Amazon’s holiday quarter was big. Its AI budget is bigger

Amazon $AMZN’s fourth-quarter earnings had two distinct personalities. The first was the one investors have been paying for lately: a cleaner, sharper Amazon that’s figured out how to make its scale throw off real profit. The second was the one Amazon can’t stop becoming when the next platform shift shows up on the horizon: a company willing to pour concrete, string fiber, and buy chips like it’s trying to build the future faster than anyone can invoice it.

Both versions showed up in the same document. The quarter itself was strong. And then CEO Andy Jassy volunteered the sentence that made everything else feel like preamble: “With such strong demand for our existing offerings and seminal opportunities like AI, chips, robotics, and low earth orbit satellites, we expect to invest about $200 billion in capital expenditures across Amazon in 2026, and anticipate strong long-term return on invested capital.”

The market’s response wasn’t “Nice quarter.” It was “Excuse me, you’re spending HOW much?”

In a week when investors have started treating AI spending plans the way Americans treat rising restaurant menu prices — with widening eyes and a sudden interest in doing the math — Amazon effectively walked up to order the tasting menu for the whole table. Shares fell as much as 10% in after-hours trading. Big Tech’s capex plans have become the market’s favorite form of indigestion, and a bad tape doesn’t exactly make investors more patient about long-dated payoffs.

AWS does the heavy lifting

The underlying print was, by most conventional measures, strong. Amazon beat the revenue number analysts were carrying into the report (about $211.4 billion), and AWS came in above expectations. Earnings per share landed at $1.95, a hair below the $1.97 many previews were modeling — the kind of miss that normally gets filed under “within the noise.” 

Amazon’s framing leaned on momentum: AWS delivered “our fastest growth in 13 quarters,” advertising grew 23%, and custom chips were up triple digits, the company said. 

But this earnings season has been running on a different fuel. The market is no longer grading these companies on whether they can post big numbers. The market is grading these companies on whether they can explain why the big numbers are worth the even bigger checks. The largest Big Tech players are expected to spend more than $500 billion on AI infrastructure in 2026, and investors have started picking favorites based on who pairs spending with clean growth and who pairs it with softer near-term outlooks.

Amazon put itself squarely in the second bucket, not because the quarter was weak, but because its guidance invited a heavy dose of skepticism.

The company’s first-quarter operating income forecast — $16.5 billion to $21.5 billion — came in below the $22.04 billion analysts were expecting; operating income is the bridge between “trust us, we’re investing” and “here’s the profit you can actually underwrite.” The same release that celebrated acceleration also warned that Q1 would absorb higher costs, including about $1 billion of incremental “Amazon Leo” expenses year over year, plus investment in faster commerce and sharper pricing in international stores.

The capex number, though, is what turned a fairly straightforward earnings story into a referendum. Amazon’s $200 billion figure forces investors to rewrite their free-cash-flow assumptions in real time. 

The bill comes due

The company already told everyone this movie’s first act: trailing-12-month operating cash flow rose 20% to $139.5 billion, but trailing-12-month free cash flow fell to $11.2 billion, driven “primarily” by a $50.7 billion year-over-year increase in property-and-equipment purchases tied to AI investment. Capex isn’t a theoretical risk here. It’s already compressing the numbers investors underwrite.

Amazon’s capex has already been running at an eye-watering pace: close to $38 billion in the fourth quarter alone, which makes the 2026 figure feel like quite an acceleration. Heading into the print, capex expectations were clustered well below $200 billion — sometimes closer to $150 billion. A $200 billion plan doesn’t just imply more data centers and more chips. It implies more depreciation, more execution risk, and a longer runway before the cash returns feel obvious. It also raises the question that now hangs over every hyperscaler call: Is this spending being pulled by customer demand or being pushed by competitive fear?

The near-term outlook didn’t help.

Amazon guided first-quarter operating income to $16.5 billion to $21.5 billion, tied to roughly $1 billion of higher year-over-year costs tied to scaling its low-Earth-orbit satellite effort, plus more investment in quick commerce and sharper pricing internationally. Sales outside North America jumped 17% in the quarter, but operating income slipped to $1 billion from $1.3 billion a year earlier. Amazon is still buying growth abroad, even as it asks investors to fun a historic infrastructure build.

Amazon made the demand case as hard as it could in its press release. It highlighted an expanding stable of AI offerings and customer wins; it pointed to momentum in its custom silicon, saying Trainium and Graviton have a combined annual revenue run rate above $10 billion; and it described large-scale infrastructure projects meant to ease capacity constraints.

That’s all well and great. But investors aren’t disputing that Amazon can spend. They’re disputing the timeline for when spending turns into a cleaner profit and cash narrative — and, for now, some investors aren't willing to take that on faith.

There’s a reason AWS remains the emotional center of this story: It generates about 60% of Amazon’s operating profit. When AWS is accelerating, the market is inclined to believe the capex will be absorbed and monetized. When the forward profit guide looks light, the market starts wondering whether the bill is arriving faster than the payoff. Meanwhile, Amazon’s rivals are posting faster percentages — so Amazon can be telling a demand story and still have to act like its playing defense.

Amazon wants to be the company that invests like a national utility and earns like a software platform, all while trimming and reorganizing elsewhere to prove it’s still serious about efficiency and margins. In late January, Amazon confirmed 16,000 corporate job cuts as part of a broader restructuring push. So: cut costs in the org chart, spend aggressively in the data center, and ask investors to focus on long-term returns.

Amazon is telling investors that it’s choosing to spend — on infrastructure, on speed, on pricing, on a satellite business it’s explicitly flagging as a scaling cost — at the exact moment investors are re-litigating what “responsible” looks like in the AI era. The quarter gave Amazon plenty of ammunition — better-than-expected revenue, stronger AWS growth, solid operating income, and a set of businesses that still know how to scale. The guidance and capex number demand a second, more difficult thing: a credible explanation of pacing and payoff.

Now, Amazon has one job: to make the $200 billion feel like an answer to demand instead of a reflex to competition. Investors can tolerate a build year. They just want to know whether they’re funding a backlog or funding an AI arms race.