Amazon Just Had Its Best Cloud Quarter in Years. Why Is the Stock Falling?

Amazon (NASDAQ: AMZN) reported its strongest quarter in a long time on July 30. Cloud revenue grew 37% — the fastest in 18 quarters — and total sales rose 20% to $200.6 billion, with operating income up 43% to $27.5 billion. Record results, in other words. Yet since early August the stock has slid about 8%, and it sits roughly a tenth below its 52-week high of about $287, even after a strong first half of the year.

That disconnect is the whole story right now. Before deciding whether the dip is a buying opportunity, it helps to understand what the market is actually arguing about — because it isn't Amazon's latest quarter. It's the $220 billion Amazon plans to spend next, and the sudden disappearance of its free cash flow.

The number that spooked investors

Amazon's free cash flow — the cash a business generates after paying for the things it needs to run — has gone negative. Over the trailing twelve months, the company burned through $7.6 billion, versus a $18.2 billion surplus a year earlier.

Dig into why, and the explanation is unusually clear. Amazon produced $161.4 billion in cash from operations. But it spent $169 billion on property and equipment in the same period — a $66 billion jump from a year earlier that management attributes almost entirely to AI. In February, Amazon expected to spend about $200 billion on capital projects in 2026; on the latest earnings call, CEO Andy Jassy raised that to roughly $220 billion, citing rising memory-chip costs.

This is not an Amazon problem alone. Microsoft, Google, and Meta are doing the same thing, and a broad "market revolt" against megacap AI spending hit the whole group this summer. But for a beginner investor, the instinct to panic makes sense from the outside: a company spending more on equipment than it generates in cash is the classic sign of trouble. The question is whether this is that kind of situation.

The other half of the story: a $496 billion receipt

This is where Amazon differs from the pure narrative. The demand these data centers are being built for isn't a hope or a pitch — a large share of it is already booked in the form of multi-year contracts. AWS exited the quarter with a contracted backlog of $496 billion, growing at triple-digit rates year over year. Amazon says most of its AI capacity is now committed under terms of five years or longer.

That backlog is the single most important number in the stock right now, because it's the bridge between the $220 billion being spent today and the revenue that will eventually pay for it. It sits behind deals with names that are themselves the very companies building AI: OpenAI, Anthropic, and Meta, plus a string of others. Amazon's in-house AI business and its custom Trainium chip business have each passed a $25 billion annual run rate and are growing at triple-digit percentages. AWS's operating margin, meanwhile, expanded to 39.4% from under 33% a year ago — meaning the growth is coming with better profitability, not worse.

The way Jassy frames the negative cash flow matters too. Data centers require capital about two years before the servers inside them can produce revenue, but they can be monetized for more than 30 years once running. In other words, the cash crunch is front-loaded: spend now, collect later. If the booked demand converts, today's negative free cash flow is a timing artifact rather than a sign the bets failed.

The honest catch in the "record" quarter

Before getting carried away, two things keep this from being an unqualified buy signal.

First, the record profit is flattered by accounting. Amazon reported net income of $62.6 billion, or $5.75 per share — but that included roughly $53.4 billion of pre-tax gains, mostly from the rising value of its stake in AI lab Anthropic. That's a non-cash, mark-to-market gain, not money from actually running the business. Strip it out and earnings per share for the quarter was closer to $1.95. This matters because it capsizes the headline price-to-earnings ratio: Amazon looks cheap at about 20x trailing earnings, but on its ongoing operating earnings the multiple is more like 30x — a growth-stock price, not a fire-sale one.

Second, there's a second-order cost hiding in all that spending. When Amazon capitalizes $220 billion of data centers, the depreciation of that equipment flows into future income statements and drags on reported profit for years — unless cloud revenue grows fast enough to absorb it. Higher memory costs also mean the budget buys less compute than originally planned. Add lingering regulatory overhangs — a preliminary EU designation of AWS as a "gatekeeper," and U.S. antitrust attention on Amazon's $70 billion-plus ad business — and the risk list is not empty.

So is now the time?

Separate the two questions, because they have different answers. Is Amazon a great company right now? Yes — the operating evidence is the strongest it has been in years, and this time the AI demand is contracted, not promised. Is it a great stock entry? That's a narrower call, and the honest answer depends on your time horizon and tolerance for a company that is deliberately running negative free cash flow.

On valuation, the pullback has restored some discipline: Amazon trades at roughly 16x EV/EBITDA, below Microsoft (~19x) and Alphabet (~23x), and its stock-based price-to-sales multiple is far lower than its AI peers' because retail revenue pads the denominator. But as the ex-gain earnings make clear, this is not a historically cheap Amazon of the kind seen in 2022. You are paying a fair growth-stock multiple for confirmed growth — reasonable, but not a steal.

The test that will tell you whether the bet is paying off runs over the next two to four quarters: Does AWS keep compounding in the high-20s to high-30s percentage range? Does the $496 billion backlog keep converting into reported revenue? And does the depreciation wave that starts flowing through — plus any normalization in spending — let free cash flow swing back toward positive, as management has signaled? If those hold, the July selloff will look like a case of the market resetting its valuation faster than the business deteriorated. If the growth stalls and the backlog stops converting, the negative cash flow stops being a timing story and becomes the problem itself.

For a patient investor who believes the backlog converts — and the contract evidence argues it should — the pullback's reward-vs-risk has improved since the spring highs. For anyone hoping this is a cheap contrarian bargain, it isn't one yet. The middle, most honest position: this is a high-quality company mid-investment-cycle, and the stock's fate over the next year hangs on AWS growth and the free-cash-flow inflection — not on the one-time Anthropic gain that made the headline look so good.