Microsoft leads $725B capex boom

AI infrastructure is the story here — not a flashy new model, but the warehouses, chips, memory, and power needed to run them. The big change this week is that Microsoft made the spending boom look even bigger. On April 29, the company told investors it expects 2026 capital expenditures to reach $190 billion, and that helped push the combined AI buildout plans of Microsoft, Amazon, Alphabet, and Meta to roughly $725 billion for the year. That is not a rounding error. It is a corporate arms race with utility-scale budgets. (cnbc.com) ### What actually moved the number? Microsoft moved it. The company’s new 2026 capex target came in far above what Wall Street had modeled, and it landed right in the middle of a week when Meta and Alphabet also raised or reaffirmed huge infrastructure plans. Meta lifted its 2026 capex range to $125 billion to $145 billion. Alphabet moved to $180 billion to $190 billion. Amazon kept it(cnbc.com) the total gets to about $725 billion. (cnbc.com) ### Why is Microsoft the center of it? Because Microsoft did two things at once. It posted a strong quarter — revenue of $82.89 billion and Azure growth guidance above expectations — but it also told investors the bill for staying in the AI race is getting steeper. Amy Hood said 2026 capex will be $190 billion, up 61% from 2025, while quarterly capex and finance leases hit $31.9 billio(cnbc.com)ody recalibrates. (cnbc.com) ### Why did costs jump so much? The short answer is memory. Microsoft said $25 billion of its 2026 capex outlook reflects higher component prices. Meta pointed to the same pressure — higher component pricing and extra data-center costs. Basically, this is not just “we want more GPUs.” It is “the whole AI hardware stack got pricier” — especially high-bandwidth memory and adjacent parts (cnbc.com)ways mean more capacity; sometimes it just means paying more for the same rack. (cnbc.com) ### So is this about training or inference? Increasingly, inference. The first phase of the AI boom was about training giant models. The next phase is serving them constantly — inside copilots, search, coding tools, cloud APIs, and enterprise software. Microsoft highlighted more than 20 million paid Microsoft 365 Copilot seats, while management also said demand signals and product usag(cnbc.com)ons of users start querying models all day, the cost of steady, low-latency capacity becomes the real bottleneck. (cnbc.com) ### Are investors comfortable with this? Comfortable is too strong. Investors still reward companies showing real AI demand, but they are getting pickier about returns. Microsoft’s quarter beat, yet guidance on revenue and margins came in a bit light, and gross margin narrowed as depreciation from data-center expansion piled up. Meta’s stock dropped after its report even as it raised s(cnbc.com)ngine that makes this sane.” (cnbc.com) ### What does this mean for the rest of tech? It raises the bar for everyone else. If the hyperscalers keep spending at this pace, they lock in supply, shape chip road maps, and make AI infrastructure even more concentrated in a few giant balance sheets. Smaller cloud players, software companies, and startups then have to rent capacity from the same firms building the moat. That is why(cnbc.com)f the AI economy. (finance.yahoo.com) ### Bottom line? This week’s news was not just that Big Tech is spending more. It was that Microsoft showed how much of the increase is coming from raw hardware inflation, even before demand is fully satisfied. So the boom is real, but it is also more expensive and more concentrated than the headline number first suggests. (cnbc.com)