Trillions of dollars in AI infrastructure spending are locked into the pipeline for the next quarter-century, but the destination is far from settled.
PwC's Global Data Center Outlook, released September 2, 2026, projects $31.6 trillion in cumulative data center investment across 46 countries and territories through 2050.
Built on modeling commissioned from Oxford Economics, the report is the first long-range capital expenditure forecast of its kind to extend through mid-century.
Power sits atop PwC's list of five forces shaping where capital lands, and PwC's report calls it "the binding constraint" in every region.
Annual data center capital expenditure is projected to climb from about $800 billion this year to $1.8 trillion per year by 2050, PwC reported.
The United States is expected to capture nearly half of all global investment, at $15.1 trillion in cumulative spending, the report indicated.
Asia Pacific follows at $8.2 trillion, led by China and India, while Europe is projected to attract $5.6 trillion over the same period.
The Middle East is set to receive $1.1 trillion, and Africa rounds out the global map at roughly $255 billion, the outlook found.
Clara Cutajar, a Senior Partner and Global Infrastructure Leader at PwC Australia, said the AI buildout is emerging as a first-order test for infrastructure investors.
AI infrastructure is becoming one of the defining capital allocation challenges of the next generation. It cuts across technology, energy, real estate, supply chains, regulation and financing. This changes how infrastructure investors need to think about capital requirements, risk and returns.
What separates this cycle from traditional infrastructure booms is that processors, servers, and networking equipment age out every four to six years, forcing operators to fund continuous hardware refreshes long after the initial construction phase.
Information and communications technology equipment is expected to rise from 70% of total data center spending today to 93% by 2050.
Energy access is the single most influential variable in PwC's model, with four additional factors shaping where capital flows: connectivity, security, access to graphics processing units, and combined policy certainty and community support.
The energy constraint appears in several major analyses published this year, reinforcing the bottleneck's structural nature.
Nvidia just made a move Wall Street wasn't ready for
Microsoft just took sides in AI policy fight
OpenAI just disclosed something genuinely alarming
Getting a new data center facility connected to the power grid can take four to ten years in many regions, far longer than construction itself, Ditlev Engel, CEO of Energy at DNV, wrote in a World Economic Forum opinion piece in May 2026.
Artificial intelligence data centers are typically built within two to three years, a mismatch that routinely stalls projects across every major market.
"There is no AI without energy, specifically electricity," International Energy Agency executive director Fatih Birol said at the AI Action Summit in Paris in February 2025, remarks that have taken on new weight as long-range forecasts converge on the same constraint.
The Goldman Sachs Global Institute's 'Tracking Trillions' report, published May 1, 2026, estimates about $7.6 trillion in cumulative capital spending on AI compute, data centers, and power from 2026 through 2031.
The firm's baseline projects $765 billion in annual AI capital expenditure this year, growing to $1.6 trillion per year by 2031.
Goldman identified four assumptions that most affect how much capital the buildout ultimately requires: the useful life of AI chips, the cost and complexity of next-generation data centers, chip architecture mix, and elongation, which is the widening gap between capital deployment and new capacity coming online.
Elongation is where Goldman's near-term model meets PwC's long-range one. The gap widens primarily due to bottlenecks in power, labor, and equipment, with grid interconnection queues the most persistent drag.
The useful life of AI chips is the most consequential variable for total dollars, but elongation determines whether that capital actually gets deployed on schedule.
PwC modeled two scenarios showing how trade policy shifts and national sovereignty campaigns could change the global distribution of AI spending.
Under tighter chip export controls, cumulative global investment falls to about $25.5 trillion through 2050, roughly $6 trillion below the baseline, the report found.
Annual investment drops to about half the central forecast by 2030 under that scenario, before supply chains gradually adapt and spending recovers.
A digital sovereignty scenario tells a different story, with cumulative spending easing to about $29.5 trillion while capital redirects toward markets with underdeveloped data center capacity.
Governments and regulated industries in those markets would prioritize building local infrastructure, drawing investment away from established hubs.
"The AI buildout is not a rising tide that will naturally lift all boats. Capturing this investment requires active positioning," Cutajar said.
Both PwC's long-range outlook and Goldman's near-term model converge on power delivery as the decisive variable.
Texas (ERCOT), Virginia's Loudoun County corridor, the UAE's nuclear-plus-solar buildout, and Nordic hydro markets are positioned to absorb the capital PwC projects.
Regions with four- to ten-year interconnection queues will cede share regardless of tax incentives.
The capital flows extend beyond hyperscalers to regulated utilities in data-center corridors, transmission equipment makers, and natural-gas turbine suppliers.
The flows also reach independent power producers signing behind-the-meter deals with AI operators, as the energy map becomes the investment map.
Related: PwC chairman challenges views on AI layoffs
This story was originally published by TheStreet on Sep 6, 2026, where it first appeared in the Technology section. Add TheStreet as a Preferred Source by clicking here.