Wall Street is finally asking the question the rest of us were told not to ask: what is all this for?
Alphabet fell about 12% after reporting negative free cash flow of $9 billion in a quarter, while holding to a plan to spend $195 to $205 billion on capital expenditures this year. Search advertising, the machine that funds everything, grew 12% and still missed forecasts. The company's own explanation is that its AI products don't yet make as much money as the search results they are replacing. Its fix is to put more ads inside AI answers.
Let that sink in. The technology sold as a civilizational leap, one that would cure disease, end drudgery and justify the layoffs already being announced in its name, has hit its first commercial test, and the plan is to sell ad space next to the chatbot. The grand vision has run into the old business model of harvesting attention and selling it.
I'm not predicting a crash. One bad quarter isn't a bubble, and Alphabet is a hugely profitable company that can absorb its own mistakes. Steve Eisman, who made his name seeing 2008 coming, says he isn't even net short. He is hedging because AI is now so entangled with the whole equity market and economy that betting against it means betting against everything. That is the real story, and it should alarm anyone who doesn't own a data center.
AI capital spending has been holding up a large share of American economic growth, and a handful of tech giants make up roughly a third of the S&P 500. That means your 401(k), your pension fund and your state's retirement system are already in this trade, whether or not anyone asked you. Executives and early shareholders get the upside. If the returns don't arrive, ordinary savers hold the bag. We have seen this movie, and the people in the cheap seats never wrote the script.
The costs also aren't confined to balance sheets. Data centers need land, water and enormous amounts of electricity. Across the country, state and local governments have handed these projects tax breaks, and regulators are fighting over who pays for the new power plants and grid upgrades. If utilities build for demand that never fully shows up, the bill goes to ratepayers, not shareholders. A family in a modest house shouldn't be quietly underwriting a trillion-dollar wager by companies with better balance sheets than most countries.
This is where my own side has been too timid. Too many Democratic governors and mayors have competed to lure data centers with subsidies and called it economic development, while the jobs are few and the energy demand is huge. The "we must beat China" argument has become a universal permission slip. It deserves the same scrutiny we would give any industry that wants public money and public infrastructure and also wants to be left alone.
Nobody voted on any of this. No public body weighed whether $200 billion a year, from one company, was the best use of that capital, labor and power. The market is making that judgment after the fact, abruptly and with a lot of collateral damage.
The response should not be to cheer for a crash. Real AI tools may well prove valuable. The response is to stop letting the industry privatize the gains while socializing the risk. Make companies pay the full cost of the power they use. Attach real conditions, like jobs, wages, clean energy and community benefits, to every tax break. Require honest disclosure of what AI revenue actually is. And be skeptical when the same executives who promise abundance start reaching for the oldest trick in tech, which is to sell your attention to the highest bidder.
The bubble question is open. The question of who pays if it pops should not be.
How it may affect me
If you own an index fund or a 401(k), you likely have heavy exposure to Alphabet and other AI-linked giants, so check your concentration. Your electric bill may also be affected: ask your state utility commission how data-center grid costs are allocated, and whether local tax breaks for new facilities are being approved near you.
