Big tech plans $650-750B AI capex push
- Microsoft, Alphabet, Meta, and Amazon just used this week’s earnings to lock in an extraordinary 2026 AI buildout worth roughly $650 billion. - The clearest tell is the guidance stack: Microsoft sees $190 billion, Alphabet up to $190 billion, Meta up to $145 billion, Amazon about $200 billion. - The fight is shifting from models to infrastructure — chips, power, networking, and data-center capacity now look like the real bottlenecks.
Data centers are the story here — not chatbots, not demos, not vague “AI strategy” talk. This week, Microsoft, Alphabet, Meta, and Amazon all came out of earnings with one message: they are still spending at a scale that barely made sense a year ago. Add the numbers together and 2026 capital spending lands around $650 billion, with some analysts pushing the total closer to $725 billion or even $750 billion if you use broader definitions. The basic point is the same either way — Big Tech has decided the AI race will be won with concrete, copper, GPUs, and power. (bloomberg.com) ### What actually changed this week? The big shift is that this stopped being a forecast from outsiders and started looking like a company-by-company commitment. Microsoft signaled about $190 billion in 2026 capex. Alphabet lifted its full-year 2026 capex range to as much as $190 billion and said 2027 should(bloomberg.com)t tied to AWS and AI infrastructure. (cnbc.com) ### Why are they spending this much? Because AI demand has moved from “interesting product cycle” to “capacity shortage.” These companies are not mainly buying optional upside. They are trying to keep up with inference demand, model training, enterprise workloads, and cloud customers that want AI capacity now. That means more GPU clusters, more custom chips, faster (cnbc.com)works without power and physical buildouts, so capex balloons fast. (bloomberg.com) ### Why does cloud matter so much? Because cloud growth is the closest thing investors have to proof that the spending is earning its keep. Alphabet’s Google Cloud jumped 63% in the latest quarter and topped $20 billion. AWS grew 28%. Microsoft said its AI business passed a $37 billion annual revenue run rat(bloomberg.com) to fill the machines as soon as they install them. (thenextweb.com) ### So why are investors still nervous? Because revenue can grow fast and still not catch up to spending fast enough. Meta’s stock dropped after it raised capex, which tells you the market is still asking the same question: when does this become clearly profitable rather than merely necessary? Investors seem more comfortable when spending is paired with obvious(thenextweb.com)s, open-ended model work, or future platform control. (cnbc.com) ### What does this do to the rest of the economy? It pushes the bottleneck downstream. If hyperscalers spend this hard, demand rises for Nvidia-class accelerators, memory, optical networking, transformers, cooling systems, construction crews, and electricity. Basically, the AI boom starts to look less like a software story and more like an industrial buildout. That is why peo(cnbc.com) the expensive part is the physical substrate. (bloomberg.com) ### Is $650 billion the real number? Close enough to be useful, but not precise enough to fetishize. One common tally for the four biggest spenders lands around $635 billion to $665 billion. Broader estimates climb toward $725 billion or $750 billion depending on which companies and definitions get included. (bloomberg.com)ucture era or the peak of a spending frenzy. (finance.yahoo.com) ### What is the bottom line? AI is no longer constrained mainly by ideas. It is constrained by who can finance, build, and power enough compute. This week’s earnings made that brutally clear: the biggest tech companies are treating infrastructure as the product. (cnbc.com)
