Record Data Sheet
The largest technology infrastructure buildout in history
PwC estimates that global investment in artificial intelligence infrastructure could reach $31.6 trillion through 2050, covering the data centers and technology required to build the computing capacity behind AI. Annual data center capital expenditure is projected to rise from approximately $800 billion in 2026 to $1.8 trillion in 2050. This is not simply a construction program. PwC expects information and communications technology equipment to represent an increasing share of total investment, meaning a significant part of the capital cycle will require continuous replacement and upgrades rather than being spent once on buildings and power infrastructure.
For CEOs and CFOs, this distinction matters. Physical infrastructure can remain productive for decades, while servers, chips and other computing equipment have much shorter economic lives. PwC estimates that ICT equipment will increase from around 70% of investment today to 93% by 2050, making a substantial proportion of the $31.6 trillion a recurring capital requirement.
Where will the $31.6 trillion go?
PwC’s baseline forecast covers 46 countries and territories and estimates that the United States will attract approximately $15.1 trillion, or 48% of global AI infrastructure investment through 2050. Asia Pacific is expected to attract approximately $8.2 trillion, led by China and India, while Europe and the Middle East are also increasing investment through national AI strategies.
| Region | Estimated cumulative AI infrastructure investment through 2050 |
|---|---|
| United States | $15.1T |
| Asia Pacific | $8.2T |
| Rest of world | ~$8.3T |
| Global | $31.6T |
The geographical distribution will not simply follow demand. PwC identifies electricity, connectivity, security, policy certainty, community acceptance and access to GPUs as factors that will influence where computing capacity is ultimately deployed.
The real bottleneck is electricity
The $31.6 trillion investment program ultimately depends on a resource that cannot be expanded at the same speed as software: electricity.
The International Energy Agency estimates that global data center electricity consumption increased 17% in 2025, while electricity consumption from AI-focused data centers increased by 50%. The IEA projects total data center electricity consumption to rise from 485 TWh in 2025 to around 950 TWh in 2030, equivalent to approximately 3% of global electricity demand. Electricity consumption from AI-focused data centers is expected to triple over the same period.
For companies, this changes the nature of AI investment. A project is no longer determined solely by the cost of GPUs, servers or software. It also depends on electricity availability, grid connections, power prices and the ability to secure reliable capacity.
The IEA also identifies bottlenecks in advanced chips, transformers, gas turbines, grid connections and permitting. High-bandwidth memory, used in advanced AI systems, is another constraint on chip production that the IEA expects to remain relevant at least through the end of 2027.
Will the $31.6 trillion reduce the number of employees?
Not directly.
The infrastructure itself does not eliminate employees. It provides the computing capacity required to deploy AI systems at scale, which can then automate or augment specific tasks.
The OECD estimates that AI could increase labour productivity by approximately 0.5 to 1 percentage point per year across G7 economies over the next decade in its central scenario. It also expects AI to increase investment in complementary assets such as cloud infrastructure, data systems, specialised hardware and software.
The effect on employment, however, will be uneven. The OECD expects AI to shift labour toward tasks and occupations that complement the technology while reducing employment in certain roles that become easier to automate. It also warns that automation risks can extend into knowledge-intensive occupations and affect some entry-level positions.
The IMF reaches a similar conclusion. Its analysis estimates that AI could affect up to 60% of jobs in advanced economies, although exposure does not mean that all of those jobs will disappear. The outcome depends on whether AI complements workers or substitutes for their tasks.
For a CFO, therefore, the relevant comparison is not simply AI investment versus headcount. It is AI investment versus cost per unit of output.
A company can invest heavily in AI and maintain a similar workforce if it produces more with the same employees. Another company may use the same technology to reduce headcount. A third could reduce unit costs and increase employment because greater productivity allows it to expand faster.
Will AI reduce labour costs?
It can, but not automatically.
If AI allows a company to produce the same output with fewer labour hours, labour cost per unit falls. If it is used to increase output while maintaining the workforce, productivity rises without necessarily reducing employment.
The IMF has already identified uneven effects across the labour market. Workers with AI-related skills tend to earn more, while some middle-skilled workers in highly exposed occupations face greater pressure. IMF research also found that employment in AI-vulnerable occupations was 3.6% lower after five years in regions with high demand for AI skills than in regions with lower demand.
This points to a potential transformation in corporate cost structures. The first areas under pressure are likely to be tasks that are repetitive, measurable and readily transferred to software, while demand can increase for employees capable of supervising AI systems, validating outputs, integrating them into business processes and working with customers.
What will happen to the economy?
The short-term and long-term effects can be different.
The IMF considers the AI investment boom a positive demand shock in the short term, because large private investments increase economic activity and can create inflationary pressure, particularly because AI infrastructure is highly energy intensive.
Over the longer term, the IMF expects AI to potentially become a positive supply shock if productivity gains materialise. Its estimates indicate that AI could eventually increase global annual potential growth by approximately 0.1 to 0.8 percentage points.
The OECD similarly estimates that AI could increase annual labour productivity growth by 0.5 to 1 percentage point across G7 economies over the next decade in its central scenario.
For companies, the distinction matters. The first phase creates demand for construction, semiconductors, energy, engineering, equipment and financing. The second should create value through higher productivity.
The question is whether those productivity gains will become large enough to justify the capital invested.
What happens if demand does not justify the investment?
This is one of the major financial risks of the entire buildout.
PwC models a scenario in which tighter export controls disrupt global chip supply chains. Under that scenario, annual investment falls to approximately half the central forecast by 2030 before recovering as supply chains adapt. Cumulative investment through 2050 falls to approximately $25.5 trillion, around $6 trillion below the baseline.
The risk does not come only from geopolitics.
The IMF warns that returns from increasingly debt-financed AI investment could prove weaker than expected, potentially resulting in sharp reversals in equity valuations, wealth destruction and layoffs. It also identifies financial relationships across the AI ecosystem as a potential source of contagion.
For CFOs, this makes the depreciation profile of AI infrastructure particularly important. A data center can remain operational for many years while the computing equipment inside it becomes economically obsolete much sooner. PwC’s projection that ICT equipment will represent 93% of investment by 2050 illustrates how different this capital cycle could be from traditional infrastructure.
Who will capture the return?
Not necessarily the companies building the largest data centers.
Investment is distributed across multiple layers of the economy: semiconductors, servers, networking equipment, data centers, electricity generation, transmission, cooling, software and businesses using AI to increase productivity.
PwC identifies power as one of the decisive factors determining where investment flows, while the IEA identifies constraints across electricity infrastructure, chips, transformers and grid connections.
The economic value of the AI buildout will therefore extend well beyond technology. Energy, construction, industrial equipment, real estate, semiconductors and financial markets are all exposed to this investment cycle.
For companies, the strategic question is no longer simply whether to buy AI software. Management must determine which parts of the value chain it wants to control, which it should outsource and which risks it is willing to leave with suppliers.
The capital allocation problem
The scale of the investment changes the question for management.
AI is no longer simply an IT expense that can be compared with the previous year’s software budget. The infrastructure required to support it increasingly competes for capital with factories, logistics, acquisitions, energy projects and other productive assets.
The IEA notes that data center investment has reached a scale that cannot be financed entirely from corporate balance sheets and that capital markets will be critical to the buildout. The pace of investment will therefore depend partly on expectations for returns from AI deployment and on financing conditions.
For a CFO, the relevant metrics are shifting from cost per user toward return on invested capital, capacity utilisation, revenue generated per unit of compute, energy cost per workload, depreciation, financing costs and productivity per employee.
The technology is new. The discipline of capital allocation is not.
What should CEOs and CFOs watch?
Five variables will determine whether the $31.6 trillion represents a productive investment cycle or an overbuild.
Utilisation. Expensive computing capacity produces little return if it remains underutilised.
Energy. Electricity availability and cost will increasingly determine where AI infrastructure can operate competitively.
Hardware depreciation. Recurring upgrades mean the capital cycle will continue long after the initial data center construction is complete.
Labour productivity. The economic case becomes stronger when AI increases output per employee rather than simply adding another technology expense.
Demand. The ultimate test is whether companies and consumers are willing to pay enough for AI-enabled products and services to generate returns above the cost of the capital deployed.
The $31.6 Trillion Question
The number is extraordinary, but the real story is not the $31.6 trillion itself.
The story is that AI is turning computing capacity into strategic infrastructure that requires capital, electricity, chips, land, financing and skilled labour at unprecedented scale.
PwC’s baseline forecast puts cumulative investment through 2050 at $31.6 trillion, with annual data center CapEx reaching approximately $1.8 trillion by 2050. The IEA expects data center electricity consumption to roughly double by 2030, while the IMF and OECD expect AI to increase productivity while also transforming labour markets and creating new economic and financial risks.
For CEOs and CFOs, the question is no longer whether AI will require extraordinary investment. That process has already begun.
The question is whether the productivity gains, additional revenue and strategic advantages created by that investment will be large enough to generate an adequate return on the capital committed.
That is the $31.6 trillion bet.
Sources
PwC, Global investment in AI infrastructure
PwC, Global investment in AI infrastructure
International Energy Agency, Key Questions on Energy and AI
IEA, Key Questions on Energy and AI
OECD, Foundations for Growth and Competitiveness 2026
OECD, Foundations for Growth and Competitiveness 2026
IMF, AI Deployment and Disruption
IMF, AI Deployment and Disruption