AI Data Centers to Draw $31.6 Trillion Through 2050, PwC Says

Global spending on data centers and the computing equipment inside them could reach $31.6 trillion through 2050 as AI drives repeated upgrades of costly chips and servers, according to a PwC forecast released Wednesday.

Annual capital expenditure is projected to rise from about $800 billion in 2026 to $1.1 trillion in 2030 and $1.8 trillion in 2050 under the central scenario, PwC said in its Global Data Centre Outlook. Faster AI adoption could lift cumulative spending to almost $50 trillion, while a slower trajectory would reduce the total to about $22 trillion.

The forecast covers both buildings and information and communications technology equipment across 46 countries and territories. PwC commissioned Oxford Economics to develop the projections, which are stated in real US dollars at 2025 exchange rates.

Most of the spending would go toward the hardware housed inside data centers rather than the buildings themselves. Servers, graphics processors, storage and networking equipment typically require replacement every four to six years, creating a recurring investment cycle that distinguishes AI infrastructure from railways, telecommunications networks and other assets with longer replacement schedules.

ICT equipment is expected to increase from 70% of total data center expenditure in 2026 to 93% by 2050. PwC estimates that each dollar committed to construction effectively generates about $12 in subsequent spending on computing equipment over time.

The spending surge is already visible among publicly traded operators. An analysis of company filings found that 12 public Bitcoin miners and three AI data center peers — Applied Digital, CoreWeave (NASDAQ: CRWV) and Nebius (NASDAQ: NBIS) — spent a combined $30.7 billion on capital assets in their latest 2026 reporting periods. That was 42.6% more than the $21.53 billion the same companies spent during all of 2025. CoreWeave and Nebius accounted for almost three-quarters of the 2026 total, spending $14.12 billion and $8.13 billion, respectively. The calculation measured net cash purchases and deposits for hardware, property, equipment and other identified productive assets.

That dependence on regular hardware upgrades raises the risk that operators could be left with underused or obsolete capacity if AI adoption or revenue growth falls short of expectations. A data center may require three to five rounds of technology investment during a 20-year operating life, according to PwC.

The US is projected to attract $15.1 trillion, or about 48% of worldwide spending, reflecting its concentration of advanced-chip suppliers, cloud-computing companies and AI developers. The Americas overall would receive $16.5 trillion.

Asia-Pacific is forecast to attract $8.2 trillion, led by China and India, while Europe would receive $5.6 trillion. PwC estimates investment of $1.1 trillion in the Middle East and $255 billion in Africa.

The projections reinforce the widening connection between technology spending and electricity infrastructure. The International Energy Agency estimates that capital expenditure by five large technology companies exceeded $400 billion in 2025 and could increase by another 75% in 2026.

The agency expects global data-center electricity use to nearly double from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030, equal to about 3% of worldwide power consumption. It has warned that grid connections, transformers, power equipment and advanced-chip supplies are already constraining some projects.

PwC identified electricity availability as the most important factor determining where investment is ultimately deployed. Transmission capacity, substation availability and multiyear lead times for transformers could delay projects even in markets where demand and financing are available.

Connectivity, security, access to advanced processors, regulatory certainty and community acceptance will also influence location decisions. Training large AI models can move to markets offering inexpensive power and access to chips, while inference and regulated workloads are more likely to remain close to users because of latency, privacy and data-sovereignty requirements.

Trade restrictions present another risk. In a PwC scenario involving tighter controls on advanced chips, annual investment falls to roughly half the central projection by 2030 before recovering as supply chains adjust. The disruption would reduce cumulative spending through 2050 by about $6 trillion, to $25.5 trillion.

Asia-Pacific expenditure would fall to $6.4 trillion in that scenario, from $8.2 trillion, while spending in the Americas would decline to $13.8 trillion from $16.5 trillion. The Middle East would experience the largest proportional reduction because several planned projects depend on imported advanced processors and internationally mobile AI workloads.

A shift toward sovereign computing would have a different effect. If governments and regulated industries require more critical workloads to be hosted domestically, total investment would decline only 6.7% to $29.5 trillion, PwC estimated. Spending would migrate from established international hubs toward countries with large domestic markets but relatively limited existing capacity.

India, Vietnam, Indonesia, the Philippines and Thailand would gain under that scenario, while some regional hubs — including Singapore, Japan and Australia — would receive less investment than in the central forecast. The Americas would record the largest absolute decline as workloads currently served from US facilities move closer to users elsewhere.