CFO Amy Hood said demand “continues to exceed available supply,” per the same 24/7 Wall St. report. Read that sentence twice. A company that wanted to spend more, had customers waiting, and could not get the buildings, chips, or power delivered on time.
That is a very different story from a company pulling back because the orders dried up. It is a supply chain story. And supply chain stories have a way of moving money around rather than making it vanish. A delayed data center does not cancel the revenue it was meant to serve. It postpones it, and it postpones the spending that goes with it.
Amazon offered the clearest evidence of the demand side. The company reported on October 29 that AWS grew 39.3% year over year, which CEO Andy Jassy described as the sixth straight quarter of acceleration and the fastest pace in nearly five years. AWS revenue was roughly $46 billion, with operating income of $17.8 billion, according to Amazon’s earnings release. Amazon says its AWS backlog stands at $578 billion and is growing at triple-digit rates. Jassy put the capacity problem in plain language: “almost as fast as we install capacity, customers are gobbling it up.” Those figures come from Amazon’s own summary of the call, so they are company-reported, but the AWS revenue and growth numbers line up across independent coverage too.
The market is not rewarding the good numbers
If demand is this strong, why did the stocks flinch? Alphabet’s Google Cloud revenue grew 103% in the quarter, yet the shares fell about 12% at midday on October 29, partly on an earnings miss that included an $8.7 billion loss on equity securities. Amazon’s stock dropped roughly 7% after its report, according to Vested Finance. Microsoft slipped about 2% midday. Meta was the lone rise among the group in that October 29 session, up about 3% — after an earlier report that showed third-quarter revenue growth of 26.1% to $64.60 billion.
Notice the pattern. The market is no longer paying for growth alone. It is paying for growth that comes with a believable path to cash. That distinction sounds subtle, but it is the whole game right now.
The cash flow math is getting uncomfortable
Here is the number that explains the nerves. Amazon’s trailing twelve month free cash flow was negative $42.4 billion in the third quarter, compared with a positive figure of roughly $15 billion a year earlier. Amazon expects about $220 billion of capital spending in 2026. One analysis cited by Vested Finance, which is an estimate and not company guidance, suggests that implies fourth-quarter spending well below the third quarter’s pace, and that either spending cools sharply or Amazon raises its forecast later.
Zoom out and the picture is the same across the group. Forbes contributor Jason Kirsch in early October pegged combined 2026 capex for Amazon, Microsoft, Alphabet, and Meta at roughly $700 billion to $725 billion, up about 75% from a record $410 billion last year. He notes that several Wall Street firms expect the total to top $1 trillion in 2027, and that capex is projected to consume nearly all of these companies’ operating cash flow this year. When a business spends close to every dollar it earns on one category, it has a very large bet on the table, and investors are entitled to ask when it pays off.
There is a second-order effect worth watching too. Kirsch relays remarks from Fed Chair Kevin Warsh, who said in September that competition for capital tied to the AI buildout is one reason Treasury yields have climbed. If that framing holds, the spending is not just a corporate story. It reaches the borrowing costs that touch everyone from homebuyers to small businesses. A buildout this size competes for money in the same pool as everything else.
Two ways to be wrong
Investors can make two opposite mistakes here. The first is to treat every capex figure as proof of a coming payoff. Spending is a cost, not a result. Depreciation on all this equipment will arrive in the income statements whether or not the demand shows up on schedule. Jassy himself acknowledged “significant depreciation from the capex we’re investing,” and that line deserves as much attention as the backlog number next to it.
The second mistake is to treat a spending miss as a warning of collapse. When Microsoft says demand exceeds supply and Amazon says customers are consuming capacity as fast as it lands, a light capex quarter looks less like retreat and more like a traffic jam. The backlog figures reinforce that. Amazon’s $578 billion is a commitment number, not cash in the bank, and commitments can change. Still, it is hard to describe a backlog that size, growing that quickly, as a sign of vanishing interest.
So which mistake is the market more likely to make over the next few quarters? My view is that it will lurch between the two, sometimes within a single week. Alphabet’s 12% drop on a 103% cloud growth rate tells you how jumpy sentiment has become. That volatility is information. It says the easy phase, where any AI announcement lifted everything, is over, and the phase where results are sorted by cash discipline has begun.
What to watch from here
A few signals deserve a place on any investor’s checklist. Watch whether delayed capacity actually arrives and whether the postponed spending shows up in later quarters, because that tells you whether the shortfall was timing or restraint. Watch free cash flow alongside capex, since the gap between them is where financing pressure builds. Watch whether companies lean more heavily on debt or leases to keep the buildout going, a theme that has run through the AI conversation for months. And watch the cost side, since Microsoft has said about $25 billion of its 2026 budget reflects higher component pricing, which means some of the spending growth is simply things costing more, not more things being built.
None of this settles whether the buildout will be remembered as a triumph or an overreach. Past technology cycles produced both durable winners and painful stretches of overbuilding, often in the same decade. The honest answer is that the data right now supports two readings at once: demand that is real and capacity that is scarce, alongside cash flows that are stretched thin. Both can be true. The investors who handle this period best will likely be the ones who can hold both ideas without picking a side too early, and who treat each earnings season as a fresh test rather than a verdict.
This article is written for educational and informational purposes only and does not constitute financial or legal advice. The views and analytical frameworks presented draw on publicly available information and reported commentary from industry participants. Readers are encouraged to consult primary sources and form their own informed views on these complex topics.
