UBS says power and grid limits are driving the next phase of the AI trade
In brief
- UBS report "Signal over Noise #14" frames power and grid capacity as the AI trade's next phase.
- UBS estimates 57 GW of data center capacity needed in 2027, against 38 GW with confirmed power.
- Data centers take 12 to 24 months to build; new power takes 5 to 10 years, UBS said.
- UBS initiated Core Scientific and Applied Digital at Buy on September 23, 2026.
Data centers move faster than the grid
The core problem is timing. UBS said a data center can go up in 12 to 24 months, while adding power capacity can take 5 to 10 years. That's a mismatch measured in years, not quarters.
The bank pointed to regulatory hurdles, interconnection problems, supply-chain bottlenecks and community opposition as the main drags on new power infrastructure. Crypto Briefing's coverage adds that US interconnection queues sit at around 2.3 TW (roughly double the country's installed capacity as of 2024), with a median of about 6 years from request to commercial operation. Large transformers face lead times of up to 210 weeks, per the same coverage.
The gap in UBS's numbers
UBS tracks 276 GW of global data center capacity that's either under construction or planned. Its estimate for 2027 is 57 GW of capacity needed, against 38 GW of sites with confirmed power.
Those are the bank's projections, not settled figures.
UBS also estimates $511 billion will be needed for new US power-generation capacity by 2030 to support a 3% compound annual growth rate in electricity demand. It isn't a purely gloomy read, though. The bank projects approximately 415 GW of power will be adequate to meet AI demand by 2030, provided the initial projects come online as scheduled.
States are reacting. New York has enacted a one-year moratorium on new data centers exceeding 20 MW, according to Crypto Briefing, and other states are requiring developers to pay for grid upgrades themselves, shifting costs from utilities and ratepayers to builders.
Colocation coverage
On September 23, 2026, UBS initiated coverage on several colocation data center stocks with a bullish stance. Its analysts started Core Scientific (CORZ) and Applied Digital (APLD) at Buy, citing constrained compute supply against rising demand and strong power development pipelines at those firms. UBS projects around 80% annual EBITDA growth for the colocation names from 2027 to 2029.
Those are UBS's analyst ratings and forecasts, not recommendations from LeoDex News.
Frequently asked questions
Why does UBS say power is the bottleneck for AI data centers?
UBS said a data center can be built in 12 to 24 months, while adding power capacity can take 5 to 10 years. The bank cited regulatory hurdles, interconnection problems, supply-chain bottlenecks and community opposition as the main drags on new power infrastructure.
Will there be enough power for AI by 2030, according to UBS?
UBS projects approximately 415 GW of power will be adequate to meet AI demand by 2030, provided initial projects come online as scheduled. For 2027, the bank estimates 57 GW of data center capacity will be needed against 38 GW of sites with confirmed power.
