Combined capital expenditures across Alphabet, Amazon, Meta, Microsoft, and Oracle rose from roughly $162 billion in 2022 to approximately $448 billion by 2025, according to a Visual Capitalist analysis of company filings. The figure represents a 177 percent increase over three years.

The inflection point arrived around the middle of 2023, when spending shifted from steady annual growth to sharp acceleration. Before that point the five companies were expanding data center footprints on a schedule set by cloud services demand. After it, capital allocation moved toward chips, power capacity, and buildings sized for model training and inference workloads.

The spending concentration matters for pricing across the AI stack. When five buyers account for a large share of accelerator purchases and data center construction, hardware lead times and cloud instance pricing move with their procurement calendars rather than with broader enterprise demand.

For enterprise buyers of AI marketing and automation tools, the capital cycle sits several layers below the products they license. Model access pricing, context window limits, and inference latency all trace back to how much compute is available and what it costs to run. Sustained capital expenditure at this level implies continued capacity growth, though it also raises the depreciation load each company must recover through service pricing over the coming years.

Source: Visual Capitalist - https://www.visualcapitalist.com/visualized-big-tech-ai-spending/