AI Capex Boom: Semiconductor Winners Emerge

Wall Street analysts have dramatically raised their forecasts for AI infrastructure spending by the largest hyperscalers. In January, the consensus expected the top five cloud companies—Alphabet, Amazon, Meta, Microsoft, and Oracle—to spend $546 billion on capex in 2026. That estimate has now jumped to $800 billion, nearly double last year's level. For 2027, forecasts have surged from $630 billion to $1.1 trillion, with Morgan Stanley projecting up to $1.6 trillion if additional players like CoreWeave and Nebius are included.

This marks the third consecutive year that analysts have underestimated AI-related capex. The revisions reflect extraordinary demand for AI infrastructure, as companies race to build out data centers and deploy AI models. JPMorgan's chief U.S. economist called the surge in AI expenditures 'the defining feature of our economic times.'

The upward revisions are a boon for semiconductor suppliers. Nvidia, AMD, Broadcom, and Micron have already delivered massive gains since January 2023, but their stocks trade well below recent highs, creating a potential buying opportunity as spending forecasts climb.

The beneficiaries of this spending surge are clear: semiconductor companies that provide the building blocks for AI infrastructure. Nvidia dominates the AI accelerator market with over 80% revenue share, and its new Vera Rubin platform positions it for continued growth. Micron, the third-largest memory supplier, is critical for high-bandwidth memory used in AI processors. Broadcom designs custom AI chips for major tech companies and controls 70% of the data center switching market. AMD is gaining traction as an alternative to Nvidia, with customers like OpenAI and Microsoft seeking diversification.

However, not all companies are equally positioned. While hyperscalers like Alphabet, Amazon, Meta, Microsoft, and Oracle are increasing spending, they are also developing custom silicon to reduce reliance on Nvidia. This could benefit Broadcom and AMD, but may eventually pressure Nvidia's margins. Meanwhile, smaller AI infrastructure players like CoreWeave and Nebius are emerging as significant spenders, but their impact on the semiconductor supply chain is less direct.

Investors should note that the stocks have pulled back significantly from their highs—AMD down 20%, Broadcom and Nvidia down 30%, and Micron down 40%—despite strong fundamentals. This disconnect suggests the market may be underestimating the duration and magnitude of the AI capex cycle.