A striking chart of CapEx among the largest U.S. companies reveals the mind-boggling scale of this tech rush:
The Hyperscale Big Four: A League of Their Own
At the top, four names utterly dominate the landscape, spending sums that resemble the annual budgets of medium-sized nations:
Amazon ($215B, +79% Y/Y): Through AWS, the giant is aggressively expanding data centers to handle skyrocketing demand for cloud-based AI.
Alphabet/Google ($160B, +105% Y/Y): Google has more than doubled its spending to scale Gemini, defend search, and supply enterprise tools through Google Cloud.
Microsoft ($136B, +97% Y/Y): Driven by its partnership with OpenAI and Copilot integration, Microsoft has nearly doubled its year-over-year spend.
Meta ($104B, +66% Y/Y): Meta is spending heavily on the high-capacity infrastructure needed to run open-source models (like Llama) and keep its ad engine sharp.
Combined, Amazon, Google, Microsoft, and Meta's CapEx exceeds $610 billion. This is a scale of concentrated spending never seen before in corporate history.
The Rest of the Field: Strategic & Specialized Spenders
Outside the top four, the spending drops off precipitously, highlighting the different strategies at play:
SpaceX ($55B): While not a pure AI player, SpaceX requires massive compute and telemetry for its launch and satellite networks.
Enterprise & Supply Chain: Memory manufacturer Micron ($31B, +94% Y/Y) is building factories for AI-essential High Bandwidth Memory, while Walmart ($29B) is integrating automation and AI into its supply chains.
Tesla ($15B): Investing to build massive supercomputers to train Full Self-Driving (FSD) and robotics.
The "Asset-Light" Giants: Apple ($9.0B, -29% Y/Y) and NVIDIA ($7.3B, +47% Y/Y) keep CapEx low by outsourcing. Apple relies on partners and consumer devices, while NVIDIA designs chips but outsources manufacturing.
The Big Takeaway
The massive disparity between the top four bars and the rest shows that the AI race is not a level playing field. For the hyperscalers, this is a high-stakes gamble: they have decided that the risk of under-investing in AI infrastructure is far more dangerous than the risk of spending too much. Whether these historic bets will pay off remains the ultimate question—but looking at these numbers, the buildout isn't slowing down.






