Amazon planned to spend $200 billion on AI infrastructure in 2026. Wall Street responded by erasing 12% of the company's stock price in February alone - its worst month since December 2022.
The disconnect between Amazon's AI conviction and investor patience is becoming a real problem. The company is pouring almost all of that $200 billion into AWS data centers to power its AI business, but the returns aren't materializing fast enough to justify the outlay. Amazon's return on invested capital (a measure of how much profit each dollar of spending generates) dropped from 14.8% to 12.4% in just two quarters. Free cash flow - the actual money left over after spending - is projected to go negative in 2026, hitting minus $524 million. That hasn't happened since 2022.
The Trillion-Dollar Confidence Problem
Amazon isn't alone in this squeeze. Combined AI spending across Big Tech is approaching $700 billion in 2026, and investors are starting to ask the uncomfortable question: what exactly are all these data centers producing in revenue? Amazon led a broader tech selloff in early February that wiped out roughly $1 trillion in market value across major AI-focused companies.
The irony is that Amazon's AI infrastructure business is actually performing better than most. AWS is one of the few players achieving respectable returns on its AI data center investment, and 78 out of 83 Wall Street analysts still rate Amazon a buy. The problem isn't that Amazon's AI business is failing - it's that the spending is so enormous that even decent returns look inadequate.
What This Means for AI Tool Prices
This financial pressure flows downstream. When the companies running the cloud infrastructure that powers most AI tools are spending faster than they're earning, something has to give. Either AI tool prices go up, the pace of new model releases slows, or these companies find ways to make inference (the cost of actually running AI models for each user query) dramatically cheaper. For now, the bet is on option three. But if that efficiency gain doesn't arrive soon, option one becomes increasingly likely.
Amazon isn't regretting AI - it's doubling down. The question is whether it can keep doubling down long enough for the bet to pay off before investors force a course correction.