AI data center spending seen at $31.6 trillion through 2050

Global investment in data centers could reach $31.6 trillion through 2050 as artificial intelligence drives demand for computing capacity, according to projections from PwC.

Annual capital spending is expected to rise from about $800 billion in 2026 to $1.1 trillion in 2030 and $1.8 trillion by 2050, the firm said on Wednesday, 2 September. The estimate is the central scenario in PwC’s Global Data Centre Outlook, based on modeling conducted with Oxford Economics across 46 countries and territories.

The outcome is not a committed spending total. PwC estimated a range of roughly $22 trillion to almost $50 trillion, depending largely on the pace of AI adoption. Unlike railways or telecommunications networks, data-center investment does not end after the basic infrastructure has been built. Servers, graphics processors, storage systems and other computing equipment must be replaced every four to six years as technology improves and older hardware becomes inefficient.

Those recurring upgrades are expected to account for most long-term spending. Information and communications technology equipment could represent as much as 93% of data-center capital expenditure by 2050.

The US is projected to attract $15.1 trillion, or about 48% of the baseline investment, reflecting its lead in cloud computing, advanced chips and AI model development.

Asia-Pacific could account for another $8.2 trillion, led by China and India. PwC did not disclose a separate public estimate for India, making it difficult to determine how much of the regional total the country would capture.

India is trying to expand domestic data-center and AI-computing capacity through infrastructure investment, cheaper power arrangements and incentives for locally hosted services. Its ability to win a larger share will depend on grid capacity, land, water, connectivity and access to advanced chips.

Power is likely to become the decisive constraint worldwide. Data centers require large quantities of reliable electricity, and suitable grid connections can take years to secure.

The concentration of advanced-chip production and restrictions on semiconductor exports create another risk. PwC estimated that serious disruptions to chip trade could reduce global data-center investment by nearly 20%.

Data-sovereignty policies could have a different effect. Requirements to store or process information locally may redistribute investment toward more countries rather than materially reduce the worldwide total.

The buildout also carries environmental and financial risks. More data centers will increase demand for electricity, cooling water and transmission infrastructure. Utilities and governments could be left with underused assets if AI demand or hardware requirements fall short of projections.

Companies must also decide whether the economic returns from AI services will justify repeated rounds of equipment replacement.