Hyperscalers Plan Major CapEx Increases for AI
Major cloud providers are significantly increasing their capital expenditures for 2026 to build out AI infrastructure, according to a Bloomberg Daybreak report. Google and Amazon are reportedly planning 40-50% increases in CapEx. Meta is also deploying millions of NVIDIA GPU units, signaling a massive, ongoing investment cycle in AI hardware across the tech industry.
- Combined 2026 capital expenditure guidance for the top five U.S. hyperscalers—Microsoft, Alphabet, Amazon, Meta, and Oracle—is projected to be between $660 billion and $700 billion, a year-over-year increase of over 60%. - Amazon is leading the spending with a $200 billion CapEx plan for 2026, significantly higher than its $132 billion in 2025 and well above Wall Street's prior forecasts of around $147 billion. - Alphabet (Google) plans to nearly double its spending, guiding for $175 billion to $185 billion in 2026 CapEx, up from $91.4 billion in 2025, with the investment focused on data centers and custom AI chips to meet demand for its Gemini models. - Approximately 75% of this aggregate hyperscaler spending, or around $450 billion, is specifically earmarked for AI-related infrastructure, including servers, GPUs, and the construction of specialized data centers. - This investment cycle is driving a surge in demand for NVIDIA's high-end GPUs, such as the H100 and Blackwell chips which cost $30,000-$40,000 per unit; Meta alone plans to have 350,000 H100 GPUs by the end of 2024. - The massive expenditures are causing capital intensity—the ratio of CapEx to revenue—to surge to historically high levels of 45-57% for some hyperscalers. - To fund this build-out, tech companies are increasingly turning to debt markets; big tech AI firms drove over $200 billion in investment-grade bond issuance in 2025, with Meta pricing a single $30 billion financing deal. - This spending creates a cascading effect down the supply chain, boosting sectors involved in data center construction, power infrastructure, and advanced cooling systems needed to manage the heat from high-density AI hardware.
