美洲电力与能源转型：Bernstein电力预测至2030年

Americas Power & Energy Transition: Baseload, Rising -Bernstein's Power Forecast until 2030 There’s never been a more important time to have a good power demand forecast. Afterdecades of no demand growth, electricity demand is growing in the US. It’s not just artificialintelligence - it’s reshoring, it’s electrification, and normal GDP growth. At current demand,Independent Power Producers (IPPs) are growingandproviding returns to shareholders.The macro debate isn’t really whether power demand is going up or how much is powerdemand increasing anymore….I think everyone agrees it’s going up (more on our forecastbelow). The debate is who captures the margin on supplying it. Sunaina Ocalan.+1 917 344 8503sunaina.ocalan@bernsteinsg.com Anshika Bajpai+1 917 344 8306anshika.bajpai@bernsteinsg.com Raphael Lee+1 917 344 8355raphael.lee@bernsteinsg.com We use an in-house built power supply and demand model to forecast power demand bysector - that we think is really easy to use and sensitize. We use the power supply portion toinform our view on our forecasts for the companies that will serve this demand. The supplymodel also helps provide a sanity check for the guidance from the companies we cover. Wego into detail below. Our proprietary power supply and demand model helps us forecast demand by sub-sector(residential, commercial, industrial, and transport) and deep-dive into key drivers of thosesub-sectors below (for example, data centers for commercial power demand). We forecastsupply by sub-sector and consider the role of intermittency of renewables in the forecast aswell. We expect power demand to rise by ~3% annually (CAGR) to 2030 with increasedconsumption occurring in all three major sectors - residential, commercial, and industrial.Since our initiation three weeks ago, we have gotten feedback that our power demandforecast might be conservative. We think 3% is a LOT when historical power demand hasrisen at only a 0.35% CAGR from 2000 to 2024. The range of other sources we’ve seen(IEA, EIA, Rystad) has power demand growing anywhere from 2 - 5% in the 2025-2030time frame, so our model falls on the more conservative side of power demand. We have updated our model through 1Q26 with actual monthly data from the EIA, keepingour forecasts the same. BERNSTEIN TICKER TABLE INVESTMENT IMPLICATIONS We are Outperform on GEV, FRVO, NEE, VST, CEG and LNGWe are Market-Perform on BE, ENPH, TE, VG, CQPWe are Underperform on FSLR and ORA DETAILS DEMAND Power demand in the U.S. has grown at a 1.2% CAGR since 2019, and 0.3% CAGR since 2010, significantly slower than overallGDP growth, as electricity use per kWh of GDP has declined by -3.7% in that time. However, we are now witnessing a once-in-a-generation increase in power demand. We bucket it into four categories: i) Residential demand from homes ii) Commercialdemand from office spaces, warehouses, and other places of business (including data centers) iii) Industrial demand frommanufacturing and iv) Transportation demand from the operation of public transport systems (which is a much smaller slice thanthe other three). We expect to see unprecedented growth in demand for power from all of these categories. The biggest driver of demand is the addition of data centers in the US. Currently, 45% of global data centers are in the UnitedStates (more than 10X data centers in the US as compared to China!). By 2030, we expect ~50% of global data centers to be inthe US. The proliferation of AI is meaningfully adding to the energy required to support data centers. Data center power demandcurrently makes up 6% of overall demand, growing to 10% of overall demand in 2030. Additionally, the push to “reshore” manufacturing in an environment of increasing global geopolitical fragmentation will addincremental power demand from the industrial sector in the U.S. Even before these more recent trends, the need to lower emissions by increasing electrification in various sectors of theeconomy (EVs, heat pumps, induction stoves) was set to increase the reliance on electricity. POWER FORECASTING While there is probably no debate that power consumption is set to increase in the U.S., the extent of this growth is widelydebated. The main tension here is that demand could materialize before supply can be connected. Power forecasting is drivenby weather, seasonality, economic activity, and large technological changes. There is a short (months - 2 years), medium (3-5years) and long term (6+ years) forecasting of power - the short term is still based on weather and seasonality and this is amature science. The long term forecasting is based on large technology shifts and has been more scenario analysis than“forecasting”. Medium term power demand is driven by economic activity and industrial capacity additions - this used to berelatively predictable and slow moving - but now with AI and data centers, this medium term piece is difficult to predict. Thecurrent projection of increased load from data