The AI boom could cost $31.6 trillion to build data centres: PwC
AI is about to fuel a $31.6 trillion data centre boom, PwC says, as the world races to build more computing power.
AI is becoming a very expensive construction project. The race to build smarter models is driving a parallel race to build the data centres, computing systems and power infrastructure needed to run them.
PwC’s Global Data Centre Outlook estimates that global spending on data centres could reach $31.6 trillion by 2050 in its central scenario. The estimate, based on modelling by Oxford Economics across 46 countries and territories, shows just how large the physical side of the AI economy could become.
If AI adoption grows faster than expected, PwC estimates spending could approach $50 trillion. If slower, the figure could fall to around $22 trillion.
The data centre bill keeps coming back
Building a data centre is only the beginning. These facilities need servers, storage, networking equipment and advanced chips. AI workloads rely heavily on graphics processing units, or GPUs, which handle the huge amounts of computing needed to train and run AI models.
PwC expects annual spending on data centres to rise from around $800 billion in 2026 to $1.1 trillion in 2030 and $1.8 trillion by 2050. A major reason is that the technology inside these buildings does not last forever. Equipment typically needs to be replaced every four to six years.
PwC estimates that ICT equipment will account for 93% of total data centre capital expenditure by 2050, compared with around 70% today. In other words, a data centre is not simply a building that companies pay for once. It comes with a recurring cycle of chips, servers and hardware upgrades.
Power could decide where AI gets built
The US is expected to attract the biggest share of this investment, with PwC projecting around $15.1 trillion through 2050, or 48% of the global total.
Asia-Pacific could account for $8.2 trillion, led by China and India, while Europe is projected to attract $5.6 trillion. The Middle East could see $1.1 trillion and Africa around $255 billion. But having money to invest will not be enough.
PwC identifies access to affordable, reliable and low-carbon electricity as the most important constraint on data centre expansion. Connectivity, security, policy certainty, community acceptance and access to GPUs will also influence where companies build.
That could make energy infrastructure just as important to the AI race as the availability of chips.
India has an opportunity, but it needs capacity
India stands to benefit from the growth of Asia-Pacific’s data centre market. Its expanding internet user base, digital economy and public digital infrastructure are creating growing demand for computing capacity.
More investment could also come from businesses and government-linked sectors looking for domestic or trusted infrastructure to handle sensitive workloads. But the global outlook also shows how vulnerable the build-out could be.
PwC estimates that disruptions to chip trade could reduce cumulative global investment to around $25.5 trillion. The bigger message is that power, chips, connectivity, land, capital and policy will all shape where the world's future computing capacity is built.