Big tech plans $725bn AI capex

The story here is data centers. Not chatbots, not splashy demos — the concrete, cables, power gear, networking, and chips underneath them. On April 29, Alphabet, Microsoft, Meta, and Amazon all reported results and either raised or reaffirmed enormous 2026 spending plans, pushing the group’s combined capex ceiling to about $725 billion. That matters because AI demand is no longer a side project inside cloud computing. It is becoming the thing that sets the pace for the whole sector. (cnbc.com) ### What actually changed? The big shift was not that these companies are spending a lot — everyone already knew that. The shift was that the number moved higher again, right after another earnings round. Alphabet raised its 2026 capex range to $180 billion to $190 billion from $175 billion to $185 billion. Meta raised its range to $125 billion to $145 billion from $115 billion (cnbc.com)pending. Amazon stuck with its earlier $200 billion plan. Add the high ends together and you get the headline number. (cnbc.com) ### Why are they spending like this? Because AI demand keeps outrunning supply. Alphabet said plainly that it is compute constrained in the near term, meaning customers want more AI and cloud capacity than Google can currently deliver. That is why this spending is not just about buying more GPUs. It is also about land, buildings, power connections, cooling systems, networking ge(cnbc.com)rastructure into the bottleneck. (cnbc.com) ### Who looked best this quarter? Alphabet did. Google Cloud revenue reached $20.03 billion in the March quarter, up 63% year over year, and management said enterprise AI products became the primary growth driver for cloud for the first time. CNBC’s cross-company comparison showed Google Cloud growing faster than Microsoft Azure and Amazon Web Services in the latest quarter. Tha(cnbc.com)s already pulling through into revenue. (cnbc.com) ### Why weren’t investors equally happy with everyone else? Because “we’re investing for the future” is starting to wear thin. Meta raised spending sharply, citing higher component pricing and extra data-center costs, but its stock fell after the report. Microsoft’s capex number also landed above Wall Street expectations. Amazon showed strong AWS growth, but investors still have(cnbc.com)sh flow. The market is getting pickier — it wants proof, not faith. (cnbc.com) ### Is this just a cloud story? Not anymore. Cloud is where the revenue shows up first, but the spending wave runs through the whole supply chain. Memory makers, server builders, networking vendors, data-center developers, utilities, and anyone who can help stand up capacity faster all get pulled in. The catch is that the physical world moves slower than software. You can order chi(cnbc.com) thin air. (cnbc.com) ### What is the real bottleneck now? Power and components. Chips still matter, but memory pricing and data-center build constraints are becoming just as important. Meta explicitly pointed to higher component pricing. Alphabet said demand is there but capacity is not. That is why this race keeps getting more expensive even for companies with huge balance sheets — the scarce thing is no longer just silicon, but fully usable AI capacity. (cnbc.com) ### Does $725 billion mean all of it is AI? Not cleanly. These companies report broad capital expenditure lines, and those budgets also support core cloud, logistics, and other infrastructure. But the direction is obvious. AI is now the main reason they keep revising the numbers upward, and management teams keep framing the buildout around model training, inference demand, and ente(cnbc.com)he strategic driver is not hard to spot. (cnbc.com) ### So what matters next? Watch two things. First, whether Google’s quarter turns out to be the template — huge spend, but obvious cloud payback. Second, whether supply constraints ease fast enough to keep revenue growth ahead of infrastructure inflation. If they do, this looks like the foundation of the next cloud era. If they do not, the industry risks building the most expensive traffic jam in tech history. (cnbc.com)