AI infrastructure spending is rapidly becoming one of the defining economic forces of the decade.

On a recent earnings call, Nvidia CEO Jensen Huang estimated that between $3 trillion and $4 trillion could be spent on AI infrastructure by the end of the decade. Much of that capital, he suggested, will come directly from AI companies themselves as they race to secure compute power.

The scale of the buildout is already reshaping power grids, construction timelines, and corporate balance sheets.

The Rise of AI Data Center Hyperscalers Capex

What was once a routine line item — capital expenditures — has become a headline figure. AI data center hyperscalers capex plans for 2026 alone approach $700 billion.

Amazon is projecting $200 billion in capital expenditures in 2026, up from $131 billion in 2025. Google estimates between $175 billion and $185 billion, nearly doubling its prior year. Meta forecasts between $115 billion and $135 billion, compared to $71 billion previously, although some projects remain off its books.

These investments are being funneled into massive data center campuses, including Meta’s 2,250-acre Hyperion site in Louisiana, expected to deliver 5 gigawatts of compute power and supported by local nuclear energy. Another Meta site in Ohio, named Prometheus, is slated to come online in 2026 powered by natural gas.

Meanwhile, xAI constructed a hybrid data center and power-generation facility in Memphis, Tennessee, which has drawn scrutiny due to emissions linked to natural gas turbines.

The sheer magnitude of AI infrastructure spending is beginning to test investor confidence. Executives argue the outlays are essential for long-term competitiveness, while some analysts question whether returns will justify the scale.

Oracle, Nvidia and the Expanding Compute Economy

Oracle has emerged as a central player. On June 30, 2025, it disclosed a $30 billion cloud services agreement with an unnamed partner later revealed to be OpenAI. Months later, Oracle announced a five-year, $300 billion compute deal set to begin in 2027 — a figure that stunned markets and briefly propelled founder Larry Ellison to the top of global wealth rankings.

Nvidia, flush with GPU demand, has pursued unconventional strategies. In September 2025, it acquired a 4% stake in Intel for $5 billion. It also structured a $100 billion investment in OpenAI, paid largely in GPUs tied to data center expansion. Similar arrangements have followed with xAI and AMD, creating a tightly interwoven ecosystem of hardware, equity, and long-term compute commitments.

The Stargate AI Project OpenAI SoftBank Oracle

Perhaps the most ambitious initiative is the Stargate AI project OpenAI SoftBank Oracle joint venture, announced shortly after President Trump’s second inauguration.

The plan calls for $500 billion in AI infrastructure investment across the United States. SoftBank is expected to provide financing, Oracle to lead buildout efforts, and OpenAI to shape technical deployment. Construction is already underway on eight data centers in Abilene, Texas, with completion targeted by the end of 2026.

While early enthusiasm was intense — with officials calling it the largest AI infrastructure project in history — Bloomberg later reported that partners were struggling to reach full consensus. Despite that, physical construction has continued.

Microsoft and the Origin of the Boom

Much of today’s AI infrastructure spending traces back to Microsoft’s 2019 $1 billion investment in OpenAI. That agreement made Microsoft the exclusive cloud provider for OpenAI, with Azure credits increasingly substituting for direct cash as training demands grew.

Over time, Microsoft’s total investment approached $14 billion. Although exclusivity has since loosened — with OpenAI granting Microsoft a right of first refusal rather than sole provider status — the model of pairing AI labs with major cloud platforms has become standard practice.

Anthropic has secured billions from Amazon, Google Cloud has positioned itself as a primary compute partner for emerging AI startups, and OpenAI has diversified across multiple providers.

The Capex Crunch

AI data center hyperscalers capex has created a delicate dynamic between Silicon Valley and Wall Street. Tech leaders remain bullish on AI’s transformative potential, but investors are increasingly cautious about debt levels and return timelines.

Power grids are under strain. Construction capacity is stretched thin. Environmental impacts are drawing attention.

Yet the trajectory of AI infrastructure spending shows little sign of slowing. For hyperscalers and AI labs alike, compute capacity is not optional — it is the foundation of future growth.

Whether these trillion-dollar bets ultimately pay off may determine the next chapter of the AI era.