AWS Grew 39% and Amazon Fell 7%: Big Tech's $720 Billion Capex Year Now Trades on Funding, Not Growth

Amazon gave investors the quarter they said they wanted and lost the stock anyway. AWS revenue rose 39% to $46.0 billion — the fastest growth in nearly five years — AWS operating income jumped 55% to $17.8 billion at a 38.7% segment margin, and consolidated operating income of $28.1 billion beat the top of Amazon's own guidance range. Revenue of $201.94 billion landed just under the $202.31 billion consensus and diluted EPS of $1.89 missed the $1.96 estimate, so shares fell about 7% in after-hours trading after closing the Oct. 29 session down 4.25% at $243.79. shifted the market's attention from growth rates to the funding gap behind them.

Add up the 2026 budgets — Amazon at about $220 billion, Alphabet at $195-205 billion, Microsoft at about $170 billion and Meta at $135-145 billion — and the four largest hyperscalers are spending $720-740 billion in a single year. Their combined trailing operating cash flow is roughly $677 billion. For the first time the build costs more than the business generates, and the shortfall has to be filled with claims on someone: bonds, preferred stock, or new shares. That is the variable the market repriced in October.

In the third quarter Amazon generated $42.4 billion of operating cash flow and spent $77.3 billion on property and equipment, against $35.1 billion of spending a year earlier. That left a free cash flow deficit of $34.8 billion in three months and pushed the trailing twelve-month figure to an outflow of $42.4 billion, from an inflow of $14.8 billion a year earlier, with the June quarter already negative at $7.6 billion. Marked the moment the build outran the business — and every dollar of the difference was borrowed.

Amazon's trailing net income of $134.8 billion looks like the cheapest earnings stream in mega-cap technology, at roughly 19 times earnings and a 5.1% earnings yield. It is not. Below the operating line, Amazon booked $69.3 billion of other income over the trailing twelve months against $104.4 billion of operating income, and the company itself attributes the bulk of it to non-operating gains on its investments in Anthropic: $53.4 billion of pre-tax other income in the June quarter and $9.5 billion more in the September quarter. Put nearly half of the reported bottom line on non-cash marks rather than on cloud billings, while the cash that would actually fund the build was negative.

Spending does not stay a cash outflow; it comes back as a cost that compounds. Amazon's depreciation and amortization reached $20.9 billion in the quarter, 25% above the $16.8 billion booked a year earlier, against revenue growth of 12%. Alphabet is further along: its depreciation rose 49% year over year to $8.4 billion while capital spending ran at 6.1 times depreciation. Turned a 39% cloud growth rate into a margin question that shareholders cannot manage from the outside.

The scale of the bill is easier to see in Amazon's own numbers than in any forecast. Trailing depreciation of $79.4 billion absorbed 76 cents of every dollar of the $104.4 billion of operating income Amazon reported. And the trailing $215.2 billion of capital spending is larger than the $184 billion annualized revenue run rate of AWS itself — the company is now spending more in a year than its cloud business runs up in a year. Converts a capital-spending boom into a fixed cost that outlives any pricing cycle.

The most telling line in Amazon's third-quarter cash flow statement is not capex at all — it is what did not happen. Amazon raised $48.1 billion of long-term debt in three months and repurchased no stock. Its 10-Q for the first half of 2026 records no repurchases in either half, leaving $6.1 billion of a 2022 authorization unused, and the treasury stock balance was unchanged at $7.8 billion at Sept. 30. Swapped share buybacks for bond issuance to keep the build on schedule, and that trade is the entire equity-versus-debt argument in miniature.

Every dollar of hyperscaler borrowing flows into a short list of listed suppliers, and they are posting the best quarters in their histories. Pays equipment and memory vendors in cash while asking hyperscaler shareholders to absorb the depreciation that creates it.

The funding arithmetic that reshaped Amazon's cash flow statement is what the credit market has been pricing on the other side of the build. Over the past month CoreWeave and Nebius have each fallen more than 30% and IREN has fallen 40%, after the spread on CoreWeave's five-year credit default swaps reached levels normally read as distress and its bond yields climbed. Drew a line between debt-funded builders with cash flow and those without, and the neoclouds sit on the wrong side of it.

There is a darker symmetry here. Amazon's internal accelerator spending is what makes AWS cheap enough to reaccelerate: revenue commitments for Graviton and Trainium rose more than twentyfold year over year, Bedrock is the fastest-growing multibillion-dollar service in AWS history, and customers saved more than $2 billion in the quarter by running on Amazon's own silicon. AWS's 39% growth and the $77.3 billion quarterly cash outflow are two sides of the same decision. That is what makes this cycle different from a normal capital-spending cycle, and what makes it dangerous: the buyer's growth and the buyer's funding need scale together.

The longer question is not whether demand holds. It is what depreciation does to the margin while the market is being asked to keep funding it. Amazon's capital spending is running at 2.7 times depreciation on a trailing basis and 3.7 times in the third quarter alone; Alphabet is at 6.1 times. Turned the capital spending decisions of 2024-2026 into a margin story that shareholders cannot manage, and the only thing that can reverse it is a slowdown in spending, which is exactly what the equity market will not reward in the near term.

For two years the market paid a premium for artificial intelligence capital spending because each dollar of it was a bet on revenue that had not yet arrived. In October 2026 it repriced that bet — not because the revenue was missing, since Amazon's cloud grew 39% and Alphabet's grew 103%, but because the bill for that revenue stopped arriving as a cash flow surplus and started arriving as a bond issuance schedule. Turned cloud growth into a bigger loan rather than a bigger multiple. The winners over the next year should be whoever gets paid in cash on the way in — memory, power and thermal vendors, plus the one hyperscaler disciplined enough to fund its build out of operating cash flow — and the losers whoever finances capacity with claims that rank ahead of shareholders.

The strongest case against this reading is that Amazon's $578 billion backlog, a 38.7% cloud operating margin and the cost advantage of its own silicon could let it convert the build into durable profit faster than the depreciation wave arrives, making today's discount the opportunity. But the bull case is a timing argument and the bear case is a claim-priority argument, and only one of them can be right about whose dollar funds the next server. What is not in dispute is the mechanism: the equity has stopped being the marginal funder of the AI build, and that changes which parts of the chain deserve a premium.