The headline numbers from Amazon’s Q3 2026 report are easy to misread. The company posted EPS of $1.89 against a $1.97 consensus estimate, while revenue of $201.94 billion fell just short of the $202.22 billion expected. Both misses were narrow. But shares fell 4.25% in the regular session to $243.79, then traded down about another 6% after hours to roughly $228.90. That reaction was not about a $278 million revenue gap.

It was about the cash.

Amazon posted negative free cash flow for the second consecutive quarter, recording an outflow of $42.4 billion for the trailing twelve-month period. The driver is straightforward: the company continues to ramp spending on infrastructure, largely for AWS and AI data centers. Amazon held its 2026 capital expenditure forecast at $220 billion and said 2027 spending would climb meaningfully higher. There is no recovery date on the calendar.

Here is where traders need to pay attention. The surface-level EPS miss is partly a distortion. Net income fell year over year largely because Q3 2025 included $9.5 billion of pre-tax gains from Amazon’s Anthropic investment that did not repeat, while operating income rose sharply. Strip out that comparison effect, and the underlying business is producing record operating income of $28.1 billion and a 13.9% operating margin, the highest in the company’s history. Those are genuinely strong numbers.

So is AWS. AWS grew 39.3% year over year, accelerating for the sixth straight quarter, its fastest pace in nearly five years. The backlog hit $578 billion, growing at a triple-digit rate year over year. CEO Andy Jassy said the cloud unit can grow north of 40% in both 2027 and 2028 and could become a trillion-dollar revenue business within five to ten years. Jassy said large customers are reserving data center capacity ranging from hundreds of megawatts to as much as one gigawatt in advance.

But the arithmetic of the spending plan deserves scrutiny. Net cash capex for the first three quarters of 2026 totals roughly $172.2 billion. If the full year is approximately $220 billion, Q4 is left with only about $48 billion, more than 30% below Q3’s run rate. Either spending eases sharply into year-end, or the $220 billion figure gets revised upward again. Amazon’s custom silicon strategy, Trainium for AI and Graviton for core compute, is yielding price-performance advantages, with customers saving over $2 billion in Q3 alone, which helps justify the outlay. But justification and cash generation are different things.

Fourth-quarter guidance calls for net sales growth of 12% to 16%, with advertising revenue growth reaccelerating to the high teens after timing-related softness in Q3. A Prime Day calendar shift pulled retail and ad growth into Q2, making top-line momentum look softer than underlying demand. That context matters for how traders frame today’s drop.

The real question is not whether Amazon is a good business. It clearly is. The question is whether the market will keep paying a premium while free cash flow stays deeply negative and 2027 capex has no ceiling. Jassy said investment costs can be recovered within three years once AI servers begin operating. That timeline pushes the payback into 2028 and 2029 for spending happening now. Traders with a 12-month horizon are being asked to trust that math. Today’s 10%-plus combined session drop suggests many are not ready to yet.

The skepticism visible in Amazon’s after-hours drop extends into the credit markets as well. Debt markets are already drawing a hard line between AI infrastructure borrowers with proven revenue and those relying on faith in future demand.