PwC Puts $31.6 Trillion Price Tag on AI Data Center Build Through 2050, With Power as the Binding Limit

The number that matters most in PwC's new AI infrastructure forecast isn't the headline figure. It's the share of that spending that has nothing to do with concrete.

<cite index="29-2">PwC's Global Data Centre Outlook, released September 2, 2026, projects $31.6 trillion in cumulative global investment in AI infrastructure through 2050.</cite> That's the central scenario. <cite index="37-11">Faster AI adoption could lift cumulative spending to almost $50 trillion, while a slower trajectory would reduce the total to about $22 trillion.</cite> The range alone tells you how many of the underlying assumptions are still unresolved.

But the structural finding is the one operators should pay attention to: this isn't a one-time build. <cite index="34-3,34-4">PwC expects chips and other information and communications technology equipment to increase from 70% of total infrastructure investment today to 93% by 2050, and unlike buildings, this equipment will need to be replaced and upgraded regularly as AI technology develops.</cite> <cite index="35-7">GPUs and servers typically turn over every four to six years, meaning a single data center may require three to five rounds of ICT investment over a 20-year asset life.</cite> That refresh cadence is what makes this cycle structurally different from prior infrastructure buildouts like fiber or highways, where you spend once and maintain.

On the supply side, the report identifies where the real friction is. <cite index="33-9">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.</cite> That finding lines up with what's happening in the deal market. <cite index="31-7">Grid interconnection queues are the most persistent drag, widening the gap between capital deployment and new capacity actually coming online.</cite>

Geopolitics adds another layer of uncertainty. <cite index="29-2">Disrupted trade flows for chips could cut global investment by nearly 20%, while a growing sovereignty push would redistribute rather than reduce global investment.</cite> Those two scenarios pull in opposite directions: export controls compress the total, while sovereignty mandates just shift it around. <cite index="30-5">Power availability, chip trade restrictions, and digital sovereignty policies will determine which markets capture the investment.</cite>

Geographically, the report is fairly blunt about where the advantage sits. <cite index="29-4">The US, which is central to the advanced-chip ecosystem, is expected to capture almost half of this investment, at $15.1 trillion.</cite> <cite index="29-5">Asia Pacific is expected to account for $8.2 trillion in cumulative capex, led by China and India, while sovereign AI strategies are accelerating investment in Europe and the Middle East.</cite>

For context, a separate Goldman Sachs Global Institute report reviewed by TheStreet adds a nearer-term data point: <cite index="33-11,33-12">Goldman'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, with its baseline projecting $765 billion in annual AI capital expenditure this year, growing to $1.6 trillion per year by 2031.</cite> Goldman and PwC converge on the same directional story even if the methodologies differ.

<cite index="31-10">PwC's report, built on modeling commissioned from Oxford Economics, is the first long-range capital expenditure forecast of its kind to extend through mid-century.</cite> That scope matters because it's forcing a different conversation than the quarterly capex disclosures hyperscalers put out. The question isn't whether the build happens. It's who controls the grid access, the chip supply chain, and the permitting pipeline when it does. Those three chokepoints are where the forecast's wide confidence interval lives, and none of them get resolved by announcing a data center campus.

Sources cited:
- PwC Global Data Centre Outlook (press release) (https://www.pwc.com/gx/en/news-room/press-releases/2026/global-investment-in-ai-infrastructure.html)
- PwC Global Data Centre Outlook (full report page) (https://www.pwc.com/gx/en/1/services/consulting/technology/data-centre-outlook.html)
- Yahoo Finance / PwC forecast analysis (https://finance.yahoo.com/technology/ai/articles/pwc-mapped-where-31-6t-170700280.html)
- TheStreet (https://www.thestreet.com/technology/pwc-ai-data-center-capex-forecast-power-constraint)
- Data Center Frontier (https://www.datacenterfrontier.com/machine-learning/article/55403120/pwc-maps-316-trillion-ai-data-center-buildout-through-2050)
- The Energy Mag (https://www.theenergymag.com/news/market-news/ai-data-center-trillion-pwc)
- ESG News (https://esgnews.com/pwc-forecasts-31-6-trillion-in-ai-investment/)
- Northeast Technology Park (https://www.northeasttechnologypark.com/post/ai-infrastructure-investment-forecast-to-reach-31-6-trillion)