Morgan Stanley hosted its second "AI Energy Summit" in New York on December 4, 2025, addressing critical challenges and opportunities in power supply for AI development in North America. The event, attended by 400 participants—a fourfold increase from last year—highlighted surging interest in AI infrastructure. Key takeaways included severe power shortages for U.S. data center developers, the rise of off-grid solutions, favorable power-and-data-center deal terms, political risks, and labor/equipment shortages.
The summit’s primary finding was an impending power deficit for U.S. data center developers, projected at 10–20% in coming years, with risks escalating from 2026 and peaking in 2027–2028. This stems from AI’s nonlinear compute demand growth: data centers now scale to 1GW (vs. 100–500MW 1.5 years ago), while grid capacity lags. High power/water consumption has also sparked political pushback, compounding supply pressures.
Off-grid solutions emerged as a consensus fix, avoiding grid-connection disputes, rising tariffs, and complex approvals. Leading setups combine gas turbines, reciprocating engines, fuel cells, and battery storage, transitioning from stopgap to permanent solutions. Improved deal terms benefit power/data center developers, though execution risks persist amid supply-demand imbalances.
Morgan Stanley dubbed 2026 the "Year of Execution," where "Time-to-Power" solution stocks (e.g., utilities enabling rapid grid access, Bitcoin miners pivoting to HPC data centers) will hinge on project delivery. Turbine makers like
Compute service providers—