UBS analysts state that infrastructure limitations, specifically power supply, grid capacity, and data center space, now dictate the pace of AI development. The bank highlights a significant timeline mismatch: while data centers can be constructed in 12 to 24 months, expanding power capacity requires 5 to 10 years.
Global data center capacity currently under construction or planned totals 276 GW. However, the near-term outlook is constrained, with 57 GW of capacity required by 2027, while only 38 GW of sites have confirmed power. US interconnection queues have reached 2.3 TW, with a median wait time of 6 years for commercial operation.
To address these challenges, UBS estimates that $511 billion in investment for new US power-generation capacity is needed by 2030. Additional bottlenecks include equipment shortages, such as large transformers with lead times of up to 210 weeks, and restrictive state policies, including a one-year moratorium on new data centers over 20 MW in New York.
On September 23, 2026, UBS initiated coverage on colocation data center stocks, including Core Scientific and Applied Digital, with a Buy rating. The bank projects approximately 80% annual EBITDA growth for these companies between 2027 and 2029, citing their strong power development pipelines and the supply-demand imbalance in compute resources.
