PwC and Oxford Economics released the Global Data Center Outlook 2026-2050 report on September 2, forecasting cumulative global data center capital expenditure of $31.6 trillion under a central scenario, with a potential rise to nearly $50 trillion if AI adoption accelerates beyond expectations [1, 2, 3, 4].
The United States is expected to capture nearly half of this spending, about $15.1 trillion, followed by the Asia-Pacific region at $8.2 trillion, Europe with $5.6 trillion, the Middle East $1.1 trillion, and Africa $255 billion [5, 1, 2, 3, 4]. Annual global data center spending is set to increase from roughly $800 billion in 2026 to $1.1 trillion by 2030 and $1.8 trillion by 2050 [1, 2, 3, 4].
Most investment will go toward information and communications technology equipment inside data centers, such as servers, GPUs, storage, and networking gear, rather than buildings or land. Equipment will be refreshed every 4 to 6 years due to rapid chip upgrades [5, 1, 2, 4]. Power supply is identified as the top factor driving investment location decisions, ahead of latency, security, GPU supply, and policy consensus. The PwC report emphasizes, "Power is the most critical. AI data centers consume massive electricity, and the ability to obtain power at reasonable price, stability, and carbon compliance will directly impact whether investments materialize" [1].
AI data centers' huge power demands are driving strong growth in transformers, power equipment, and advanced cooling systems, especially liquid cooling, which is expected to reach 70% of new AI data center installs by 2030 [6, 7, 8]. Companies such as Taiwan's Delta Electronics, South Korea's HD Hyundai Electric, and China's Jinpan Intelligent Technology and Envicool are benefiting from growing demand for power and thermal management equipment supporting AI data centers [6, 7, 8].
HD Hyundai Electric reported a transformer order backlog of $8.5 billion by mid-2026, a 23% increase from six months earlier. Its production capacity is largely booked for the next three years, with some delivery schedules being discussed through 2030. The company said, "By the end of June, the backlog increased 23% compared to six months ago, reaching $8.5 billion, with capacity heavily booked for the next three years and some orders discussing delivery into 2030" [6, 7, 8]. Taiwan data center CEO Zhang Yongjian noted, "Outside the industry, people talk about GPUs; inside the circle, the absolute question is the delivery time of generators and transformers" [8].
Solid-State Transformers, which improve power efficiency by about 4% and reduce costs, are gaining traction, with market penetration expected to grow by 2030 [6, 7, 8].
Local opposition has blocked or delayed at least 75 data center projects worldwide, representing roughly $130 billion in investments, mainly due to environmental, resource, and social concerns [5, 3]. While super-scale cloud operators aim to deliver new data centers in about six months, delays to grid connection can range from two years or more in emerging markets to over eight years in developed ones [6, 8].
Despite the massive AI-driven investments in data centers, overall productive asset investment in the US—factories and infrastructure—has stagnated, raising challenges for industrial competitiveness compared with countries like China, which invests more heavily [9].
The report and industry data highlight growing infrastructure bottlenecks as AI adoption accelerates. Transformer and cooling system supply chains will remain under strain, with delivery schedules extending to 2030. The sector will closely monitor power availability and grid connection timelines to meet rising data center demand over the coming decades.
