What's the read on power demand, utilities, and data-center load, and which names look most mispriced?

After nearly two decades of flat power demand in the U.S. [2, 18], an unprecedented surge in electricity consumption is underway, driven primarily by the buildout of AI data centers [2, 5, 19, 22]. Projections for annual demand growth vary significantly, from a baseline of 3-4% in developed countries to 4-6% , with regional hotspots like Georgia expecting **8% to 10% annual increases** from data centers alone . Some analyses suggest a national requirement for 15-20% annual growth to meet the demand . This surge is not limited to the U.S., with Japan also projecting a 20% increase in national power demand driven by data centers and AI . While the U.S. narrative is dominated by data centers, it's worth noting a tension with global projections from the IEA, which forecasts that industrial electrification and transport will be larger drivers of electricity demand growth than data centers through 2030 [8, 30].

This demand shock is colliding with an unprepared energy grid, creating significant strain and economic consequences . Power markets like PJM, the nation's largest, are experiencing unprecedented load growth, leading to grid stress, interconnection backlogs, and record-high capacity auction prices [1, 6, 10, 13]. The Texas grid (ERCOT) faces similar pressures, with interconnection requests at 2.5 times peak demand . This imbalance is translating directly to consumers, with utilities increasing rate hike projections from 5% to 9% and publicly attributing the increases to data center load [14, 16]. Some forecasts warn that the collision of AI power demand with grid constraints could raise consumer electricity prices by **15-40%** . The strain is palpable in market signals, such as PJM's implementation of a price cap, which indicates regulators anticipate continued supply tightness and are acting to mitigate extreme price volatility [1, 10].

In response, U.S. energy policy is pivoting from "energy subtraction" to "energy addition," with a focus on expediting new power supply, which is seen as the most effective lever to suppress rising wholesale prices [1, 7, 9]. This includes halting coal plant closures and restarting shuttered nuclear plants . Natural gas is viewed as the critical short-term bridge fuel, evidenced by over 100 GW of turbines on order . However, this rush is creating severe supply chain bottlenecks for critical generation equipment, with capital costs for new combined-cycle gas plants projected to double by 2030 . The long-term solution is seen as advanced nuclear, with commercial small modular reactors (SMRs) expected by 2028-2029 .

Alongside supply expansion, demand-side management and new data center architectures are emerging as key strategies. Making data centers flexible, controllable grid assets that can curtail load during peak demand is considered crucial for maintaining reliability and reducing price spikes, even if it doesn't significantly lower average annual prices [1, 6, 12]. The industry is also exploring a shift from massive, centralized projects to smaller, distributed "edge AI" data centers to bypass regulatory hurdles and achieve faster speed-to-power . While the sources do not