Concerns Rise Over AI Market Concentration and Sustainability Amid High Spending and Power Demand
A significant portion of the S&P 500's returns is increasingly concentrated in AI-related stocks, with 75% of the index's performance attributed to just 41 AI companies. Within this group, the 'magnificent seven'—Nvidia, Microsoft, Amazon, Google, Meta, Apple, and Tesla—alone account for 37% of the index's gains.
These firms are making massive investments in AI, with Microsoft, Amazon, Google, Meta, and Oracle projected to spend around $1 trillion by 2026. OpenAI plans to spend approximately $1.4 trillion over the next three years, while forecasting a profit of about $20 billion in 2025.
The expansion of data centers to support AI demand is monumental. For instance, the Stargate data-center complex in central Texas is expected to rival the size of Central Park by mid-2026, while Meta's Hyperion facility in Louisiana is similarly vast. These developments place increasing strain on the US power grid, although some companies have the capacity to build their own power sources.
There are also concerns about depreciation risks associated with AI hardware. If AI chips lose their competitive edge every three years, the combined value of five major tech firms could drop by approximately $780 billion. This impact could escalate to around $1.6 trillion if the depreciation cycle shortens to every two years.
Despite the surge in AI adoption, profitability has not yet caught up with the massive capital expenditures. Estimates suggest that to justify these costs, about $2 trillion in profits would be necessary by 2030.
Adoption rates indicate mixed signals. OpenAI has about 800 million weekly active users but only about 5% are paying subscribers. Early 2025 data indicate that 8–12% of firms use AI in producing goods and services, with large firms' usage estimated at 14% in June but recent figures closer to 12%. Additionally, McKinsey reports most firms remain in early stages of AI integration, underscoring that widespread adoption is still evolving.
