Google, Amazon plan $670B capex

Data centers are the story here. Not chatbots, not demos, not vague “AI transformation” talk — actual buildings, servers, networking gear, and power contracts. What changed over the past week is that Big Tech earnings made the scale hard to ignore. Alphabet raised its 2026 capital spending target to $180 billion-$190 billion, and Amazon’s latest results reinforced expectations that it will spend around $200 billion this year to keep AWS and its AI stack fed. (abc.xyz) ### Why are Google and Amazon spending this much? Because demand is no longer theoretical. Google Cloud revenue hit $20 billion in the March quarter, up 63% year over year, and Alphabet said its cloud backlog reached $462 billion. Amazon said AWS grew 28% to $37.6 billion, its fastest growth in 15 quarters, with AWS now running at roughly a $150 billion annualized pace. (abc.xyz) ### What is the money actually buying? Mostly compute capacity. That means AI servers packed with accelerators, high-speed networking, storage, and the physical data-center shell around them. It also means custom silicon. Amazon said its chips business — Graviton, Trainium, and Nitro — has already crossed a $20 billion annual revenue run rate, which tells you this isn’t just renting Nvidia boxes anymore. (ir.aboutamazon.com) ### Is this just Google and Amazon? No — that’s the bigger point. Microsoft just laid out a $190 billion 2026 capex plan, and Meta raised its 2026 capex range to $125 billion-$145 billion. Add Oracle and the rest of the infrastructure buildout, and analyst estimates now cluster around roughly $690 billion to $725 billion of 2026 hyperscaler capex. That’s why this stopped being a company story and became an economy story. (cnbc.com) ### Why does Wall Street care so much? Because this spending is starting to explain both the upside and the risk in mega-cap tech. The upside is obvious — cloud growth is reaccelerating, AI services are monetizing faster, and backlog numbers suggest customers are committing real dollars. The catch is that capex is now so large it can crush free cash flow, (cnbc.com)ed to believe that demand will stay hot long enough to earn back the biggest infrastructure buildout the industry has ever attempted. (abc.xyz) ### What’s the real bottleneck now? Power and physical capacity. Google said it remains compute-constrained in the near term, meaning demand is already outrunning available infrastructure. That’s the useful tell. If the limiting factor were customer interest, these companies would slow spending. Instead they’re accelerating, becaus(abc.xyz)strial inputs behind AI. (alphaspread.com) ### Does the $670 billion number hold up? Broadly, yes — but it depends on who’s included. If you mean just the biggest hyperscalers, current analyst ranges are mostly around $690 billion to $725 billion for 2026. If you narrow the basket or use older estimates, you can get into the high-$600 billions. So the headline is directionally(alphaspread.com)precedented.” (futurumgroup.com) ### So what matters next? Watch whether revenue keeps matching the buildout. If cloud growth holds, backlog converts, and AI services keep scaling, this capex wave will look disciplined. If demand slips, the same spending spree starts to look like overcapacity with a very expensive power bill. Either way, Google and Amazon have made one thing clear — AI is no longer mainly a software race. It’s a construction race. (abc.xyz)