Key Takeaways
- A recent U.S. Department of Energy (DOE) report warns of increasing blackout risks as electricity demand from artificial intelligence and data centres collides with the accelerated retirement of reliable coal and natural gas power plants.
- Grid operators, particularly PJM Interconnection, have sharply revised their long-term demand forecasts upwards, citing the unprecedented power requirements of new data centres, which are clustering in specific regions and straining local infrastructure.
- The structural gap between retiring dispatchable generation (coal, gas) and adding intermittent renewables (solar, wind) is forcing a pragmatic reassessment of energy policy, with several utilities already delaying planned coal plant closures to ensure grid stability.
- This dynamic creates distinct investment implications, favouring natural gas as a crucial bridge fuel, nuclear energy as a long-term solution, and companies involved in grid modernisation and energy efficiency.
The United States is facing a structural deficit in its electricity supply, a predicament where surging demand, primarily driven by the computational needs of artificial intelligence, is on a collision course with a shrinking inventory of reliable power generation. A recent report from the U.S. Department of Energy (DOE) has lent significant weight to this concern, projecting a heightened risk of blackouts as the nation’s grid is squeezed from both ends.1 While the retirement of coal-fired power plants continues in line with long-term environmental objectives, the growth in replacement capacity, particularly from dispatchable sources, is failing to keep pace with an aggressive new demand profile shaped by technology and industrial policy.
The New Arithmetic of Power Demand
The scale of new electricity demand is startling and is fundamentally reshaping load forecasts that were stable for decades. The primary catalyst is the proliferation of data centres, which are essential for powering AI models and cloud computing. The Boston Consulting Group projects that data centres alone could consume up to 7.5% of all U.S. electricity by 2030, a substantial increase from the 2.5% they consumed in 2022.2 This is not a gradual, distributed increase; it is concentrated and intense, with single data centre campuses often requiring as much power as a small city.
This is reflected in the revised forecasts from grid operators. PJM Interconnection, which manages the grid across 13 eastern states and is home to the world’s largest concentration of data centres in Northern Virginia, recently doubled its 15-year demand growth forecast.3 This surge is compounded by demand from the reshoring of manufacturing facilities and the broader electrification of transport and heating. The previous energy paradigm, built on predictable, slow-moving demand growth, is no longer fit for purpose.
A Widening Supply-Side Chasm
Simultaneously, the supply side of the grid is undergoing its own radical transformation. For decades, a fleet of coal and nuclear plants provided consistent, dispatchable baseload power. However, a combination of regulatory pressures, environmental goals, and unfavourable economics has accelerated their retirement. According to the U.S. Energy Information Administration (EIA), between 2011 and 2020, approximately 100 gigawatts (GW) of coal-fired capacity was retired. A further 83 GW, or 42% of the remaining fleet, is slated for retirement by 2035.4
While renewable generation from solar and wind has grown significantly, it is intermittent and cannot, by itself, guarantee the 24/7 reliability required by a modern economy. The table below illustrates the growing divergence between projected demand and the availability of firm, dispatchable capacity.
| Metric | 2023 Status | 2030 Projection | Key Driver |
|---|---|---|---|
| Data Centre Power Consumption (% of U.S. Total) | ~3% | 7.5% | AI & Cloud Computing |
| 5-Year Electricity Demand Growth Forecast (National) | 4.7% | N/A (Represents 2023-2028 growth) | Data Centres & Industrial Reshoring |
| Cumulative U.S. Coal Capacity Retirements (since 2011) | >100 GW | ~150 GW | Policy & Economics |
Sources: Boston Consulting Group, Grid Strategies LLC, U.S. Energy Information Administration.
This gap between what is needed and what is available is the source of the DOE’s warning. Without sufficient dispatchable capacity to back up renewables during periods of low generation or peak demand, grid operators may have no choice but to implement controlled outages to prevent catastrophic failures.
Pragmatism over Policy: The Market Reaction
The market is already reacting to this impending reality. Several utility companies have announced delays to their planned coal plant retirements, prioritising reliability over previously stated climate timelines. Wisconsin’s WEC Energy Group and Alliant Energy, for instance, have postponed closures, citing concerns over maintaining adequate power reserves.5 This trend is likely to accelerate, creating a messy but necessary state of affairs where legacy fossil fuel assets are kept online longer than anticipated.
For investors, this complex environment presents clear risks and opportunities. Technology firms heavily reliant on AI may face rising energy costs and will need to factor grid stability into their location and infrastructure decisions. Conversely, companies providing essential bridging solutions stand to benefit. This includes natural gas producers and pipeline operators, as gas is the most viable fuel for providing large-scale, dispatchable power to complement renewables. There is also a renewed, and more serious, interest in nuclear energy, including small modular reactors (SMRs), as a long-term source of carbon-free baseload power. Finally, companies focused on grid modernisation, from advanced transmission technologies to energy storage solutions, are critical to managing this difficult transition.
The warnings about grid fragility are not abstract or distant; they are a direct consequence of quantifiable trends in supply and demand. The coming decade will be defined by a fundamental trade-off between energy transition goals and the immediate economic need for reliable power. As a speculative hypothesis, the most significant policy shift will not come from Washington, but from state-level utility commissions that will increasingly approve life extensions for existing gas and even coal plants under the non-negotiable mandate to “keep the lights on,” making the energy transition a slower and more fossil-fuel-dependent process than many currently forecast.
References
1. Department of Energy. (2024, July). Ensuring Resource Adequacy in a New Era of Growing Electricity Demand. Retrieved from https://www.energy.gov/
2. Boston Consulting Group. (2024, March). AI’s Unquenchable Thirst for Energy. Retrieved from https://www.bcg.com/publications/2024/ai-power-surge-data-center-energy-demand
3. PJM Interconnection. (2023, December). PJM Load Forecast Report. Retrieved from PJM’s official reports portal.
4. U.S. Energy Information Administration (EIA). (2024, January). Annual Coal Report. Retrieved from https://www.eia.gov/coal/annual/
5. Daily Climate. (2024, July). DOE warns aging coal plants may stay online as AI drives up power demand. Retrieved from https://dailyclimate.org/doe-warns-aging-coal-plants-may-stay-online-as-ai-drives-up-power-demand-2673045600.html
6. Bloomberg. (2024, July 7). US Warns of Blackout Risk From Killing Coal as Trump Snubs Renewables. Retrieved from https://www.bloomberg.com/news/articles/2024-07-07/us-warns-of-blackout-risk-from-killing-coal-as-trump-snubs-renewables
7. @unusual_whales. (2024, July 7). [Trump administration warns that U.S. blackouts could double by 2030 due to increased power demand driven by AI and coal plant closures, per Bloomberg]. Retrieved from https://x.com/unusual_whales/status/1809980893325603043
