Goldman Sachs Warns: AI Energy Demand Equivalent to "Building Another Japan"; Turbines, Transformers, and Political Resistance Are the Biggest Bottlenecks

AI-driven data center power demand is expanding at a faster-than-expected pace, and the core issue on the supply side has evolved from "is there enough electricity" to "can it be built in time."

Goldman Sachs' latest research warns that from the beginning of 2024 to the end of 2030, the additional electricity consumption from AI will be equivalent to the total consumption of Japan, the world's fifth largest electricity consumer. Shortages of turbines, transformers, transmission lines, and skilled workers, along with local political resistance, are becoming the true bottlenecks constraining supply deployment.

Goldman Sachs has raised its forecast for U.S. power demand growth from a compound annual growth rate of 3.2% to 3.5%. U.S. data center power demand for 2030 has been significantly revised upward from 83 GW to 108 GW. The forecasted global data center power demand increase from 2025 has also jumped from 117% to 170%. Meanwhile, capital expenditure forecasts for hyperscale cloud service providers have been raised in tandem: from $1.2 trillion to $1.7 trillion for 2027, and from $1.5 trillion to $2.1 trillion for 2029.

Goldman Sachs emphasizes that its baseline scenario does not predict large-scale blackouts nationwide, but regional power tightness, rising electricity costs, and public opposition will profoundly impact actual construction progress. The PJM region is currently at a critical level of tension, and ERCOT is expected to come under pressure around 2028 as loads rise. Public and political support will largely determine whether the predicted power capacity can actually be built.

Demand Continues to be Revised Upward: Efficiency Gains Offset by Larger Scale Applications

Goldman Sachs analyst Brian Singer points out that improvements in chip and model efficiency are real, but they have not resulted in an overall reduction in computing expenditure. Lower computation costs have expanded the economically viable scope of applications, with more token generation, broader intelligent agent workloads, and the ever-expanding budgets of hyperscale cloud service providers jointly offsetting and exceeding efficiency gains.

The vacancy rate in major U.S. data center markets has plummeted from 2%-7% to 1%-2%, and Goldman Sachs expects it to recover to around 3% by 2030. This tightening supply-demand situation further supports the logic for upward demand forecasts.

Singer quantifies this trend with an intuitive benchmark: "From early 2024 to the end of this decade—a period of seven years—the additional power consumption from AI will equal that of Japan, the world’s fifth-largest power-consuming country. This scale is hugely significant."

Supply Path: Natural Gas Leads, Renewables Follow, Nuclear Power in the Rear

Goldman Sachs outlines a phased path for data center electricity supply: in the short term, primarily simple cycle gas units and renewables with storage; a medium-term transition to combined cycle gas units; and, in the long term, a reliance on nuclear power, with some existing nuclear plant restarts as a transitional measure. By 2030, data center electricity supply is expected to be composed of approximately 60% natural gas and 40% renewables (including storage).

Due to the lengthy grid interconnection queue of 2 to 7 years, behind-the-meter (BTM) gas generation is becoming a key "shortcut." Goldman Sachs has raised its BTM gas installed capacity forecast to around 30 GW by 2030 (with actual generation exceeding 20 GW), accounting for about 20% of total data center demand at that time. Goldman Sachs sees BTM as a transitional rather than permanent solution, asserting that large customers will still prefer grid connection in the long term for lower costs and higher reliability.

In terms of regional distribution, the rise of MISO (Midcontinent Independent System Operator) is structurally significant—regulated utilities can provide a one-stop service for generation, transmission, distribution, and regulatory relations, and electricity prices are set by state regulators rather than wholesale markets, helping to manage the political sensitivity of electricity cost increases.

The Four "T"s and Seven "P"s: Bottlenecks Shifting from Megawatts to Supply Chain and Politics

Goldman Sachs analyst Allison Nathan summarizes the current core challenges as the four "T"s: turbines, transformers, transmission, and tradespeople. A major turbine manufacturer has indicated that it expects to have sold half its 2031 capacity by the end of this year, and equipment delivery cycles are profoundly affecting choices of power sources. At the same time, after a decade of stagnant U.S. power load, shortages of electricians and high-voltage welders are becoming more acute, with the structurally long training cycles unlikely to be resolved quickly.

Singer further introduces the seven "P" framework, covering pervasiveness (AI adoption), productivity (of chips and models), price of power, policy, parts, people, and physical environment. Physical environment risks are particularly noteworthy: over half of data centers are located in areas with high physical risks, facing challenges such as high temperatures, humidity, and drought. The water-power tradeoff from cooling systems is most prominent in West Texas—accessing water resources there is more difficult and expensive than adding new power supply.

Political Resistance as the Ultimate Constraint: Over 300 Local Bans Already in Place

Goldman Sachs analyst George Lee characterizes interconnection as "the single most important issue in the U.S. utility sector," and notes that there are already over 300 local and regional moratoria, while the number of state-level bans remains relatively limited. He warns that if the AI and power industries cannot develop a coordinated public communication strategy, political factors will become the ultimate hard constraint.

Community opposition is mainly focused on five concerns: blackout risk, rising electricity rates, water consumption, noise pollution, and waste heat emissions. Goldman Sachs believes that projects will migrate to more regulation-friendly jurisdictions, potentially increasing geographic concentration. Meanwhile, local governments competing for tax bases, construction jobs, and infrastructure investments are providing opportunities for some projects to take root.

On utility balance sheets, regulated utilities within Goldman Sachs’ coverage universe are expected to increase capital expenditure by about 60% in the next five years compared to the previous five, with incremental financing expected to be 30% to 50% equity-based. There remains a cushion of roughly 100 basis points on leverage before triggering rating downgrades.