The scale of the AI infrastructure buildout is becoming easier to describe in trillions than billions. PwC’s inaugural Global Data Centre Outlook 2026–50 projects $31.6 trillion in cumulative global data center capital expenditure through 2050 under its central scenario, with annual spending rising from roughly $800 billion in 2026 to $1.1 trillion in 2030 and $1.8 trillion by 2050.

There is also an enormous range around that central case. PwC, working with Oxford Economics, puts plausible cumulative investment at roughly $22 trillion to nearly $50 trillion, depending primarily on how quickly AI adoption progresses.

But the most important finding may not be the $31.6 trillion headline. PwC argues that the economics of AI infrastructure are creating a fundamentally different capital cycle from previous infrastructure booms. Data centers are long-lived assets, but the increasingly expensive computing equipment inside them is not.

Servers, GPUs, networking systems and other information and communications technology equipment are expected to require replacement on roughly four- to six-year cycles. PwC calculates that every $1 of construction spending can effectively commit the market to approximately $12 of subsequent ICT investment. ICT equipment accounts for about 70% of total data center CapEx in 2026 under its model, rising to 93% by 2050.

That creates something closer to a continuously renewing technology platform than a conventional construction cycle.

Over a 20-year data center asset life, PwC estimates that a facility could undergo three to five rounds of ICT investment. Increasing rack densities can force corresponding power and cooling upgrades, but the largest recurring expense remains the compute hardware itself.

For data center developers and operators, that distinction matters. The economic life of the building increasingly diverges from the technical and financial life of the infrastructure filling it.

AI Fragments the Data Center Demand Model

The report also sees AI broadening the customer base for data center capacity. Cloud infrastructure demand over the previous decade largely revolved around hyperscalers. PwC identifies at least five additional AI-era buyer categories: neocloud providers, model developers, inference platforms, enterprises and governments. Each creates a different combination of credit risk, density requirements, geographic constraints and infrastructure needs.

Hyperscalers continue to provide enormous scale but place a premium on development speed. Neocloud providers and AI model developers bring dense GPU deployments, but their demand can be more exposed to semiconductor access and AI monetization. Enterprises can provide more durable workloads while carrying tighter security, governance and integration requirements. Governments can anchor sovereign infrastructure, although procurement and policy processes can lengthen development cycles.

The workloads themselves are also becoming more geographically differentiated. Traditional cloud capacity generally follows population, economic activity and established digital markets. Inference becomes more location-sensitive as AI moves into enterprise and regulated workloads, where latency, data access, privacy, security and sovereignty can pull computing closer to users.

Large-scale AI training can behave very differently. Because training is less sensitive to end-user proximity, these workloads can move toward markets capable of supplying cheap, reliable electricity, advanced GPUs, large development sites, technical talent and established AI ecosystems.

PwC estimates that roughly 30% of workloads already carry localized requirements, with that share growing rapidly. The implication is a data center market that becomes less geographically uniform even as global demand continues to rise.

Power Moves to the Top of the Site Selection Stack

PwC identifies five forces that it believes will determine how that capital gets distributed: power; latency and connectivity; security and trusted-region hosting; GPU access and ecosystem depth; and policy certainty and community consent. Power ranks first.

The report describes affordable, reliable and increasingly low-carbon electricity at scale as the most difficult requirement for many markets to deliver. More specifically, it points to transmission capacity, substation availability and multiyear transformer lead times as constraints increasingly capable of determining whether projects break ground at all.

On-site or behind-the-meter generation can help individual projects, PwC notes, but does not eliminate the requirement for broader grid expansion where entire markets are trying to add gigawatts of data center load.

Policy and community acceptance then determine whether those underlying advantages can actually be converted into operating capacity.

PwC does not model sustainability as a standalone variable. Instead, factors such as renewable-heavy grids, cooler climates and credible decarbonization strategies flow through the model's assessment of power and siting competitiveness. Grid, land and planning constraints work in the opposite direction.

That framing is significant for a market in which available land alone increasingly says little about whether a proposed data center site can actually be delivered.

The United States Captures Nearly Half

No country benefits more under PwC's central scenario than the United States. The Americas account for $16.5 trillion in cumulative data center CapEx through 2050. Of that, the U.S. alone accounts for approximately $15.1 trillion, or 48% of the global total.

PwC attributes that advantage to the country's position within the advanced-semiconductor ecosystem, its hyperscalers and AI model developers, availability of capital and technical talent, and an expanding AI-native business base.

The U.S. position also creates substantial exposure to the pace of AI adoption. Under PwC's accelerated scenario, cumulative Americas investment climbs to $27.1 trillion, the largest absolute increase of any region. A weaker AI trajectory produces correspondingly large downside because so much of the U.S. development pipeline is GPU-intensive.

Canada and Chile emerge as notable power-driven alternatives. PwC points to Canada's grid stability and renewable generation base, while Chile benefits from competitively priced renewable power and substantial solar resources.

Asia-Pacific Has More Room to Move

Asia-Pacific accounts for $8.2 trillion in the central case, below the region's share of global economic output.

China and India provide the largest sources of incremental demand, driven by population, expanding digital economies and the potential for broader adoption of AI across businesses and consumers.

PwC sees China as one of the relatively few markets capable of supporting large-scale AI training infrastructure.

Asia-Pacific also has the widest scenario spread. Cumulative CapEx rises 69% under faster AI adoption and falls 34% under the slower scenario. Japan and Australia prove comparatively resilient because of deeper domestic markets and more diversified workloads.

Europe Runs Into Physical Constraints

PwC's central forecast puts European data center CapEx at $5.6 trillion through 2050, below the region's share of global GDP. The reasons are largely infrastructure-related: power constraints, planning difficulties and fragmented regulation.

Those supply restrictions also limit Europe's upside in PwC's accelerated-AI case. The region receives only a 23% increase, the smallest proportional gain among the major regions, because many European markets are already physically constrained in the central forecast.

There are important exceptions. PwC sees the Nordic countries gaining credibility as alternatives to constrained Western European hubs due to renewable-heavy grids, cooler climates and electricity prices it estimates to be 40% to 50% below those elsewhere in Europe.

Middle East Bets Heavily on GPUs

At $1.1 trillion, the Middle East's cumulative investment is much smaller in absolute dollars, but PwC projects the fastest compound annual growth among the regions it examines.

One advantage is the region's ability to coordinate energy, capital, planning and development through centralized policy structures that can compress construction timelines. There is a corresponding risk.

Much of the Middle Eastern buildout is GPU-intensive and designed not simply to accommodate domestic requirements, but also to capture internationally mobile AI workloads. That makes the region unusually sensitive to disruptions in advanced semiconductor supply.

Africa Is a Different Kind of Growth Story

Africa represents only about $255 billion of cumulative investment under PwC's central case, but its trajectory is notable because it is less dependent on AI assumptions.

PwC characterizes much of the region's investment as foundational digital infrastructure that would be required regardless of whether AI adoption meets today's most aggressive expectations.

South Africa remains the region's established data center anchor, with Kenya, Nigeria and Ghana identified among the promising emerging markets. The report singles out Kenya's electricity system, which it says is roughly 95% renewable, as a potentially important infrastructure advantage.

What Happens if the Chips Stop Flowing?

Perhaps the most revealing part of the Outlook is PwC's attempt to model what happens when geopolitics interferes with the buildout.

Its first scenario assumes U.S.-China export controls escalate significantly, access to advanced GPUs becomes restricted across additional markets and retaliatory controls on critical raw materials create additional semiconductor supply-chain friction.

The immediate effect is severe. Global annual data center CapEx falls to roughly half the central forecast by 2030. Supply chains eventually adjust, and by 2050 annual spending actually runs 8% above the central case as delayed infrastructure catches up.

But the lost years matter. Cumulative investment through 2050 falls from $31.6 trillion to $25.5 trillion — a roughly $6 trillion reduction.

The Middle East experiences the largest proportional regional loss, dropping 29%. Asia-Pacific falls from $8.2 trillion to $6.4 trillion, while the Americas decline from $16.5 trillion to $13.8 trillion.

Markets possessing deeper semiconductor supply chains recover faster. PwC sees the U.S. proving relatively resilient, along with Taiwan, Japan and Singapore in Asia-Pacific and Germany in Europe.

It is an important addition to the infrastructure constraint discussion: GPUs themselves can become a gating resource in much the same way that transformers, substations or generation capacity can.

Sovereign AI Changes the Map Instead

PwC's second geopolitical scenario produces a very different result. Rather than restricting chips, it assumes governments and regulated industries become increasingly unwilling to place critical workloads — including public services, financial systems, healthcare and sovereign AI — on foreign infrastructure.

Global spending barely changes. Cumulative CapEx declines only 6.7%, from $31.6 trillion to $29.5 trillion. But trillions of dollars move between markets as countries begin building infrastructure that might otherwise have been supplied by international data center hubs.

Asia-Pacific gains 7% versus the central scenario, with India, Vietnam, Indonesia, the Philippines and Thailand among the beneficiaries. Africa receives the largest proportional gain, rising approximately 12% to $284 billion.

The Americas lose the most dollars. Regional CapEx falls from $16.5 trillion to $13.7 trillion, while the U.S. alone loses approximately $2.9 trillion as workloads previously served from American infrastructure move closer to domestic users elsewhere.

Europe changes relatively little overall because existing data-sovereignty requirements already keep much of its demand within the region. Internally, however, PwC sees the UK, Türkiye and Poland gaining while established cross-border hubs including Ireland, the Netherlands and Germany surrender some workload share.

From Real Estate Asset to Hybrid Infrastructure Platform

For operators and investors, PwC's analysis ultimately comes down to how a data center is underwritten. The firm argues that the asset increasingly needs to be understood as three overlapping layers: property, utilities and semiconductor exposure, each operating on different economic lives and carrying different risks.

For operators, its recommendations are similarly infrastructure-focused: compete on speed to power, preserve optionality through phased utility connections and modular capacity, and build cooling systems capable of accommodating changes in hardware.

PwC also argues for policies that allow utilities and governments to procure long-lead transmission equipment, substations and transformers earlier — while warning that pre-procurement without credible demand can create stranded grid investment of its own.

That may be the clearest takeaway from the entire $31.6 trillion forecast. Capital itself is not the limiting resource PwC sees. The limiting resources are increasingly megawatts, grid equipment, advanced compute hardware, execution speed and the political permission to build.

And because the technology inside each data center will turn over repeatedly during the useful life of the facility, this infrastructure cycle does not end when today's wave of campuses reaches completion. It begins again with the next generation of compute.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. 

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn, following us on X/Twitter and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.