Big Tech Will Spend $725 Billion on AI This Year

The numbers behind the AI boom are getting hard to comprehend. Google, Microsoft, Amazon, and Meta together plan to spend about 725 billion dollars on capital projects in 2026, up roughly 77 percent from last year's record. Almost all of it is going into AI. This is the largest corporate building spree in history.

A year ago the figure was around 410 billion, itself a record. Now the four hyperscalers have raised guidance again, with Microsoft alone setting 2026 capex at 190 billion, well above what analysts expected. Roughly three quarters of the total, around 450 billion dollars, goes straight into AI infrastructure, the GPUs, servers, networking, and data centers that train and run the models.

The spending tells you who wins. Nvidia still captures roughly 90 percent of AI accelerator spending, so every capex plan effectively starts with a Nvidia order. The rest flows to data-center builders, power suppliers, networking firms, and the memory makers, where AI demand has created its own shortage. The scale is so large that some analysts now model 7.6 trillion dollars of AI-related capex across 2026 to 2031.

Investors have mostly cheered it, then started asking harder questions. The AI trade has run for two years on the idea that this spending turns into revenue, and the bull case still holds while cloud growth stays strong. But the bigger the number gets, the louder the doubt, since 725 billion dollars a year only pays off if AI demand keeps compounding. Microsoft's own capex jump, partly driven by surging memory prices, shows the costs are climbing too.

This is the bet that defines the market. If AI keeps scaling into real, paid usage, the spending looks visionary and the infrastructure owners win for a decade. If demand plateaus, hundreds of billions in data centers become a very expensive overhang, and the stocks that rode the capex story unwind. There is no middle path at this scale. The hyperscalers have decided the risk of underbuilding is worse than the risk of overbuilding.

So 2026 is the year Big Tech bet the balance sheet on AI, 725 billion dollars of it. Nvidia cashes most of the checks, the build-out is real, and the only question that matters is whether the demand shows up to match it. Watch cloud revenue growth. That is the scoreboard for all of this.