AI infrastructure supercycle worth $7 trillion: The race for data-center infrastructure and global energy demand

The AI infrastructure race is accelerating toward a potential $7 trillion scale, driving sharp increases in electricity and resource demand and placing AI at the center of a global macroeconomic supercycle. Reuters reported that the total value of global data-center deals reached nearly $61 billion in the first 11 months of 2025, exceeding the $60.81 billion recorded for all of 2024, indicating rapid growth in AI infrastructure within a single year.

Deal momentum and regional concentration

For the 2019-2025 period, the United States and Canada accounted for about $160 billion in deals. Asia-Pacific reached nearly $40 billion, while Europe recorded about $24.2 billion, reflecting a North America-led concentration of infrastructure investment.

Capital expenditure outlook for the decade

Drawing on a JPMorgan report, the Financial Times said total capex for AI infrastructure and data centers globally could exceed $5 trillion in the next decade. In an upside scenario, this could rise to $7 trillion, potentially making it one of the largest investment cycles in history.

Data-center capex needs

Analyses cited in the article indicate that capital needs for data centers alone could reach around $700 billion in 2026 and rise to more than $1.4 trillion by 2030.

Reinvestment by cloud providers

The article also notes that large cloud-service providers generate over $700 billion in annual operating cash flow, with around $300 billion reinvested into AI and infrastructure. This highlights the scale of internal reinvestment supporting the buildout.

Energy demand becomes a binding constraint

Energy pressure is a central issue. A report published in November 2025 estimated that energy consumption by global data centers could increase from 448 TWh in 2025 to 980 TWh by 2030—an increase of more than 118% over five years. It also projected that AI server systems would account for about 44% of total energy consumption by 2030.

Power requirements per project rising

AI data-center capacity powering AI has grown to about 500 megawatts per project, eight times higher than the previous level of around 60 MW, making electricity infrastructure a key growth constraint.

Capacity growth targets

Global data-center capacity for AI is expected to grow by roughly 122 gigawatts in 2026-2030. However, absent energy and resource constraints, the figure could reach 144 gigawatts in just three years, suggesting demand could outpace deployment.

Large deals underpin the buildout

According to The Business Standard, the AI infrastructure wave is being driven by mega deals, including computing contracts worth up to $300 billion over five years, data-center projects worth $500 billion, and numerous chip deals worth tens of billions of dollars involving OpenAI, Nvidia, Microsoft, and Amazon.

Profitability thresholds and efficiency concerns

The article highlights questions about financial efficiency across the AI ecosystem. The Financial Times said that to achieve roughly 10% profitability, the market must generate about $650 billion in annual revenue—equivalent to about 0.58% of global GDP—described as a very large revenue threshold compared with prior tech cycles.

Key risk: revenue growth lagging expansion

The risks described are not primarily about demand, but about the market expanding faster than actual revenue, raising the possibility of oversupply similar to earlier infrastructure cycles.

Early commercialization signals

Despite these concerns, the article points to early signs that AI is moving gradually into commercialization. It notes that some companies have already reported annual revenues around $20 billion, suggesting compute demand is beginning to translate into cash flow.

Source: Reuters, FT, The Business Standard