Триллион долларов на кону: почему рынок перестал бояться трат на ИИ

Imagine you walked into a restaurant, ordered dinner, and at the end received a bill not just for the food, but also for building a new kitchen, buying the adjacent plot of land, and training the chef in Paris. That's roughly how investors felt last week when the largest American technology companies revealed their financial plans. Cloud giants — Microsoft, Alphabet, and Amazon — directly stated that their capital expenditures this year would exceed a staggering $650 billion. The market reacted like a capricious teenager: in three days, the combined market capitalization of these companies plummeted by a trillion dollars. It wasn't just a correction, but a true manifesto of distrust in the "AI bubble."

However, by Friday, sentiment shifted sharply, and shares climbed again. Why did panic give way to buying? It all comes down to context that investors initially preferred to ignore. We're in a phase that could be called "infrastructure hangover." Companies aren't spending money on abstract ideas, but on quite concrete hardware from Nvidia and building data centers that will consume electricity on the scale of small countries. Investors feared that return on these investments (ROI) wouldn't come anytime soon, but then they remembered Silicon Valley's old truth: in technology races, it's not the most economical that survives, but the one who managed to stake out the territory first.

If we look at history, we'll see that such cycles repeat constantly. Amazon once spent billions on logistics centers, causing analysts to laugh, and now that infrastructure is its main asset. Today the situation is identical, but the stakes are tens of times higher. Companies aren't just buying graphics cards; they're laying the foundation for the next decade. Without these expenditures, they risk becoming the next Yahoo or Nokia — companies that once dominated but missed the turn toward mobile technologies or cloud computing. The market understood that $650 billion isn't the whim of CEOs, but an insurance premium for the right to stay in the game.

It's also interesting how quickly confidence was restored. This signals that liquidity in the market remains abundant, and faith in AI as the main driver of economic growth remains unshaken. The trillion-dollar drop turned out to be just a temporary fire sale for those who believe in the long-term trend. Of course, skeptics will continue to point out that revenues from AI itself don't yet cover even a tenth of these expenses. But in the Big Tech world, revenue often follows dominance, not the other way around. Right now we're witnessing an attempt to buy this dominance at any cost before competitors do the same.

The connection between hardware spending and stock growth has become almost linear. As long as Nvidia reports record profits and cloud providers confirm demand from their clients, investors will grumble but keep buying. The current rebound shows that the market has made peace with the new reality: cheap AI won't exist. Either you spend hundreds of billions on developing your own models and infrastructure for them, or you use other people's solutions and pay rent to those who weren't afraid to take the risk. In this all-or-nothing game, going all-in seems to be the only strategy that makes sense to the board of directors.

In the end, the trillion dollars that temporarily evaporated from balance sheets came back home. It was an important stress test for the entire industry. It showed that even the scariest numbers in financial reports can't outweigh the fear of missing the next technological revolution. We're entering an era where business efficiency will be measured not only by profits in the current quarter, but also by the number of teraflops a company can throw at training its neural networks.

The bottom line: the market recognized that the AI race is a long game, where enormous spending is not a sign of madness but a necessary condition for survival. Will companies be able to monetize this infrastructure before investors' patience finally runs out?