Google just spooked the market by hiking its annual spending forecast to as much as $205 billion — a stark jump from last quarter’s $190 billion ceiling. Even the new floor, $195 billion, exceeds the old high-water mark. For investors, the real alarm isn’t the raw number. It’s that a company of Google’s caliber essentially admitted it can’t forecast its own costs. That’s the kind of uncertainty that makes people sell.
The math is unforgiving. Google is spending more cash than it’s generating while simultaneously facing pricing pressure from competitive Chinese AI tools. You don’t need an MBA to see the problem: costs are ballooning while revenue per unit is flat or dropping. This isn’t just a Google problem. Meta, Amazon, and Microsoft report this week, and plenty of analysts expect similarly bloated capex numbers tied to the data center arms race.
The anxiety is rippling outward in telling ways. SpaceX shares have cratered to nearly half their peak value. Oracle, which functions as a public-market proxy for OpenAI’s buildout, has investors side-eyeing its mounting datacenter debt. And Nvidia is now structuring deals worth a combined $750 billion, including guaranteeing $250 billion of OpenAI’s obligations. Billy Leung, tech strategist at Global X Management, called that move “as much a reminder of funding strain in the AI build-out as it is a demand signal.”
I’ve spent three years asking AI companies how they actually plan to make money, and I still don’t have a satisfactory answer. Most smart money I talk to expects massive data center overbuilding and a wave of AI company failures when the correction hits. They’re staying in because they think the survivors will generate enough returns to offset the wreckage. But the jitters are unmistakable now. Whether this week’s other earnings reports calm the waters or confirm the fear, Elon Musk taking SpaceX public feels like a pretty decent top signal. We’ll see how this ages.
