The Great AI Capital-Spending Reckoning Has Begun

For two years, the largest companies in technology have spent on artificial intelligence with the conviction of people who could not afford to be left behind. Data centres rose, chip orders ballooned, and electricity contracts were signed for power plants not yet built. The spending was treated as self-evidently wise. Now the question that capital always eventually asks has arrived: where is the return?

The sums are staggering enough that even patient investors have begun to fidget. Building the infrastructure for advanced AI consumes cash on a scale usually reserved for railways and national grids, and it depreciates far faster. A data centre full of the latest chips is a wasting asset, obsolete in a few years, and the revenue meant to justify it has so far arrived in a trickle rather than a flood.

The market is no longer asking whether AI is transformative. It is asking who actually pays.

This is not a verdict that the technology has failed. It is the ordinary, overdue moment when a gold rush meets a balance sheet. Plenty of genuinely useful applications exist, and some firms are quietly earning real money from them. But useful is not the same as profitable at the price paid, and the gap between the capital deployed and the income returned has grown wide enough to make even believers nervous.

What follows is a sorting. Companies with a clear path from spending to revenue will keep raising money on easy terms. Those whose business case rests on a future that keeps receding will find capital suddenly expensive and shareholders suddenly curious. The era of spending first and explaining later is closing.

None of this means the build-out stops. It means it gets disciplined — pruned to the projects that can show their work. The reckoning is not the end of the AI story. It is the chapter where the market stops applauding the announcement and starts reading the accounts.