Amazon bets $200 billion on AI, and the market wants proof it will pay off

Imagine you set out to build a house, then realize you don't need just a cottage but an entire city with its own power plant. That is roughly what is happening at Amazon right now. The company has announced plans to raise its capital expenditure to a staggering $200 billion. That is 54 percent more than last year. And almost all of that money will flow into one "black hole" called artificial intelligence. AWS used to be a convenient cloud for storing your photos and websites. Now it is a giant forge where future GPT-class models and their rivals are hammered out. Demand for compute is growing faster than the servers in data centers can cool down.

Why is Amazon spending this much right now? The answer is both simple and complicated: it has no choice. The cloud market has changed for good. Competition used to be about interface convenience and the range of tools; now the winner is whoever has more Nvidia chips and more powerful transmission lines. AWS has long been the cash cow of Jeff Bezos's empire, delivering the lion's share of profit that covered losses in logistics and retail. Now the cow has to be fed golden hay. Investors, used to steady reports, look at these figures with barely concealed horror. For them, $200 billion is not an investment in the future but a risk that may not pay off for years.

The situation is complicated by the fact that Amazon is not the only player with bottomless pockets. Microsoft and Google are also putting tens of billions into their cloud platforms. This creates an "AI tax". To stay in the game and not lose customers who are moving en masse toward generative models, tech giants have to buy up all the available hardware. It recalls a gold rush, where the sellers of shovels, that is, the chipmakers, earn the most, while the prospectors spend their last money hoping to find a nugget. The stock market's reaction was predictable: Amazon shares dropped, because Wall Street loves profit here and now, not vague promises of dominance in 2030.

There is another problem: the physical world is starting to push back against digital progress. Building a data center at this scale is not just a matter of buying servers. It is a fight for land, for water for cooling and, above all, for electricity. We already see Big Tech starting to invest in small nuclear reactors because ordinary grids cannot withstand AI's appetite. Amazon is effectively turning from a software company into an infrastructure behemoth that does more energy and construction work than coding. This is a fundamental shift in the business model, one that many analysts are still trying to grasp.

What does this mean for us? Most likely, the cost of AI services for consumers and businesses will only rise. Free neural networks will become even more limited, and access to powerful APIs more expensive. Amazon is trying to claim the position of the main landlord of the AI world, but the price of this strategy may prove too high even for it. If in the next couple of years we do not see a real economic boom from AI adoption in the real economy, these $200 billion could become a monument to the unmet expectations of the "large models" era.

The bottom line: Amazon is betting that AI will become the new oil, but for now the company is spending more to extract that oil than it is worth on the market. Who will blink first in this race of budgets?