Trending
Written by: Stephanie Aliaga
As AI capabilities continue to advance, leading U.S. tech companies are deploying record levels of capital to expand cloud infrastructure, enhance computing power and integrate AI more deeply into their operations. In the wake of DeepSeek’s breakthrough, which sparked both global enthusiasm and a significant selloff among many AI names, the open-sourcing of a world-class reasoning AI model has seemingly increased – rather than decreased – demand for computing resources1. Major U.S. tech companies have reaffirmed their commitments to AI infrastructure spending and Nvidia, at the center of this investment surge, reported record data center revenue growth and robust demand for its latest chip model.
The beneficiaries of AI capex are expected to evolve as companies shift their spending towards AI-driven automation and industry-specific applications. However, this earnings season has highlighted that the AI infrastructure buildout is still in full swing, with implications beyond the tech sector.
The AI Capex Craze
Four major hyperscalers – Microsoft, Alphabet, Amazon and Meta – are projected to spend a cumulative $318 billion this year, accounting for roughly 50% of the entire market’s capex growth2. Meanwhile, the four-year Stargate Project, a $500 billion initiative led by OpenAI, SoftBank, Oracle and MGX, plans to build AI data centers and energy facilities in the U.S., with an initial $100 billion phase in Texas.
The ripple effects of this investment surge extend beyond the tech sector, influencing various facets of the global economy.
- Supply chain expansion: Infrastructure spending is stimulating demand for semiconductor manufacturing, data center construction and equipment, energy resources and related services.
- Productivity enhancements: As AI tools become more widely integrated into various industries—from healthcare and finance to manufacturing and logistics—the productivity gains should translate into overall economic growth. Recent studies suggest AI could add between 1.5-3% to annual productivity growth over the next decade.
- Job creation and transformation: While direct job creation from AI infrastructure investments may be limited,3 AI advancements could indirectly stimulate job growth in various sectors—and job transformation. AI-enabled workers will need a unique skillset. Per the World Economic Forum’s 2025 Future of Jobs report, the most in-demand skills over the next 5 years will involve a combination of technological literacy with socio-emotional traits like creativity and social influence.
However, the investment arms race also raises concerns about overbuilding and capital misallocation. While the hyperscalers have had the cash flow to justify strong levels of investment, there are signs investors are getting skittish. Microsoft, Alphabet and Amazon underperformed in the week they reported earnings this season, and on average market responses were the worst they’ve been since 3Q22. Robust fundamentals should provide support for valuations, but waning enthusiasm and decelerating earnings growth may limit price momentum for these tech giants this year.
For investors, a nimble approach to AI exposure is increasingly necessary. After a significant run in Mag 7 stocks, diversifying across the AI value chain and mitigating concentration risks will be key to capturing long-term value and growth opportunities in an evolving AI landscape.
The major hyperscalers are doubling down on capex commitments
USD billions; Microsft, Amazon (AWS), Meta and Alphabet
Related: Are Small Caps Set to Outshine the Market?
Source: Bloomberg, J.P. Morgan Asset Management. Data for 2024, 2025 and 2026 reflects consensus estimates. Capex shown is company total, except for Amazon, which reflects an estimate for AWS spend (2004 to 2012 are J.P. Morgan Asset Management estimates and 2012 to current are Bloomberg consensus estimates). CFO reflects cash flow before capital expenditures. Data are as of February 26, 2025.
1 The cost efficiencies achieved by DeepSeek is likely to be more than offset by the compute intensity of agentic AI, which can consume 100x to 1000x as much tokens.
2 Source: Empirical Research Partners, “Hyperscalers, Hyperspenders, Hyperextended?” February 19, 2025.
3 WSJ, “The AI data-center boom is a job-creation bust.” Large-scale data centers typically employ around 100 full-time workers once completed, a fraction of the workers that might service a similar size factory or warehouse.