Amazon can afford to build aggressively. It still has to earn a return on the build. It said it expected to spend about $200 billion in capital expenditures in 2026. That is spending across Amazon, including AI, chips, robotics, and satellites. Calling the entire amount an “AI bet” makes the headline cleaner than the disclosure.
The demand signal is real. AWS sales grew 24% year over year to $35.6 billion in the fourth quarter of 2025, according to the same release. Amazon’s trailing free cash flow fell to $11.2 billion from $38.2 billion a year earlier as investment rose. Those two facts can coexist: customers bought more cloud capacity, and the company spent cash faster than it converted that demand into returns.
I think the useful question is whether new capacity earns enough over its life to justify the build. Quarterly AWS growth alone cannot answer it. Amazon has to power and fill the data centers, keep the chips competitive, and earn a return after depreciation. Management expects a strong long-term return, but that is a forecast.
Analysts projected negative 2026 free cash flow, but that was not a reported result. A recovery in the share price was even less certain. I would watch AWS revenue against capex and operating income over several periods. A 9% share move is a reaction; return on invested capital is the verdict.