Transactions in the AI space continue to move rapidly. OpenAI and Amazon on Monday announced a seven-year, US$38 billion deal for the former to access advanced chips on the latter’s cloud infrastructure platform, while Microsoft announced it would buy US$9.7 billion worth of computing power from Australian data center operator IREN.
These deals are the latest in a series of big AI deals announced in recent months that have fueled the ongoing stock rally. We believe these investments, along with increased capital spending announced by major technology companies over the past week, highlight the growing need for computing power driven by increasingly complex AI applications.
We currently forecast global AI capital spending to reach USD 423 billion this year (previously estimated at USD 375 billion) and USD 571 billion (USD 500 billion) by 2026. By 2030, overall spending is expected to reach USD 1.3 trillion, implying a compound annual growth rate (CAGR) of 25% over the next five years. There are several reasons why these numbers are realistic rather than overly bullish.
Computing demand has exceeded expectations and monetization is accelerating. Third-quarter earnings and recent company commentary confirm that demand for AI and compute resources remains strong. For example, Google’s Gemini reported that its consumption of AI tokens (small units of data used by large language models to process and generate output) has increased 130 times over the past 18 months, while Meta said its computing needs continue to “meaningfully expand” and exceed expectations. The anticipated growth of agent AI (AI systems capable of autonomous decision-making and actions) and physical AI, such as robots and autonomous vehicles, should further drive demand for AI computing. Meanwhile, accelerating cloud revenue growth across major platforms strengthens our confidence in AI’s significant monetization potential, even relative to the size of capital investment plans.
Despite capital spending, big tech companies’ profit margins and balance sheets remain strong. Due to the large upfront investments required for AI computing and infrastructure, the capital expenditure intensity (capital spending as a percentage of revenue) of the four largest US tech companies has nearly doubled to 20.8% over the past five years and is expected to reach 27% by 2030. But the company’s margins are expected to remain relatively resilient as growth in other operating costs, which account for nearly 90% of Big Tech companies’ expenses, slows. These companies also maintain strong cash positions and robust balance sheets. In our view, the risks of underinvestment are likely to outweigh the risks of overinvestment.
Spending on AI remains modest compared to global GDP. The International Monetary Fund estimates that AI capital investment is expected to reach USD 1.3 trillion by 2030, accounting for approximately 1% of global GDP. This is lower than the historic infrastructure investment boom of the past 150 years (ranging from 1.5% to 4.5% of global GDP), including railways, automotive infrastructure, computers, and telecommunications. Similar to these past investments, AI is already driving productivity gains. An Adecco global study found an average savings of 1 hour per day, while Forbes reported 52 minutes. Our calculations show that future productivity gains will be sufficient to justify continued AI investment and associated depreciation.
As such, we maintain the belief that AI stocks should drive the stock market, and believe that under-allocated investors should add exposure to this theme through a diversified approach.
For more information, see Intelligence Weekly #88: Learn more about new AI capital expenditure forecasting.
