Sometimes, it can seem like the AI industry is racing to see who can spend the most money on data centers. Whoever builds the most data centers will have the most compute, the thinking goes, and thus be able to build the best AI products, which will guarantee victory in the years to come. There are limits to this way of thinking — traditionally, businesses eventually succeed by making more money and spending less — but it’s proven remarkably persuasive for large tech companies.
If that is the game, Amazon does seem to be winning.
The company announced in its earnings on Thursday that it projects $200 billion in capital expenditures throughout 2026, across “AI, chips, robotics, and low earth orbit satellites.” That’s up from the $131.8 billion in capex in 2025. It’s tempting to attribute the whole capex budget to AI. But unlike most of its competitors, Amazon has a significant physical plant, some of which is being converted for use by expensive robots, so the non-AI expenses aren’t so easy to wave away.
Google is close behind. In its earnings on Wednesday, the company projected between $175 billion and $185 billion in capital expenditures for 2026, up from $91.4 billion the previous year. It’s significantly more than the company spent on fixed assets last year, and significantly more than most of its competitors are spending.
Meta, which reported last week, projected $115 to $135 billion in capex spending for 2026, while Oracle (once the poster child for AI infrastructure) projects a measly $50 billion. Microsoft doesn’t have an official projection for 2026 yet, but the most recent quarterly figure was $37.5 billion, which pencils out to roughly $150 billion, assuming it keeps up. It’s a notable increase, and one that has led to investor pressure on CEO Satya Nadella — but it still puts the company in third place.
From within the tech world, the logic here is simple. The revolutionary potential of AI is going to turn high-end compute into the scarce resource of the future, and only companies that control the …