PwC Projects $31.6 Trillion in Global Data Center Spending by 2050 as AI Buildout Accelerates

Today’s Focus

A new forecast from consulting firm PwC projects that global spending on data centers will reach roughly $31.6 trillion cumulatively by 2050, driven by the buildout of infrastructure to train and run artificial intelligence models, Bloomberg reported.

The figure, first published in a PwC analysis and cited by Bloomberg and Seeking Alpha, dwarfs prior industry estimates. It assumes annual capital expenditure climbing from hundreds of billions of dollars today to more than $1 trillion per year by mid-century.

Goldman Sachs analysts issued a parallel note calling near-term AI capex “shocking” in scale, Yahoo Finance reported, and raised their forecasts for hyperscaler spending through 2027. Nvidia, Microsoft, Amazon, Alphabet, Meta, and Oracle account for the bulk of current outlays.

The numbers are landing in Washington at a delicate moment. PYMNTS reported that a bipartisan group of lawmakers is weighing proposals to give the federal government an equity stake or royalty claim in AI infrastructure that benefits from public subsidies, tax credits, or federal land.

At the same time, local opposition is intensifying. A Vox analysis published this week found that new data center construction now polls about as poorly as a range of unpopular public policies, with organized protests reported in California, Virginia, and Georgia over water use, noise, and electricity prices.

President Donald Trump’s administration has pushed federal permitting reforms and opened public land for AI campuses, and the Department of Energy has fast-tracked interconnection reviews for several gigawatt-scale sites. Congress has not yet acted on the public-stake proposals floated in the Senate Commerce Committee.

The Debate

Supporters argue

Industry groups and administration officials frame the projected spending as a generational opportunity. PwC’s report argues the buildout will support millions of construction and operations jobs and anchor U.S. leadership in AI, according to Bloomberg’s summary.

Goldman Sachs strategists told clients the capex wave is “the defining investment cycle of the decade,” Yahoo Finance reported, and predicted productivity gains large enough to offset the electricity draw. The American Enterprise Institute has argued that restrictive local zoning threatens to push data centers, and the tax base they bring, to friendlier jurisdictions abroad.

Sen. Todd Young (R-IN), quoted by PYMNTS, said federal partnership with hyperscalers is “how we keep the compute here” rather than ceding ground to China. Energy Secretary Chris Wright has said grid upgrades tied to data center loads will ultimately lower costs for residential ratepayers by financing new generation.

Nvidia CEO Jensen Huang, cited in Seeking Alpha’s coverage, said demand for AI compute continues to outrun supply and that the $31.6 trillion figure may prove conservative if agentic AI adoption accelerates.

Critics argue

Opponents say the forecast masks who pays and who benefits. Sen. Elizabeth Warren (D-MA) told PYMNTS that taxpayers should not “hand tax breaks to trillion-dollar companies” without receiving equity, royalties, or binding price commitments in return.

The Vox report cited polling in which majorities in Virginia, Ohio, and Arizona opposed new data center construction near their communities, with concerns focused on rising electric bills, aquifer depletion, and diesel backup emissions. Data Center Watch, an advocacy group tracking local fights, documented more than 140 organized protests in 2025 and 2026.

The Center for American Progress has argued that current tax and utility structures let hyperscalers “socialize the costs and privatize the returns,” pointing to state-level rate cases in which residential customers absorbed transmission upgrades tied to specific corporate campuses.

Rep. Ro Khanna (D-CA) told Reason he would support permitting reform only if paired with mandatory community benefit agreements and caps on ratepayer cost-shifting.

What the experts say

Independent researchers say the headline number is plausible but highly sensitive to assumptions about chip efficiency and AI adoption. Lawrence Berkeley National Laboratory’s 2024 data center energy report projected U.S. data center electricity consumption could reach 6.7% to 12% of national demand by 2028, up from about 4.4% in 2023.

The International Energy Agency estimated in its April 2025 outlook that global data center electricity use will roughly double by 2030, to about 945 terawatt-hours, comparable to Japan’s total consumption. IEA analysts noted that efficiency gains historically offset a large share of computing growth, but the pace of AI training is testing that pattern.

Brookings Institution scholar Mark Muro has written that data center jobs, while high-paying, are relatively few per facility, typically 50 to 200 permanent positions on campuses drawing hundreds of megawatts. A 2025 RAND Corporation report found that state and local tax abatements for data centers averaged roughly $2 million per permanent job created, among the highest ratios of any industry.

By the Numbers

$31.6 trillion: cumulative global data center capital expenditure projected through 2050, according to PwC as reported by Bloomberg.

945 terawatt-hours: projected global data center electricity use by 2030, per the International Energy Agency’s April 2025 outlook.

12%: upper-bound share of U.S. electricity demand that data centers could reach by 2028, according to Lawrence Berkeley National Laboratory.

140+: organized local protests against U.S. data center projects documented in 2025-2026 by Data Center Watch, cited by Vox.

$2 million: average state and local tax abatement per permanent data center job, according to a 2025 RAND Corporation analysis.

50 to 200: typical range of permanent jobs at a hyperscale data center campus, per Brookings Institution research.

4.4%: U.S. data center share of national electricity consumption in 2023, the baseline in the Berkeley Lab report.