Markets opened November with one question hanging over every trading desk: how long can Big Tech keep pouring money into AI data centers? Finimize framed the week that way on November 2, noting that the Nasdaq limped out of a rough October while the price of insuring against corporate defaults ticked higher. The AI capex boom now has a price tag near $828 billion for 2026, according to Fortune’s tally of the five biggest spenders, and the latest earnings season gave investors their first real reason to doubt the payoff timeline.
The doubt is not abstract. Alphabet fell 12% in a single session after its results, Amazon dropped 7% and Meta slid 10%, TechCrunch reported on October 30. The Nasdaq closed down 2.9% in the sell-off, and our own tracking of the month, in the Nasdaq 100’s $2 trillion October wipeout, shows how quickly sentiment turned. This analysis walks through the spending numbers, the cash-flow strain, the debt that now funds the buildout, and what the next 12 to 24 months could look like.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Why Markets Question How Long the AI Capex Boom Can Last
For two years the rule on Wall Street was simple. Spend more on GPUs and data centers, and cloud revenue would follow. That rule held through most of 2025 and the first half of 2026. What changed in late October is that spending outran the cash these companies generate, and the gap showed up in the quarterly filings rather than in analyst footnotes.
Finimize described the mood as unease about duration, not direction. Few investors argue that AI demand has vanished. The argument is about whether $800 billion a year in hyperscaler spending, which Finimize says analysts expect this year, can climb toward $1.2 trillion in 2027 without cash flow, credit markets or power grids pushing back. Goldman Sachs strategist Ben Snider, cited by Finimize, said earnings growth looks strong but is no longer accelerating. That detail matters. It suggests spending could stay high even if the pace of new AI investment cools.
The practical question for readers is therefore narrower than the headlines suggest. Nobody is asking whether the boom ends this quarter. They are asking which constraint binds first: free cash flow, financing costs, chip and power supply, or customer demand. Each of those has a different clock, and the evidence from the past week points to cash flow and financing as the nearest deadlines.
What Happened During the Big Tech Earnings Week
Four of the five largest AI spenders reported within days of each other. The results landed on a market already nervous after weeks of AI-related selling. According to TechCrunch, Alphabet’s stock suffered its worst day on record, erasing roughly $500 billion in market value. Fortune reported that Nasdaq 100 companies together lost more than $1 trillion in market value during the rout, and the S&P 500 fell 1.8% on the day the Nasdaq dropped 2.9%.
The damage spread beyond the spenders themselves. TechCrunch listed Nvidia down 4%, Broadcom down more than 5% and Qualcomm down 11% in the same stretch. When the biggest buyers of AI hardware look financially stretched, the suppliers feel it immediately. A cut to a customer’s capex plan is a cut to a vendor’s revenue forecast.
CNN’s market coverage on October 30 described the Nasdaq as having its worst month in more than a year. By the Nasdaq 100’s measure, the index fell more than 5.8% in October, according to a market roundup circulating before the weekend. October ended with tech investors asking whether they had priced the buildout correctly, not whether AI was real.
The 2026 AI Capex Numbers, Company by Company
Fortune put combined 2026 capital spending for Amazon, Alphabet, Microsoft, Meta and SpaceX (which includes xAI) at up to $828 billion. Strip out SpaceX and the four classic hyperscalers land above $740 billion, per TechCrunch. Estimates vary by outlet because companies guide in ranges and some analysts count leases differently, so the table below uses the Fortune figures and flags where TechCrunch adds quarterly detail.
| Company | 2026 capex guidance | Latest quarterly detail |
|---|---|---|
| Amazon | $220 billion | Negative free cash flow of $42.4 billion over 12 months |
| Alphabet | Up to $205 billion | $51.4 billion of Q3 capex, spending expected to rise again in 2027 |
| Microsoft | About $170 billion | $45.2 billion in the quarter, below the $50 billion it projected |
| Meta | Up to $145 billion | Free cash flow of negative $2.1 billion in Q3 |
| SpaceX (incl. xAI) | Up to $88 billion | Included in Fortune’s $828 billion total |
Note the Microsoft line. The company spent less than it planned because of capacity delays at third-party providers, according to TechCrunch. That is a supply-side limit, and it shows that money alone does not turn into compute on a schedule. Earlier in the cycle our readers saw the same pattern in Big Tech’s $720 billion data center bet, where the guidance kept climbing faster than physical delivery.
Free Cash Flow Is Where the Strain Shows
Capex headlines get the attention, but free cash flow tells you whether a company can sustain the spending without outside money. By that test, only one of the five could say yes in the latest period. Fortune reported that Microsoft stayed positive while the other four did not. TechCrunch put numbers on the divide.
| Company | Free cash flow | Comparison point |
|---|---|---|
| Microsoft | Positive $18.6 billion (latest quarter) | Only positive result among the five |
| Alphabet | Negative $9.1 billion (Q3) | Second straight negative quarter |
| Amazon | Negative $42.4 billion (trailing 12 months) | Positive $14.8 billion a year earlier |
| Meta | Negative $2.1 billion (Q3) | Positive $10.6 billion a year earlier |
Amazon and Alphabet Carry the Biggest Swings
Amazon’s swing from positive $14.8 billion to negative $42.4 billion is the sharpest on the list. We covered the mechanics in Amazon’s negative $42.4 billion free cash flow, and the figure has only become more central since. Alphabet’s second consecutive negative quarter turns a one-off into a pattern, which is what investors punished.
Microsoft Is the Exception, For Now
Microsoft’s positive cash flow gives it room the others lack, though Fortune noted it also announced a $25 billion bond offering on Thursday. A company can be free-cash-flow positive and still choose to borrow. The more relevant signal is that even the strongest balance sheet in the group is tapping debt markets.
Cloud Growth Is Real but Uneven
Bulls have a legitimate counterargument, and it sits in the cloud revenue lines. Fortune reported Azure growing 48% and AWS expanding about 39%. TechCrunch added that Meta’s advertising revenue rose 26%. Demand for AI services is not imaginary, and the backlog of customer commitments keeps growing. Our breakdown of Azure’s 48% growth to $33.2 billion shows why Microsoft escaped the worst of the selling.
The weak spot is Google Search. TechCrunch reported 12% growth in Search advertising, the slowest in five quarters, and described Google’s AI Mode as eating into that revenue. That is the uncomfortable pattern: a company spends tens of billions to build AI products that may cannibalize its own most profitable business. Meta’s CEO Mark Zuckerberg, per TechCrunch, described the company’s new Muse agent as very compute-intensive to run, which is another way of saying margins are not guaranteed.
Growth of 39% to 48% at the cloud level would be extraordinary in any other era. In this one it has to justify spending that grows even faster. The test is whether revenue growth can stay ahead of depreciation, which will start hitting income statements as this year’s servers age.
Debt Is Now Funding the Buildout
When cash flow stops covering capex, borrowing fills the hole. TechCrunch reported that Microsoft, Alphabet, Amazon, Meta and Oracle raised a combined $326 billion in debt in 2026, triple the prior year’s amount. Beyond that, Fortune reported that Meta announced a share issuance of up to $95 billion, so equity markets are being asked to help as well.
The wider AI sector shows the same direction. Morgan Stanley estimated that AI-related companies have raised $466 billion this year, more than four times the $101 billion raised over the same period in 2025. Those numbers cover a broader group than the hyperscalers, including chip designers, neocloud providers and model labs. Even so, the shape of the curve is clear. Financing, not revenue, has become the marginal source of capex funds.
Anthropic’s IPO filing, covered in Fortune’s report, underlines the point from the model-lab side. Fortune said Anthropic carries $265 billion in future minimum purchase commitments, and TechCrunch relayed a warning from Blackstone that the company’s debt might earn a high-yield rating after listing. The labs that buy the compute are leveraged to the same question as the clouds that sell it.
Credit Markets Begin Pricing the Risk
Credit default swaps, the contracts that pay out if a borrower defaults, are the market’s cleanest read on stress. Finimize noted that their prices rose heading into this week. A separate tally of trading data put the net notional of outstanding swaps on major tech names at a record $12.5 billion, and reported that the cost of protecting debt from issuers ranging from Broadcom to Oracle reached levels more typical of junk-rated borrowers. Treat that last figure as one outlet’s calculation rather than an official index reading.
What matters is the direction. In 2024 and early 2025, hyperscaler bonds traded like utilities. Now investors want insurance. Rising hedging costs raise the effective price of every new data center, because the borrower pays more to fund it. If spreads widen further, projects that cleared internal return hurdles last year may not clear them next year.
Finimize also flagged the sensitivity of long-duration tech shares to discount rates. Stocks priced on profits far in the future fall hardest when borrowing costs rise. That links the AI trade directly to bond yields, so a surprise from the Federal Reserve or the Treasury market can hit AI valuations without any change in AI demand.
Goldman’s Read: Strong, but No Longer Accelerating
The most useful framing in the Finimize piece comes from Goldman Sachs strategist Ben Snider. He argued that earnings growth at the hyperscalers is strong but has stopped accelerating. That distinction is easy to miss. A market that paid for acceleration will punish deceleration even if absolute growth stays high.
The same logic applies to capex. Spending growth was running near 75% this year, with analyst estimates putting the four main hyperscalers between $700 billion and $725 billion against roughly $410 billion in 2025. Wall Street projections point above $1 trillion for 2027. Growth from $725 billion to $1 trillion is a 38% rise, which would be a slowdown in rate even as it adds the largest dollar increase in the history of corporate investment.
That is Snider’s point in practice. If capex growth slows from 75% to the high 30s, suppliers such as Nvidia, Broadcom and memory makers can still post record sales, yet their stocks may fall because the second derivative turned negative. Markets trade the change in the rate of change, and that is exactly what late October looked like.
Historical Context: How Past Infrastructure Booms Ended
Every large infrastructure build, from nineteenth-century railroads to the fiber networks of the late 1990s, followed a similar arc. Capital flowed in because the underlying technology was real and demand was growing. Competition then drove returns down, financing tightened, and the weakest builders failed while the survivors inherited cheap assets. The technology survived and the investors in the middle of the cycle often did not.
The telecom comparison is the one analysts reach for most often, and it has limits. In the fiber era, much of the spending came from leveraged upstarts with no profits. Today’s spenders are among the most profitable companies ever created, and Microsoft still produces positive free cash flow. By contrast, the share of spending funded by new debt is rising quickly, and the customer base includes labs that have yet to turn a profit. The balance sheets at the top are strong, but the ecosystem beneath them is more fragile.
Our earlier work on the cash-flow squeeze, in Alphabet’s drop on its $9.07 billion cash burn, noted that the spending cycle looks different from past ones because the purchases depreciate fast. A fiber cable lasts decades. A GPU cluster may need replacement in five or six years. That shortens the window in which investments must pay back.
Competitive Comparison: Who Can Afford to Keep Spending
The five spenders do not face the same constraints. Microsoft has the cleanest cash position and the fastest-growing cloud in the group. Amazon has the largest absolute capex plan and the deepest cash drain. Alphabet has a profitable core business under pressure from its own AI products. Meta has no cloud business to sell compute to outsiders, so every dollar it spends must pay back through ads and new products. SpaceX, with xAI inside, is the newest entrant to the capex club and has the thinnest track record as a public-market borrower.
| Name | Reported stock move in the sell-off | Growth signal |
|---|---|---|
| Alphabet | Down 12% in one day, about $500B lost | Search ads up 12%, slowest in five quarters |
| Amazon | Down 7% | AWS up about 39% |
| Meta | Down 10% | Ad revenue up 26% |
| Nvidia | Down 4% | Supplier to all five spenders |
| Broadcom | Down more than 5% | Custom-chip supplier |
| Qualcomm | Down 11% | Not a core capex supplier |
| Nasdaq | Down 2.9% at the close | Index-level worst month in over a year |
Meta illustrates the pressure best. It reported negative free cash flow, plans up to $145 billion in capex, and announced an equity issuance of up to $95 billion. We unpacked the earlier version of that bet in Meta’s $34.91 billion capex quarter, and the numbers have only grown since.
Market Impact for Chips, Memory, Power and Software
The knock-on effects reach well past the five names. Chip suppliers depend on hyperscaler orders for a large share of revenue, so any slowdown in capex growth reaches their guidance first. The 4% drop in Nvidia and 5%-plus drop in Broadcom show how closely their shares track customer spending plans.
Memory and power are the quieter constraints. Microsoft’s miss against its own capex projection came from third-party capacity delays, which points to bottlenecks in construction, grid connections and equipment. A boom limited by physical supply cannot run faster than that supply even if financing is cheap. That cuts both ways. It slows spending, which hurts suppliers, but it also protects against overbuilding in the near term because the capacity simply cannot arrive all at once.
Software and application companies are in a different position. They benefit if compute gets cheaper, which happens if the buildout overshoots. Enterprises and developers buying AI services therefore have an unusual stake in the outcome: a mild glut would lower their costs, while a sudden pullback in capacity could raise them. For the broader picture of who wins and loses in model and chip markets, see our AI and machine learning coverage.
Five Predictions for the AI Capex Boom
These are our analytical judgments based on the data above, not reported facts, and they could prove wrong.
- Capex guidance rises again in early 2027, but the growth rate falls. Alphabet already said spending will keep rising next year. Expect dollar increases with slower percentage growth, which is the pattern Snider’s remark points toward.
- Bond and equity issuance keeps climbing through the winter. With four of five spenders cash-flow negative in the latest period and Microsoft already issuing a $25 billion bond, more offerings are the default path.
- Credit protection costs stay elevated for at least two more earnings seasons. Hedging demand rarely fades while free cash flow remains negative.
- Weaker AI labs consolidate or seek strategic backers. Commitments such as Anthropic’s $265 billion in minimum purchases only work if revenue scales, and the market is now scrutinizing that.
- Depreciation becomes the next headline. As 2025 and 2026 servers enter the income statement, operating margins at the clouds will face pressure, shifting the debate from capex to profit.
What Investors and Builders Should Watch Next
Several markers will show which constraint binds first. Start with the next round of cloud backlog figures, since a shrinking backlog would undercut the whole demand case. Next, track free cash flow at Alphabet and Meta, where a third straight negative quarter would harden the debt narrative. Then follow bond spreads and swap pricing for the big issuers, because those move faster than earnings.
Watch the Anthropic listing as well. A public filing from a major model lab exposes its commitments and revenue to outside scrutiny, which gives the market a rare direct look at the demand side. Finally, keep an eye on any capex cuts or delays, however small. Microsoft’s $45.2 billion quarter against a $50 billion plan was a delay for supply reasons. A future shortfall caused by weaker demand would be a very different signal.
For developers and companies buying AI services, the practical advice is modest. Avoid locking into long contracts priced at peak compute rates without escape clauses, keep workloads portable across providers, and treat price cuts as a likely feature of the next 12 months rather than a surprise.
Frequently Asked Questions
How much will Big Tech spend on AI in 2026?
Fortune reported combined 2026 capex of up to $828 billion across Amazon, Alphabet, Microsoft, Meta and SpaceX. TechCrunch put the four classic hyperscalers above $740 billion. Figures differ by outlet because companies guide in ranges.
Is the AI capex boom ending?
No outlet cited here reports an end. Alphabet expects spending to rise again in 2027, and Finimize cites analyst expectations of about $1.2 trillion next year. The debate is about the pace of growth and how it is financed.
Which Big Tech company has the weakest cash flow?
Amazon, by size. TechCrunch reported negative free cash flow of $42.4 billion over the past 12 months, against positive $14.8 billion a year earlier. Alphabet and Meta were also negative in Q3, while Microsoft stayed positive.
Why did Alphabet stock fall 12%?
TechCrunch tied the drop to a second straight quarter of negative free cash flow, rising capex plans and a 12% growth rate in Search advertising, the slowest in five quarters. The decline erased roughly $500 billion in market value.
How is the buildout being financed?
Increasingly with debt and equity. TechCrunch reported $326 billion in debt raised by five major firms in 2026, triple last year’s total. Fortune reported a $25 billion Microsoft bond offering and a Meta share issuance of up to $95 billion.
What do credit default swaps say about AI risk?
Finimize said swap prices rose heading into this week. Other reports described record trading volume in swaps on tech issuers. Higher prices mean investors are paying more to insure against losses, which signals rising caution rather than expected default.
What should AI developers and buyers do now?
Keep workloads portable, avoid rigid long-term compute commitments, and watch cloud pricing. A cooling buildout could lower prices, while supply bottlenecks could keep them firm. Neither outcome is certain.
