Powering the 21st Century Digital Surge: How AI, Crypto, and EVs are Rewiring the Grid

Introduction: The Grid of Today, Tomorrow

The United States electrical grid, arguably the most complex machine ever built, stands at a precipice. For the first two decades of the 21st century, the narrative of the American power sector was one of decoupling: economic growth continued while electricity demand remained largely stagnant, thanks to significant gains in energy efficiency and the structural shift away from heavy manufacturing. This era of stagnation has abruptly ended. The grid is now facing a tripartite convergence of aggressive new load drivers—Artificial Intelligence (AI), cryptocurrency mining, and Electric Vehicles (EVs)—that threaten to overwhelm the physical and market mechanisms designed to manage them.

This report provides an exhaustive analysis of how these distinct yet compounding pressures are reshaping the US power grid. Unlike previous periods of load growth, which were broadly correlated with population and GDP, the current surge is characterized by extreme localization, differing load physics, and a collision with an aging, constrained infrastructure. AI data centers are creating hyper-dense load pockets in regions like Northern Virginia, driving wholesale capacity prices to historic highs and forcing a re-evaluation of fossil fuel retirements. Cryptocurrency mining, particularly in the ERCOT (Texas) market, introduces gigawatt-scale, price-responsive loads that offer theoretical flexibility but practical volatility. Simultaneously, the electrification of the transportation sector is creating a distributed crisis at the "grid edge," threatening to thermally overload millions of residential distribution transformers and exacerbating the ramping challenges of the "duck curve" in renewable-heavy markets like California.

The analysis synthesizes data from federal agencies (EIA, FERC, DOE), national laboratories (NREL, LBNL), grid operators (PJM, ERCOT, CAISO), and industry research. It concludes that the "business as usual" approach to utility planning—characterized by slow, retrospective rate cases and incremental infrastructure build-out—is fundamentally incompatible with the speed and magnitude of the digital and electrified transition. The grid requires a radical architectural shift involving the deployment of Grid Enhancing Technologies (GETs), the commercialization of Virtual Power Plants (VPPs), and arguably, a return to firm baseload generation through the restart of nuclear assets, as exemplified by the recent Microsoft-Constellation agreement for Three Mile Island.

1. The End of Stagnation: The New Physics of Load Growth

1.1 The Statistical Reversal and Forecasting Shock

For nearly twenty years, the US power grid operated in a regime of flat demand. Annual growth rates hovered below 0.5%, and in many years, total consumption actually declined despite economic expansion. This stability allowed utilities and grid operators to prioritize decarbonization and maintenance over massive capacity expansion. That paradigm has been shattered. The Energy Information Administration (EIA) now forecasts that US annual electricity consumption will increase in 2025 and 2026, surpassing the all-time highs reached in previous years.¹

The magnitude of this shift is difficult to overstate and has caught the planning community off guard. In 2022, utility forecasts filed with FERC projected a five-year cumulative demand growth of approximately 24 gigawatts (GW). By 2025, those same forecasts had been revised to project 166 GW of growth over the subsequent five years—a nearly seven-fold increase in expected demand.² This is not merely an incremental adjustment; it represents a structural break in the forecasting models used to plan multi-billion-dollar infrastructure investments. The North American Electric Reliability Corporation (NERC), the body responsible for grid reliability, has explicitly identified this rapid load growth as a significant near-term reliability challenge, noting that demand growth is now higher than at any point in the past two decades.³

1.2 The Divergence of Load Characteristics

While often grouped together under the banner of "electrification" or "digitalization," the three primary drivers—AI, Crypto, and EVs—exert fundamentally different physical pressures on the grid. Understanding these distinctions is critical for effective policy and engineering responses.

Table 1: Comparative Physics of Emerging Loads

Characteristic	AI Data Centers	Cryptocurrency Mining	Electric Vehicles
Grid Level	Transmission / Sub-transmission	Transmission (mostly)	Distribution (Edge)
Reliability Requirement	Tier 4 (99.999% uptime)	Flexible / Interruptible	Flexible (if managed)
Load Profile	Flat, continuous baseload	Flat, but price-responsive	"Bursty," coincident peaks
Location Strategy	Latency-dependent (near users) or Power-dependent	Pure arbitrage (chases cheapest power)	Population-dependent (where people live)
Thermal Impact	High localized heat density	High localized heat density	Distributed transformer heating
Grid Service Potential	Low (requires firm power)	High (Demand Response)	High (V2G / V1G storage)

1.3 The Capacity Crisis and the "Phantom" Queue

The surge in demand is colliding with a grid that is struggling to add new supply. The queue of power generation projects waiting to connect to the grid has ballooned to over 10,000 projects representing 1,400 GW of generation and 890 GW of storage.⁵ However, the throughput of this queue is abysmal. The typical project built in 2024 took 55 months—nearly five years—from the initial interconnection request to commercial operation.⁶

This bottleneck creates a dangerous temporal mismatch: data centers and crypto mines can be built in 18 to 24 months, while the transmission lines and generation assets needed to power them take 5 to 10 years to permit and construct.³ This discrepancy forces grid operators to rely on existing, often aging, fossil fuel assets to bridge the gap, complicating decarbonization goals. Furthermore, the "speculative" nature of many interconnection requests—where developers flood the queue with projects that may never be built—complicates the ability of planners to discern real capacity from "phantom" capacity, leading to inefficiencies and paralysis in grid expansion.³

2. The Digital Baseload: Artificial Intelligence and Data Center Architecture

2.1 The AI Energy Multiplier Effect

The transition from traditional cloud computing to Generative AI (GenAI) represents a phase change in energy intensity. A standard Google search requires roughly 0.3 watt-hours of energy. A query to a GenAI model like ChatGPT requires approximately 2.9 watt-hours—nearly ten times the energy input.⁷ When scaled across billions of daily interactions, this multiplier effect drives massive aggregate demand.

Data centers consumed approximately 4.4% of total US electricity in 2023. Forecasts suggest this could rise to between 6.7% and 12% by 2028, a doubling or tripling of sector-wide consumption in just five years.⁹ This growth is not linear; it is exponential, driven by the escalating computational requirements of larger models. The infrastructure required to support this is immense: a single hyperscale AI campus can now demand loads ranging from 100 MW to over 500 MW, rivaling the consumption of mid-sized cities or heavy industrial smelting plants.¹⁰

2.2 The Jevons Paradox in Silicon

A critical economic mechanism driving this consumption is the Jevons Paradox. First observed in the 19th century regarding coal efficiency, the paradox states that as technological improvements increase the efficiency with which a resource is used, the total consumption of that resource increases rather than decreases.

In the context of AI, hardware manufacturers like NVIDIA are relentlessly improving the energy efficiency of their Graphics Processing Units (GPUs). The transition from the H100 to the H200 chip, for instance, offers greater performance per watt. However, these efficiency gains lower the cost of compute, making new applications viable and encouraging the training of even larger, more complex models.¹¹

• The Rebound Effect: Instead of banking the energy savings, the industry "spends" them on more compute. The lowered barrier to entry allows startups, researchers, and enterprises to deploy AI into new domains, broadening the user base and increasing the total number of inference queries.

• Result: The grid faces a "DeepSeek moment" where efficiency gains accelerate, rather than dampen, the aggregate demand curve.¹² The World Economic Forum warns that without a deliberate shift to "Net-Positive AI"—where AI is used to optimize the grid itself—total consumption could triple to 1,200 TWh globally by 2035.¹³

2.3 Training vs. Inference: The Bifurcation of Load

To understand the grid impact, one must distinguish between the two primary modes of AI operation: Training and Inference.

AI Training:

• Process: Feeding massive datasets into a model to teach it patterns. This can take weeks or months.

• Load Profile: Extremely high intensity, continuous, and "bursty." It requires massive clusters of GPUs running at peak thermal design power (TDP).¹⁴

• Geography: Training is generally not latency-sensitive. A training cluster can be located in rural North Dakota, West Texas, or wherever power is cheapest and cooling is efficient. This offers some flexibility to grid planners to site these loads in areas with excess renewable capacity.

AI Inference:

• Process: The model responding to user inputs in real-time.

• Load Profile: Continuous, fluctuating with user traffic patterns (diurnal cycles), and critically latency-sensitive.

• Geography: To minimize lag, inference servers must be located close to end-users (the "edge") or internet exchange points. This forces these massive loads into already congested urban and suburban corridors like Northern Virginia, Silicon Valley, and the Dallas-Fort Worth metroplex.¹⁴

2.4 The Northern Virginia Case Study: PJM's Crisis

Northern Virginia, specifically Loudoun County, serves as the global epicenter of the data center industry, often called "Data Center Alley." The density of infrastructure here is staggering, and it is the canary in the coal mine for grid congestion. Dominion Energy, the utility serving the region, has reported that 7 out of its 54 largest customers account for 72% of the year-to-date demand growth.¹⁶

The concentration of load in this specific pocket of the PJM Interconnection (the regional grid operator) has created severe transmission constraints. There simply isn't enough wire capacity to import the necessary power. The result has been a violent price signal in the capacity market.

• Price Explosion: In the PJM capacity auction for the 2024/2025 delivery year, prices cleared around $28.92 per megawatt-day. For the 2025/2026 auction, prices skyrocketed to approximately $269.92 per megawatt-day—a nearly ten-fold increase.¹⁷

• Implication: This price spike is a direct market response to scarcity. It signals that the region is dangerously short of firm generation capacity. Consequently, PJM and Dominion have been forced to delay the retirement of fossil fuel plants and are proposing massive new transmission lines, including 765 kV backbones, to import power from Ohio and West Virginia.¹⁸

• Local Impact: Residents and other states are beginning to revolt. Pennsylvania and Maryland have launched legal challenges, arguing that their ratepayers are effectively subsidizing the infrastructure upgrades required to serve Virginia's data center boom, raising profound equity questions about the allocation of transmission costs.¹⁹

3. The Cryptographic Load: Bitcoin Mining in the ERCOT Market

3.1 Texas as the Global Mining Capital

Following the Chinese ban on cryptocurrency mining in 2021, the United States, and specifically Texas, emerged as the global hub for Bitcoin mining. The state offered a perfect trifecta: a deregulated energy market (ERCOT), a "friendly" regulatory environment, and abundant, often stranded, renewable energy in West Texas.

The scale of this migration is immense. ERCOT has reported 41 GW of requests for new cryptocurrency mining capacity in its interconnection queue, with 9 GW of planning studies already approved.²⁰ To put this in perspective, 9 GW is roughly equivalent to the peak demand of the entire state of Minnesota.

3.2 The "Large Flexible Load" (LFL) Paradigm

Unlike AI data centers, which require Tier 4 reliability and cannot tolerate interruptions, cryptocurrency mines are uniquely flexible. The mining process (Proof-of-Work) is probabilistic; a mine can turn off 90% of its machines instantly without "breaking" the process—they simply stop earning Bitcoin for that duration.

This characteristic has led to the creation of the "Large Flexible Load" (LFL) designation in ERCOT. Miners essentially act as a grid reliability tool:

1. Normal Operations: They consume massive amounts of baseload power, soaking up excess wind and solar generation in West Texas that might otherwise be curtailed due to negative pricing.

2. Scarcity Events: When the grid is stressed (e.g., during a heatwave or winter storm), prices spike. Miners, who are economically sensitive, voluntarily curtail their load to avoid high costs or are paid via demand response programs (like the Responsive Reserve Service) to shut down.²¹

3.3 The Economic and Reliability Double-Edged Sword

While the flexibility of crypto miners is touted as an asset, their presence introduces significant volatility and cost to the system.

• Wholesale Price Pressure: Even if miners shut down during peaks, their presence during the other 95% of the year increases the "base" demand. This shifts the supply curve, forcing more expensive marginal generators (gas peakers) to run more often, which raises wholesale prices for all customers. Research indicates that without policy intervention, the demand growth from data centers and crypto mining could increase wholesale electricity costs by 8% nationally.¹⁹

• Forecasting Difficulty: The erratic behavior of miners—who may shut down based on the price of Bitcoin rather than grid conditions—complicates load forecasting. ERCOT has had to implement a "Large Load Adjustment" to its methodology to account for this uncertainty. If Bitcoin prices surge, miners may be willing to pay higher electricity prices, refusing to curtail even during grid stress, potentially threatening reliability.²²

• Transmission Congestion: By locating in remote areas to chase cheap wind power, miners can exacerbate local transmission constraints, using up the line capacity that was intended to export renewable energy to urban centers. This "cannibalization" of transmission capacity can hinder the broader integration of renewables.²¹

3.4 Regulatory Divergence: Texas vs. New York

The political response to crypto mining is bifurcated. Texas has largely embraced the industry, passing legislation like HB 5066 to facilitate transmission planning for these loads.²² In contrast, New York has taken a hostile stance, passing a moratorium on new Proof-of-Work mining operations that rely on carbon-based fuel, citing environmental concerns and the revival of dormant fossil fuel plants for private gain.²¹ This regulatory patchwork is driving miners toward "crypto-friendly" grids like ERCOT and SPP (Southwest Power Pool), further concentrating the load and the associated risks.

4. The Electrification of Transport: A Distributed Crisis

4.1 From Transmission to Distribution

While AI and crypto stress the high-voltage transmission grid, the electrification of transportation (EVs) creates a crisis at the distribution level—the "last mile" of wires and transformers that deliver power to homes. The aggregate energy demand of EVs is significant but manageable on a national generation scale. The acute danger lies in the coincidence and locality of the demand.²³

4.2 The Physics of Transformer Failure

The residential distribution transformer—the canister mounted on utility poles or sitting on green boxes in front yards—is the weakest link in the EV transition. These devices are sized based on diversified load assumptions: planners assume that in a cluster of 5-10 homes, not everyone will run their AC, dryer, and oven simultaneously.

An EV charger breaks this assumption. A typical Level 2 home charger draws roughly 7 kW continuous load. This is equivalent to running a central air conditioning unit or an electric oven at full power. However, unlike an oven that cycles on and off, an EV charger draws this high load continuously for hours.

• Clustering: Socio-economic factors mean EV adoption is not random; it is clustered. Neighbors influence neighbors. If three households on a single 25 kVA transformer buy EVs and plug them in at 6:00 PM, the transformer load can easily exceed 150% of its rating.

• Thermal Dynamics: Transformers degrade based on heat. The Arrhenius equation dictates that for every 6°C rise in hotspot temperature above the design limit (typically 110°C), the insulation life of the transformer is cut in half.

• The "Cool-Down" Loss: Traditionally, transformers rely on the cool nighttime hours to dissipate the heat generated during the day. EV charging, which typically happens at night, eliminates this recovery period. The transformer stays hot, accelerating insulation degradation. A transformer designed to last 40 years may fail in less than 5 years under unmanaged EV clustering.²⁴

4.3 The "Duck Curve" and Ramping Stress

In markets with high solar penetration like California (CAISO), EVs exacerbate the "Duck Curve." The Duck Curve describes the shape of the net load chart: low demand during the day (when solar is generating) and a steep ramp up in the evening (when solar sets and people return home).

• The Collision: Unmanaged EV charging typically begins between 5:00 PM and 7:00 PM—exactly when the sun is setting and the grid is already under maximum stress to ramp up gas generators to replace solar.

• The Scale: A study by the University of Washington suggests that in a worst-case scenario with full EV adoption and unmanaged charging, the evening peak demand in the Western US could increase by 25%, requiring hundreds of gigawatts of new peaking capacity.²⁷

• Curtailment: Conversely, during the day, the grid often has too much solar power, leading to negative prices and curtailment (wasted energy). In the first four months of 2025 alone, CAISO curtailed more than 738,000 MWh of clean energy.²⁸

4.4 Harmonics and Power Quality

Beyond simple power demand, EVs introduce "dirty" power issues. EV chargers use power electronics (rectifiers/inverters) to convert AC grid power to DC battery power. These non-linear loads inject harmonic distortions into the grid.

• Implication: High Total Harmonic Distortion (THD) causes additional heating in transformer cores and neutral wires, independent of the actual power load. This "invisible" heating further de-rates the capacity of grid equipment and can cause malfunctions in sensitive electronic equipment in neighboring homes.²⁹

4.5 The Equity and Cost Dilemma

Upgrading the distribution grid to handle this load is enormously expensive. The California Public Advocates Office estimates that distribution grid upgrades for electrification could cost between $15 billion and $37 billion by 2040.³¹

• Regressive Impact: These costs are typically recovered through general rate increases. This raises a profound equity issue: low-income ratepayers, who are less likely to own EVs, may end up subsidizing the transformer upgrades required for their wealthier neighbors' electric cars. This "equity gap" is a major policy hurdle for regulators trying to balance decarbonization with affordability.³³

5. Infrastructure Realities: The Hardware and Supply Chain Deficit

5.1 The Aging Asset Base

The ambitions of the energy transition are colliding with the reality of a geriatric industrial base. The US grid infrastructure earned a "D+" grade from the American Society of Civil Engineers.³⁴

• Transformers: Approximately 70% of the nation's large power transformers (LPTs) are over 25 years old. The average age of distribution transformers is over 33 years. With a typical design life of 40 years, a massive wave of end-of-life failures is statistically imminent, precisely as load stress is maximizing.³⁵

• Transmission Lines: 70% of transmission lines are also over 25 years old. These lines were designed for a one-way flow of power from central plants to cities, not the dynamic, bi-directional flows required by VPPs and renewable integration.³⁴

5.2 The Supply Chain Crisis and National Security

Compounding the age problem is a severe supply chain crisis. The manufacturing base for transformers has atrophied domestically.

• Lead Times: The lead time for procuring a new Large Power Transformer (LPT) has extended from months to years. Current wait times can exceed 12 to 24 months, with some specialized units taking up to three years.³⁶

• Vulnerability: This scarcity creates a national security vulnerability. A physical attack on substations (like the Moore County incident) or a severe weather event could destroy transformers that cannot be replaced for over a year, leading to long-duration regional blackouts. The Department of Energy has identified LPTs as one of the most vulnerable components of the grid.³⁶

• Cost Inflation: The shortage has driven prices up by 40-60%, blowing holes in utility capital budgets and ultimately increasing costs for consumers.³⁵

5.3 The Interconnection Queue Nightmare

The process of connecting new generation to the grid is broken. The "Interconnection Queue"—the waiting list for new power plants—is overwhelmed.

• Volume: As of the end of 2024, there were ~10,300 projects seeking interconnection.

• Throughput: Only ~20% of projects that enter the queue are ever built. The rest withdraw due to unexpected costs.

• Cost Allocation: The primary killer of projects is the "cost causation" principle, where the developer of a new solar farm is asked to pay for massive upgrades to the transmission system (e.g., a $100 million substation upgrade) that benefit the whole grid. This "first-mover disadvantage" leads to a game of chicken where developers drop out, triggering a cascading restudy of all other projects in the queue.⁶

6. The Generation Response: Gas, Nuclear, and the Future

6.1 The Natural Gas Bridge

Faced with the immediate need for firm capacity to serve data centers, utilities are pausing their retreat from fossil fuels. In 2024, active natural gas capacity in the interconnection queue increased by 72% year-over-year.³⁸ Grid operators like PJM are explicitly citing data center load growth as the justification for delaying the retirement of coal and gas plants.³⁹ This represents a pragmatic but controversial pivot: climate goals are being temporarily subordinated to the imperative of reliability and the economic pressure of the AI race.

6.2 The Nuclear Renaissance: The Microsoft-Constellation Deal

The most significant market signal in recent years is the renewed embrace of nuclear power by the technology sector. Nuclear energy is the only carbon-free source capable of providing the 24/7, high-capacity-factor (90%+) power that data centers require.

• The Deal: In September 2024, Microsoft signed a 20-year Power Purchase Agreement (PPA) with Constellation Energy to restart the Three Mile Island Unit 1 reactor in Pennsylvania. The plant, to be renamed the Crane Clean Energy Center, had been retired in 2019 for economic reasons.

• Significance: This deal changes the economics of nuclear power. Microsoft is effectively paying a "green premium" for firm, carbon-free power to meet its "carbon negative" goals. This "behind-the-meter" or dedicated-use model allows the tech giant to bypass the volatility of the wholesale market and secure long-term energy security.⁴⁰

• Implication: This sets a precedent for other retired nuclear assets (like the Palisades plant in Michigan) to be brought back online, funded by private capital from tech hyperscalers rather than public utility ratepayers.⁴³

6.3 Small Modular Reactors (SMRs): Hype vs. Reality

Beyond restarting old plants, tech companies like Google and Amazon are investing in Small Modular Reactors (SMRs). These are next-generation reactors, typically under 300 MW, designed to be factory-fabricated and deployed in clusters.

• The Vision: Co-locate SMRs directly with data center campuses, eliminating the need for long transmission lines and insulating the facility from grid outages.⁴⁴

• The Skepticism: While promising, SMRs face significant hurdles. Critics point out that "modular" does not automatically mean "cheap." The "nth-of-a-kind" cost savings rely on a volume of orders that doesn't exist yet. Furthermore, regulatory requirements (e.g., NRC rules requiring control room operators) may damage the unit economics of smaller plants. There is also the unresolved issue of waste disposal, which becomes more complex with distributed nuclear sites.⁴⁶

7. Grid Management and Policy: The Software Infrastructure

7.1 Virtual Power Plants (VPPs)

To manage the distributed chaos of EVs and the flexibility of crypto, the grid is turning to software. Virtual Power Plants (VPPs) aggregate thousands of distributed energy resources (DERs)—smart thermostats, home batteries, EV chargers—and control them as a single, dispatchable power plant.

• Mechanism: When the grid is stressed, the VPP operator sends a signal to reduce demand (dimming lights, pausing EV charging) or discharge batteries.

• Progress: In ERCOT and California, VPPs are now qualified to participate in wholesale markets alongside gas plants. Companies like Voltus and aggregators in Texas are effectively creating gigawatts of "virtual" capacity that is cheaper and faster to deploy than new physical power plants.⁴⁸

• EV Integration: "Smart Charging" (V1G) is the killer app for VPPs. By shifting EV charging from 6:00 PM to 2:00 AM, VPPs can mitigate the transformer aging crisis and the duck curve without requiring new hardware.²³

7.2 Grid Enhancing Technologies (GETs)

With transmission lines taking a decade to build, utilities are deploying Grid Enhancing Technologies to squeeze more capacity out of existing wires.

• Dynamic Line Rating (DLR): Traditional line ratings are static, based on conservative assumptions about hot, still days. DLR uses real-time sensors to monitor wind and temperature. On a windy day, a line might be able to carry 30-40% more current because the wind cools the conductor. Since wind generation coincides with wind, DLR unlocks massive capacity exactly when it's needed.⁵¹

• Topology Optimization: Software that reconfigures the grid switching in real-time to route power around congested bottlenecks, maximizing the utilization of the existing mesh network.

7.3 The Federal-State Conflict

The divergence in load growth is straining the federalist structure of US energy regulation.

• Cost Allocation: Who pays for the new transmission lines? In PJM, states like New Jersey and Maryland are pushing back against cost-sharing for projects that primarily benefit Virginia's data center industry. This has led to lawsuits and regulatory infighting that threatens to stall necessary upgrades.¹⁹

• FERC Order 2023: The Federal Energy Regulatory Commission has issued orders to clear the interconnection backlog, imposing stricter penalties on developers who withdraw projects and moving to a "cluster study" approach (studying groups of projects together) rather than one-by-one. While a step in the right direction, the implementation is complex and slow.⁵

Conclusion: Synthesizing the Divergence

The United States power grid is in the early stages of a traumatic but necessary metamorphosis. The convergence of Artificial Intelligence, Cryptocurrency, and Electric Vehicles is not merely a story of "more demand." It is a story of diverging demand.

• AI demands hyper-centralized, hyper-reliable, firm baseload power, driving a renaissance in nuclear energy and a stubborn retention of fossil fuels.

• Cryptocurrency acts as a volatile, price-responsive arbitrageur, offering economic flexibility but introducing operational uncertainty and wholesale price inflation.

• EVs demand a hyper-distributed, edge-focused upgrade cycle, requiring millions of small interventions at the transformer level and sophisticated software orchestration to prevent grid failure.

The "business as usual" model—characterized by slow-moving utility monopolies, retrospective rate-making, and years-long permitting processes—is structurally incapable of meeting these simultaneous challenges. The data suggests that the successful grid of 2035 will look radically different: it will be a hybrid machine, combining the massive, firm generation of revived nuclear giants with a nimble, software-defined edge network of Virtual Power Plants.

Achieving this requires a coordinated policy response:

1. Radical Transparency: Mandating better data sharing on interconnection queues and load forecasts to eliminate "phantom" projects.

2. Market Evolution: transitioning from simple volumetric pricing to dynamic, time-and-location-based tariffs that incentivize VPP participation.

3. Industrial Policy: Direct investment in the domestic supply chain for transformers and critical grid hardware to close the national security gap.

4. Nuclear Realism: Acknowledging that the data center boom makes a 100% renewable grid physically and economically largely impractical in the near term, necessitating the preservation and expansion of nuclear capacity.

The alternative is a future of rising costs, frequent constrained supply, and a stalled digital and energy transition. The physics of the grid will not negotiate; the only variable is how quickly the US regulatory and industrial complex can adapt to the new reality.

Works cited

1. After more than a decade of little change, U.S. electricity consumption is rising again - U.S. Energy Information Administration (EIA), accessed January 14, 2026, https://www.eia.gov/todayinenergy/detail.php?id=65264

2. Power Demand Forecasts Revised Up for Third Year Running, Led by Data Centers - Grid Strategies, accessed January 14, 2026, https://gridstrategiesllc.com/wp-content/uploads/Grid-Strategies-National-Load-Growth-Report-2025.pdf

3. Rapid Data Center Development Challenges Grid Reliability, NERC Report Says - NRECA, accessed January 14, 2026, https://www.electric.coop/rapid-data-center-development-challenges-grid-reliability-nerc-report-says

4. 2024 Long-Term Reliability Assessment, accessed January 14, 2026, https://www.nerc.com/globalassets/our-work/assessments/2024-ltra_corrected_july_2025.pdf

5. Queued Up: Characteristics of Power Plants Seeking Transmission Interconnection | Energy Markets & Planning - Lawrence Berkeley National Laboratory, accessed January 14, 2026, https://emp.lbl.gov/queues

6. Queued Up: 2025 Edition - Characteristics of Power Plants Seeking Transmission Interconnection As of the End of 2024, accessed January 14, 2026, https://eta-publications.lbl.gov/sites/default/files/2025-12/queued_up_2025_edition_12.15.2025.pdf

7. As generative AI asks for more power, data centers seek more reliable, cleaner energy solutions - Deloitte, accessed January 14, 2026, https://www.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2025/genai-power-consumption-creates-need-for-more-sustainable-data-centers.html

8. AI Energy Consumption: Statistics from Key Sources [2026] - Research AIMultiple, accessed January 14, 2026, https://research.aimultiple.com/ai-energy-consumption/

9. DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers, accessed January 14, 2026, https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers

10. "Nuclear Powered Artificial Intelligence (AI): Small Modular Reactors as an Emerging Power Source for AI Data Centers" - Shumaker, Loop & Kendrick, LLP, accessed January 14, 2026, https://www.shumaker.com/insight/nuclear-powered-artificial-intelligence-ai-small-modular-reactors-as-an-emerging-power-source-for-ai-data-centers/

11. How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference, accessed January 14, 2026, https://arxiv.org/html/2505.09598v6

12. The Jevons Paradox: Why Efficiency Alone Won't Solve Our Data Center Carbon Challenge, accessed January 14, 2026, https://www.sigarch.org/the-jevons-paradox-why-efficiency-alone-wont-solve-our-data-center-carbon-challenge/

13. AI’s Power Problem: Why the WEF Wants Net-Positive Energy Models, accessed January 14, 2026, https://openthemagazine.com/technology/ais-power-problem-why-the-wef-wants-net-positive-energy-models

14. AI Inference vs. Training – What Hyperscalers Need to Know - EdgeCore Digital Infrastructure, accessed January 14, 2026, https://edgecore.com/ai-inference-vs-training/

15. AI Inference Costs in 2025: The $255B Market's Energy Crisis and Path to Sustainable Scaling | Tensormesh, accessed January 14, 2026, https://www.tensormesh.ai/blog-posts/ai-inference-costs-2025-energy-crisis

16. Dominion Energy Virginia - PJM.com, accessed January 14, 2026, https://www.pjm.com/-/media/DotCom/committees-groups/subcommittees/las/2025/20250916/20250916-item-04ai---dominion-data-center-large-load-request.pdf

17. Projected data center growth spurs PJM capacity prices by factor of 10 | IEEFA, accessed January 14, 2026, https://ieefa.org/resources/projected-data-center-growth-spurs-pjm-capacity-prices-factor-10

18. PJM Reviews Preliminary Recommended Projects for 2025 RTEP Window 1, accessed January 14, 2026, https://insidelines.pjm.com/pjm-reviews-preliminary-recommended-projects-for-2025-rtep-window-1/

19. Electricity Grid Impacts of Rising Demand from Data Centers and Cryptocurrency Mining Operations - Scott Institute for Energy Innovation, accessed January 14, 2026, https://energy.cmu.edu/_files/documents/electricity-grid-impacts-of-rising-demand-from-data-centers-and-cryptocurrency-mining-operations.pdf

20. Bitcoin Mining's Surging Demand Strains US Power Grids Amid Energy Transition, accessed January 14, 2026, https://www.globalelectricity.org/bitcoin-minings-surging-demand-strains-us-power-grids-amid-energy-transition/

21. Bitcoin and Large Loads in ERCOT | Thought Leadership - Baker Botts, accessed January 14, 2026, https://www.bakerbotts.com/thought-leadership/publications/2023/april/2022-energy-memo_bitcoin-and-large-loads-in-ercot

22. ERCOT Releases Capacity, Demand and Reserves (CDR) Report; Planning Reserves Pressured as Texas Economy Grows, accessed January 14, 2026, https://www.ercot.com/news/release/02132025-ercot-releases-capacity

23. Electric Vehicle Smart-Charge Management and Flexibility Analysis | Transportation and Mobility Research | NLR - NREL, accessed January 14, 2026, https://www.nrel.gov/transportation/smart-charge-management-flexibility-analysis

24. Analyzing the influence of electric vehicle charging scheduling on distribution transformer lifespan - PMC - NIH, accessed January 14, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC11566691/

25. Reducing the impacts of electric vehicle charging on power distribution transformers - Edith Cowan University, accessed January 14, 2026, https://ro.ecu.edu.au/context/ecuworkspost2013/article/10269/viewcontent/Reducing_the_impacts_of_electric_vehicle_charging.pdf

26. Avoiding gridlock: The distribution impacts of EV charging - EnergyHub, accessed January 14, 2026, https://www.energyhub.com/news/avoiding-gridlock-the-distribution-impacts-of-ev-charging

27. How Would EV Charging Demands Affect the Electric Grid in California? | Sustainable Transportation Lab - UW Sites - University of Washington, accessed January 14, 2026, https://sites.uw.edu/stlab/2023/06/15/how-would-ev-charging-demands-affect-the-electric-grid-in-california/

28. The Duck Curve Explained: Impacts, Renewable Energy Curtailments, and Market Strategies, accessed January 14, 2026, https://www.yesenergy.com/blog/the-duck-curve-explained-impacts-renewable-energy-curtailments

29. Evaluating Effects of Electric Vehicle Chargers on Residential Power Infrastructure - MDPI, accessed January 14, 2026, https://www.mdpi.com/2076-3417/15/11/5997

30. Power quality impact of electric vehicle charging station on utility grid - ResearchGate, accessed January 14, 2026, https://www.researchgate.net/publication/355441778_Power_quality_impact_of_electric_vehicle_charging_station_on_utility_grid

31. Costs of Upgrading the Distribution Grid for Electrification - Fact Sheet, accessed January 14, 2026, https://www.publicadvocates.cpuc.ca.gov/press-room/reports-and-analyses/costs-of-upgrading-the-distribution-grid-for-electrification---fact-sheet

32. Distribution Grid Electrification Model (DGEM) 2.0 - Public Advocates Office - CA.gov, accessed January 14, 2026, https://www.publicadvocates.cpuc.ca.gov/press-room/reports-and-analyses/distribution-grid-electrification-model-2024

33. policy brief: The Equity Challenge: Ensuring Grid Upgrades Don't Leave Communities Behind in California, accessed January 14, 2026, https://ucits.org/research_products/policy-brief-the-equity-challenge-ensuring-grid-upgrades-dont-leave-communities-behind-in-california/

34. How An Aging Grid, AI, And Electrification Are Impacting Energy Rates - VECKTA, accessed January 14, 2026, https://veckta.com/2025/10/22/how-an-aging-grid-ai-and-electrification-are-impacting-energy-rates/

35. Increased Transformer Demand and Aging Grids, accessed January 14, 2026, https://www.maddox.com/resources/articles/increased-transformer-demand-and-aging-grids

36. Addressing Security and Reliability Concerns of Large Power Transformers, accessed January 14, 2026, https://www.energy.gov/oe/addressing-security-and-reliability-concerns-large-power-transformers

37. Securing America's grid through transformers and workforce resilience - Breaking Defense, accessed January 14, 2026, https://breakingdefense.com/2025/11/securing-americas-grid-through-transformers-and-workforce-resilience/

38. Queued Up: 2025 Edition, Characteristics of Power Plants Seeking Transmission Interconnection As of the End of 2024 | Energy Markets & Planning - Lawrence Berkeley National Laboratory, accessed January 14, 2026, https://emp.lbl.gov/publications/queued-2025-edition-characteristics

39. Powering the US Data Center Boom: Why Forecasting Can Be So Tricky, accessed January 14, 2026, https://www.wri.org/insights/us-data-centers-electricity-demand

40. Constellation to restart Three Mile Island unit, powering Microsoft - World Nuclear News, accessed January 14, 2026, https://www.world-nuclear-news.org/articles/constellation-to-restart-three-mile-island-unit-powering-microsoft

41. Constellation to Launch Crane Clean Energy Center, Restoring Jobs and Carbon-Free Power to The Grid, accessed January 14, 2026, https://www.constellationenergy.com/newsroom/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html

42. The Atomic AI Renaissance: Why Tech Giants are Betting on Nuclear to Power the Future of Silicon, accessed January 14, 2026, https://markets.financialcontent.com/wral/article/tokenring-2026-1-12-the-atomic-ai-renaissance-why-tech-giants-are-betting-on-nuclear-to-power-the-future-of-silicon

43. Constellation plans 2028 restart of Three Mile Island unit 1, spurred by Microsoft PPA | Utility Dive, accessed January 14, 2026, https://www.utilitydive.com/news/constellation-three-mile-island-nuclear-power-plant-microsoft-data-center-ppa/727652/

44. Data Centres, Artificial Intelligence and Cryptocurrencies Eye Advanced Nuclear to Meet Growing Power Needs | International Atomic Energy Agency, accessed January 14, 2026, https://www.iaea.org/bulletin/data-centres-artificial-intelligence-and-cryptocurrencies-eye-advanced-nuclear-to-meet-growing-power-needs

45. Nuclear-Powered Data Centers | Small Modular Reactors & AI, accessed January 14, 2026, https://www.datacenters.com/news/exploring-small-modular-reactors-as-the-next-energy-frontier

46. Expert assessments of the cost of light water small modular reactors - PNAS, accessed January 14, 2026, https://www.pnas.org/doi/10.1073/pnas.1300195110

47. Five Things the “Nuclear Bros” Don't Want You to Know About Small Modular Reactors, accessed January 14, 2026, https://blog.ucs.org/edwin-lyman/five-things-the-nuclear-bros-dont-want-you-to-know-about-small-modular-reactors/

48. Virtual Power Plant Services for DER Providers, Developers, & OEMs - Voltus, accessed January 14, 2026, https://www.voltus.co/der-owners

49. 'Virtual Power Plants' to Provide Power to ERCOT Grid for the First Time - Welcome to the Public Utility Commission of Texas FTP Server, accessed January 14, 2026, https://ftp.puc.texas.gov/public/puct-info/agency/resources/pubs/news/2023/PUCT_Virtual_Power_Plants_to_Provide_Power_to_ERCOT_Grid_for_the_First_Time.pdf

50. High-Fidelity Analysis of EV Integration on Real Utility Feeders in Colorado - Publications, accessed January 14, 2026, https://docs.nrel.gov/docs/fy26osti/94231.pdf

51. Grid-Enhancing Technologies: A Case Study on Ratepayer Impact - Department of Energy, accessed January 14, 2026, https://www.energy.gov/sites/default/files/2022-04/Grid%20Enhancing%20Technologies%20-%20A%20Case%20Study%20on%20Ratepayer%20Impact%20-%20February%202022%20CLEAN%20as%20of%20032322.pdf

52. Application of Distribution Transformer Thermal Life Models to Electrified Vehicle Charging Loads Using Monte-Carlo Method: Preprint - National Laboratory of the Rockies, accessed January 14, 2026, https://research-hub.nrel.gov/en/publications/application-of-distribution-transformer-thermal-life-models-to-el-2/

53. EIA projects US electricity demand growth through 2027 - POLITICO Pro, accessed January 14, 2026, https://subscriber.politicopro.com/article/2026/01/eia-projects-us-electricity-demand-growth-through-2027-00726148