PwC Puts Potential Global Data-Center Spending Above $30 Trillion by 2050

PwC has estimated that global data-center spending could exceed $30 trillion by 2050. The figure is a long-range projection, not a measure of deployed capacity, completed facilities or spending already incurred.

The publication places that forecast alongside a broader set of AI investment assumptions. It reports that Anthropic plans to spend $518 billion in the coming years, although it does not establish that the full amount has been spent or that it is entirely attributable to data centers. It also cites Bain’s estimate that US companies scaling compute infrastructure would need more than $4.2 trillion in additional revenue over the next five years.

The implied scale of investment is substantial. One estimate cited in the publication puts US AI investment at about $9 trillion in 2025–2032, equivalent to roughly 3.2% of annual US GDP. For the US AI industry to achieve a 10% return on investment by 2032, annual revenue would need to reach about $3.55 trillion, according to the analysis referenced in the article.

The central constraint is the gap between anticipated infrastructure spending and demonstrated economic output. JPMorgan’s calculation, as cited, suggests US productivity would need to grow by 3%–5% annually over the next decade, compared with a 1.75% annual baseline forecast from the Congressional Budget Office. The publication says broad evidence of the required productivity acceleration remains difficult to identify.

For the data-center sector, the outlook therefore combines a potentially long-lived demand signal with material execution and demand-risk exposure. If projected investment materialises, it could support sustained compute-infrastructure demand. But a weaker-than-expected monetisation cycle or delayed productivity gains could lead to slower ordering, project deferrals, asset-value pressure and lower returns after an expansionary build phase—particularly where projects rely materially on debt financing.

The estimates cover AI investment in a broad sense and do not consistently isolate direct data-center CAPEX from other AI-related expenditure. They should therefore be read as scenario-based indicators of the possible scale and economic requirements of the build-out, rather than as confirmed data-center investment or operational capacity.