I. The Prediction Meets Reality

On election night in Georgia, Crystal finally gave up trying to make the numbers work.

In Load to Ruin, she was the substitute teacher in Macon whose street transformer began humming after two neighbors bought electric cars — the stand-in for every ratepayer left holding the bill for progress. When the utility replaced the 25-year-old unit at a cost of roughly $10,000, it added about $11 a month to her $126 bill. That was before the latest round of increases.

This November, the envelope on her kitchen table carried two numbers that mattered more than any line item on her statement: the names on the Public Service Commission ballot.

For the first time in nearly twenty years, Georgia voters ousted two incumbent Republican commissioners. In District 3, Democrat Angela Speir Phelps defeated Lauren “Bubba” McDonald 61.4 to 38.6 percent; in District 4, Democrat Patty Durand defeated Fitz Johnson 63.1 to 36.9 percent. These were low-profile, down-ballot races that suddenly became a referendum on electric bills that, as one voter told Reuters, “look like monthly car payments.”

The backdrop was brutal. Georgia Power’s two new reactors at Plant Vogtle came online in 2023 and 2024, seven years late and at more than twice the original $14 billion budget — roughly $31 billion in total. Public Service Commission staff estimated customers will pay $36 billion to $43 billion more over the reactors’ lifetime than if the utility had built gas plants instead. That cost showed up quickly: summer bills for a typical household are now up 41 percent since 2021, driven by six rate increases in just two years.

On the same day, Jefferies Equity Research downgraded Southern Company from Buy to Hold, warning that the Republican loss of two PSC seats introduced material political risk to future rate-case outcomes and signaled that the era of automatic approvals for Georgia Power’s capital plans is likely over. Markets rarely care about down-ballot elections. This one, they did — because the entire cost-recovery model depends on political predictability.

In the same news cycle, Republican Ohio gubernatorial frontrunner Vivek Ramaswamy told his party that “lesson number one from this year’s elections: affordability — electric bills, groceries, housing, health care — is the issue that unites voters across party lines.”

Five months ago, Load to Ruin argued that affordability would become the defining political fault line in energy — that the real question isn’t how to build the grid, but who pays for it in an economy where AI is eroding the very wages that fund it.

Georgia just supplied the first clear signal.

Affordability now sits alongside reliability as voters’ statutory priority. Until regulators systematically track and price the relationship between employment, arrearages, and capital plans, today’s build-out risks hardening into tomorrow’s stranded debt — carried, once again, by people like Crystal.

⸻

II. The Georgia Signal in Context

Georgia didn’t just validate the affordability thesis; it made it visible.

For years, energy politics lived inside committee rooms — debated in IRP filings, modeled in LCOE spreadsheets, translated through rate-case consultants. But the Georgia election moved the conversation to the ballot box.

Voters didn’t parse cost-allocation tables; they responded to a pattern that had become undeniable. The grid was no longer a public utility that happened to finance private growth. It had become a private growth engine funded by the public.

The recognition came faster than anyone expected. The speed wasn’t ideological — it was structural.

Scale. Georgia Power projected 8,200 megawatts of new load over six years — 27 times its previous-decade forecast. Every ratepayer could feel the magnitude of that ask.

Sequence. Vogtle’s overruns and the AI-driven load surge merged into one continuous price shock — six rate filings in twenty-four months against a backdrop of generational inflation.

Accountability. Georgia’s commissioners face voters directly —a mechanism present in only 11 of the 50 states.

Together, they created the first political feedback loop that Load to Ruin had only described in theory:

Rising bills → voter backlash → market repricing → policy reset

Jefferies’ downgrade of Southern Company wasn’t a moral judgment; it was math. Political risk now sits beside fuel and interest-rate risk in utility valuations. You can already see it in analyst notes — language that pairs “customer sentiment” with “regulatory headroom” for the first time. That marks a new era: affordability as measurable uncertainty.

For decades, affordability crises were treated as cyclical. Georgia suggests they’re systemic — a permanent constraint running parallel to carbon and reliability.

Inside boardrooms, that realization is quietly changing the question from “How much can we build?” to “Who can pay for what we’ve already promised?”

Investors are beginning to model what Load to Ruin defines as electoral elasticity — the threshold where affordability pressure becomes electoral risk for regulators and measurable uncertainty for capital deployment and rate recovery.

Five months after the hypothesis, the affordability signal has left the lab. Now we see how different laboratories respond to the same experiment.

III. Two Laboratories, Two Futures

Georgia — Where Voters Still Have a Voice

The old guard in Atlanta is still stunned. Two incumbents gone, Jefferies’ downgrade still blinking across the Bloomberg terminals. No new policy yet — just a chill through every rate case in the queue. Whether that chill becomes reform or retrenchment will define the next decade.

Georgia showed that electoral accountability still works. Voters reached into the machine and pulled a lever that mattered. Wall Street noticed within hours. The feedback loop closed: rising bills became voter backlash became market repricing. Political risk is now measurable, which means it can be managed — if regulators act.

That’s the Georgia path: pressure forces adaptation. The system can still self-correct when voters have direct power to fire commissioners.

West Virginia — The Logic of Extraction

Governor Patrick Morrisey announced his “50 by 50” plan in Wheeling with the kind of certainty that comes from repetition: more capacity, more exports, more revenue. PJM’s capacity prices are up, data-center demand looks endless, and coal is the one asset West Virginia still controls. Build and sell. Lock in the contracts. Restart the engine.

In market terms, it’s rational. When PJM raises capacity payments, states with dispatchable generation see a short-term windfall: every new megawatt cleared earns regional revenue, no matter where it’s consumed. The budget benefits through severance taxes and payroll. On paper, it’s growth.

But the arithmetic is brutal. The plan would more than triple the state’s current generating capacity and require new high-voltage transmission to move it west into Ohio and north into Pennsylvania. Those lines don’t stop for the ratepayer in Morgantown or Parkersburg — they bypass them. The power will leave; the costs will stay. PJM’s regional cost-sharing spreads the expense across the footprint. Still, in practice, poorer states bear proportionally higher rate impacts because their load base is smaller and their household incomes are lower.

That’s the paradox. West Virginia’s grid is already fragile, built for a smaller, slower economy. Its rural counties need distributed resilience — microgrids, local generation, targeted modernization — not gigawatts of export capacity. Yet every dollar of new investment is being routed toward infrastructure designed to serve someone else’s load. A 233 percent capacity increase might look like ambition from the Capitol, but at the meter, it’s austerity.

The contrast with Georgia is structural, not moral. Georgia’s voters could register frustration; West Virginia’s can’t. One state converted affordability pressure into political feedback, the other into fiscal dependency. Both are responding to the same signal — rising load and falling patience — but only one has the means to course-correct before the next cycle of debt and dislocation begins.

And that cycle is already shifting from infrastructure to employment. The same impulse driving West Virginia’s buildout — secure cash flow in uncertain markets — is appearing inside corporate payrolls, where “Day One” efficiency replaces reform. Georgia’s voters felt the strain through their bills; West Virginia’s workers will feel it through their paychecks.

IV. The Converging Trends

It doesn’t matter whether you’re in Macon or Morgantown. The crisis comes everywhere because it rides on the same social contract: that wages will keep pace with the wires.

That contract is failing in slow motion.

The headlines call it a tech layoff cycle. It’s not. It’s Day One culture — the quiet prelude to automation. Companies are flattening org charts not because AI has replaced the worker, but because it soon will. Over 150,000 workers lost their jobs in October 2025 alone; roughly 33,000 in tech. Industry reports cite “AI adoption” or “efficiency initiatives” as the driving factor in approximately one-fifth of announced reductions.

That isn’t the robot uprising. It’s management performing preemptive surgery on its own payroll—a cultural rehearsal for automation, a cleansing of the middle before the machine arrives.

None of this is malicious; it’s mechanical.

And the grid? Most utilities still model steady wage growth in their load forecasts — typically 2-3 percent annually.

Utilities are issuing 30-year bonds to build out data-center infrastructure for firms that are simultaneously removing the income that pays for it. The same technology driving the new load is hollowing out the load-payer. Every dollar of “productivity” on Wall Street is an arrearage risk on Main Street.

The AI boom isn’t killing jobs directly; it’s killing the idea of latency. When models improve monthly, reorgs follow quarterly.

Markets move at inference speed; wages move at regulatory speed.

Five months ago, Load to Ruin warned of the “double-build trap” — that 30-year infrastructure financed by households would collide with five-year labor disruption cycles driven by AI. It’s happening faster than forecast. The system we’re modernizing for a digital future is being financed by a disappearing present.

Affordability isn’t a slogan anymore. It’s the denominator of civilization — and it’s cracking.

⸻

V. The Narrowing Window

Georgia fired the first shot, but the echo is everywhere.

Everywhere, the same math hums: 30-year debt, 5-year labor cycles, and fewer people earning the wages that keep the grid alive.

The new gospel isn’t replacing the worker — it’s never hiring them at all. “Day One” culture treats human payroll like spinning reserve: an inefficiency to be optimized. The middle is being hollowed out before the machines even arrive.

Crystal feels it. She still teaches in Macon, still budgets her paycheck, but the second job she once relied on — remote grading work — vanished when the company “streamlined with AI.” Her electric bill, once $126, is climbing toward $180 — the exact $11 transformer surcharge multiplied across six rate cases. Her side income is down 30 percent. That’s the loop: the same system billing her for modernization is deleting her work. She’s paying for her own obsolescence.

Across the country, families are living inside that paradox. The grid is expanding, but the payrolls that sustain it are contracting.

Like your 401(k)? Then you’d better hope the machine eats your job. If it doesn’t, the bubble bursts; if it does, the wage base collapses. Either way, the grid’s math fails before the bonds mature.

Georgia proved voters can still interrupt the feedback loop — for now. But interruption isn’t the same as reform. Each model release, each rate case, each merger tightens the circuit between energy and employment, until one outruns the other.

Crystal doesn’t read analyst notes. She just knows the envelope on her kitchen table feels heavier. And that’s where the affordability war is really being fought — in living rooms lit by the system that’s pricing them out.

Five months ago, Load to Ruin called this a hypothesis. Now it’s an observation.

The fuse is lit. The timer isn’t fixed.

Subscribe now

⸻

Endnotes & Sources

Part I: The Prediction Meets Reality

[1] Georgia PSC Election Results — “Georgia voters oust GOP utility commissioners over rising electricity rates,” Reuters, November 5, 2025, https://www.reuters.com/business/energy/georgia-voters-oust-gop-utility-commissioners-over-rising-electricity-rates-2025-11-05/; vote margins from “2025 Georgia Public Service Commission special election,” Wikipedia, accessed November 9, 2025, https://en.wikipedia.org/wiki/2025_Georgia_Public_Service_Commission_special_election (citing Georgia Secretary of State data).

[2] “Monthly Car Payments” Quote — Voter quoted in “Georgia voters oust GOP utility commissioners over rising electricity rates,” Reuters, November 5, 2025.

[3] Vogtle Cost Estimate — “Georgia Power customers to pay billions more for Plant Vogtle nuclear expansion,” Associated Press, August 27, 2024; “Georgia PSC staff warns Vogtle overruns saddle customers with $40B tab,” E&E News, July 2024.

[4] Rate Increases and Bill Growth — “Georgia Power customers see rates climb again as PSC approves fuel costs,” Atlanta Journal-Constitution, May 16, 2024; “Georgia Power rate hikes to continue as Vogtle costs mount,” Utility Dive, June 2024.

[5] Jefferies Downgrade — “Southern Co. stock rating downgraded by Jefferies on Georgia PSC election,” Investing.com, November 5, 2025, https://www.investing.com/news/analyst-ratings/southern-co-stock-rating-downgraded-by-jefferies-on-georgia-psc-election-93CH-4332972; “Southern Company Shares Fall After Jefferies Downgrade,” MarketScreener / MT Newswires, November 5, 2025, https://www.marketscreener.com/news/southern-company-shares-fall-after-jefferies-downgrade-ce7d5cdddb89f027.

[6] Vivek Ramaswamy Statement — “Ramaswamy launches Ohio governor bid, vows affordability focus,” WKYC Cleveland, November 6, 2025; see also X post, @VivekGRamaswamy, November 6, 2025, 8:14 AM ET.

Part II: The Georgia Signal in Context

[7] Georgia Power Load Projection — Georgia Power Company, 2025 Integrated Resource Plan (January 2025); “Georgia Power projects 8,200 MW of new load growth through 2030,” Utility Dive, June 2025, https://www.utilitydive.com/news/georgia-power-8200-mw-load-growth-2025-irp.

[8] Elected PSC Commissioners — Eleven states elect PSC commissioners: Alabama, Arizona, Georgia, Louisiana, Mississippi, Montana, Nebraska, New Mexico, North Dakota, Oklahoma, and South Dakota. Source: National Association of Regulatory Utility Commissioners (NARUC), 2024 State Commission Profiles,

https://www.naruc.org

.

Part III: Two Laboratories, Two Futures

[9] West Virginia 50 by 50 Plan — “Morrisey shares new energy plan for WV relying heavily on coal, natural gas to grow power creation,” West Virginia Watch, September 11, 2025, https://westvirginiawatch.com/2025/09/11/morrisey-shares-new-energy-plan-for-wv-relying-heavily-on-coal-natural-gas-to-grow-power-creation/; West Virginia population and median income data from U.S. Census Bureau, 2023 American Community Survey.

Part IV: The Converging Trends

[10] Tech Layoffs and AI Adoption — Layoff figures compiled from Challenger, Gray & Christmas, Inc., Monthly Job Cut Report, October 2025; Layoffs.fyi database, accessed November 2025. “AI adoption” and “efficiency initiatives” language from company announcements compiled by TechCrunch Layoff Tracker and Bloomberg Technology, October-November 2025.

Definitions

Electoral Elasticity — The threshold at which affordability pressure translates into electoral risk for regulators, creating measurable uncertainty for capital deployment and rate recovery. In states with elected commissions, this threshold is observable through polling and election outcomes; in states with appointed commissions, it manifests through legislative pressure, gubernatorial intervention, or credit rating adjustments.

Double-Build Trap — The phenomenon where households fund two simultaneous infrastructure expansions: massive generation and transmission systems for AI loads, plus local distribution upgrades for their own electrification needs (electric vehicles, heat pumps). Originally defined in Load to Ruin: The Infrastructure Paradox (June 2025).

⸻

Wesley Whited is Principal Consultant for VPP & Flexibility Strategy at DNV and author of Load to Ruin: The Infrastructure Paradox. The views expressed are his own.

TL;DR

AI data centers are forcing utilities to spend over $1 trillion on grid infrastructure through 2030, paid for by regular customers through higher bills, while simultaneously eliminating the middle-class jobs needed to afford those rising costs. This creates a dangerous 'double-build trap' that could collapse the traditional utility system unless regulators act quickly.

Curious about my work or personal interests?

You can explore my professional background on LinkedIn or delve into my life at the grid’s edge on my Carrd site.

New here?

Check out my Welcome Post to choose the reading format that works best for you.

Executive Summary

What You'll Learn

For Utility Regulators: How AI data centers are creating unprecedented cost-allocation challenges that threaten traditional rate recovery mechanisms, plus specific policy tools to restore equity before investment-grade utilities drift toward junk status.

For Utility Executives: Why the hyperscale boom represents both the largest growth opportunity and greatest financial risk in decades, with practical strategies to avoid stranded asset exposure while managing unprecedented demand.

For Energy Policy Advocates: How current infrastructure investments are creating a regressive wealth transfer from households to hyperscalers, and the regulatory interventions needed to prevent a two-tier energy system.

For Community Leaders: Why local electric bills are rising to fund AI infrastructure that may eliminate local jobs, and what policy changes could protect household affordability during the employment transition.

The Crisis

America's electric grid faces an unprecedented crisis: AI data centers are triggering over $1 trillion in infrastructure investments just as artificial intelligence eliminates the middle-class jobs needed to pay for them. This represents a fundamental break from the century-old utility business model, where energy-intensive industries created local employment that sustained the communities, funding grid expansion.

Hyperscale data centers now consume electricity equivalent to that of entire cities, while employing fewer than 50 people each. A single AI campus draws as much power as 50,000 homes but generates virtually no local payroll. Meanwhile, the algorithms these facilities host are systematically eliminating cognitive work—from legal research to financial analysis—that once supported household incomes between $65,000-$150,000 annually.

The result is a "double-build trap" where households fund two simultaneous infrastructure expansions: massive generation and transmission systems for AI loads, plus local distribution upgrades for their own electrification needs. In Virginia, average monthly electric bills are projected to rise from $142 to $315 by 2039, primarily driven by the expansion of data center infrastructure. In Georgia, the demand for AI has forced utilities to abandon their coal retirement plans and extend fossil fuel operations indefinitely.

The Timing Mismatch

This timing mismatch poses a threat to the utility industry's financial stability. Infrastructure costs are typically recovered over 20-30 years, but AI-driven employment disruptions are already accelerating. National utility debt has exploded from $12 billion to $21.1 billion since 2020, with nearly 21 million households now carrying overdue balances. As hyperscalers retreat to private power systems, they leave public ratepayers financing potentially stranded assets while serving an increasingly impoverished customer base.

The Solution Framework

The crisis demands immediate regulatory intervention, including the establishment of Utility Arrearage Elasticity metrics to track unemployment-to-default relationships, mandatory fair-share infrastructure contributions from hyperscalers, and treating affordability as a statutory priority equal to reliability. Without these reforms, the traditional utility system faces a death spiral where investment-grade companies drift toward junk status, serving customers whose AI-disrupted incomes cannot sustain the infrastructure debt that eliminated their careers.

The Choice

The choice is binary: engineer equity into the next-generation grid now, or inherit a collapsing two-tier system where private power islands serve profitable AI loads while an over-leveraged public network struggles to serve increasingly impoverished households. We are financing tomorrow's stranded assets with today's disappearing paychecks—a 30-year mortgage on infrastructure that eliminates the jobs needed to pay for it within the first decade. Chapter 1: Generation & Interconnection

Chapter 1: Generation & Interconnection

A Tsunami of Load

For two decades, America's electricity system operated in a state of predictable equilibrium. Demand growth held steady at roughly 1% annually—a measured pace that allowed utilities to expand their infrastructure methodically and distribute investments across manageable timelines (1). This era of stability abruptly ended in late 2022. The North American Electric Reliability Corporation (NERC), the nonprofit guardian of grid stability, now warns of "explosive" demand growth driven by hyperscale data centers—a surge so rapid it threatens to outrun the deliberate processes of generation and transmission planning that have long anchored the industry's approach to resource adequacy (2). The scale is staggering: 2025 alone will witness an additional 83 TWh of hyperscale demand, representing a 14% surge over 2024 and equivalent to powering 7.7 million American homes (3).

This demand surge is not abstract. Lawrence Berkeley National Laboratory (LBNL) reports that data centers used about 2% of U.S. electricity in 2020—a share expected to triple to 6% by 2025 (4). The Electric Power Research Institute projects that AI workloads will draw 9–12% of all U.S. electricity by 2030 (5). Bain & Company estimates that these loads add roughly 1% to customer bills every year through 2032 (6).

The surge in demand shows no signs of slowing, and it's a load that virtually no one was planning for just a few years ago. The Department of Energy projects that by 2035, we'll need more than five times the current capacity to move electricity between regions compared to what we had in 2020 (7). LBNL forecasts that this growth will accelerate dramatically, with demand increasing by 13% to 27% annually through 2028 (8).

The Rise of Private Power Islands

Hyperscale data centers—facilities with 5,000+ servers and over 10,000 square feet—serve giants like Google, Amazon, Microsoft, Meta, and Apple (9). From a grid operator's perspective, these facilities are uniquely challenging: you can't ask them to use less power during peak times, you can't cut them off during emergencies, and they'll continue to consume electricity regardless of how expensive it becomes.

Utility interconnection queues have become the bottleneck choking AI expansion. What once took 12-18 months now stretches 3-5 years, with some projects facing even longer delays as grid studies reveal cascading upgrade requirements. For hyperscalers racing to deploy AI infrastructure, these timelines are commercially unacceptable.

The capital response has been staggering. McKinsey projects that companies will need to invest $5.2 trillion in AI data center infrastructure by 2030, with speed to market becoming the overriding imperative (10). When AI training windows represent billions in potential revenue that competitors might capture first, cost becomes secondary to deployment velocity.

This has unleashed a wave of infrastructure circumvention. Companies are paying premiums to acquire "shovel-ready" sites with existing electrical infrastructure, purchasing retired industrial facilities for their grandfathered interconnection rights, and pre-building substations and transmission lines on speculation. The most aggressive players are constructing entire parallel energy systems—dedicated generation, private transmission, and behind-the-meter storage—designed to operate independently of utility timelines and approval processes.

What began as temporary workarounds has crystallized into a permanent strategy: build first, interconnect later, and bypass traditional utility planning entirely when possible. Yet this creates a troubling dynamic—hyperscalers leverage public infrastructure to establish their footprint, trigger costly grid upgrades, then pivot to private systems that could leave public ratepayers holding the bill.

The Public Still Pays

Now consider Bob and Linda, retirees in Loudoun County living on a fixed income of $3,000 per month. Dominion projects that their electricity bill will rise from $142 to $315 by 2039—an increase primarily driven by data center infrastructure they neither use nor benefit from. They are not alone (11).

For most of the past century, the U.S. grid operated under a simple social contract: shared infrastructure, shared cost, and shared benefit. That contract is fraying. As hyperscale customers exit the traditional model—forming private energy enclaves—the cost burden is socialized while the benefits are privatized.

The result is a grid where public customers pay for infrastructure stress they didn’t cause, and where the largest, fastest-growing loads no longer participate in the very system, they’re transforming.

Pressure Points & Pushback

Virginia: The Epicenter of Excess

Northern Virginia’s “Data Center Alley” now hosts more than 35 % of global hyperscale computing capacity, according to the Virginia Economic Development Partnership (12). Dominion Energy Virginia—the utility that serves Loudoun County, the world’s largest data center market—faces infrastructure demands without precedent. Its 2024 Integrated Resource Plan (IRP) projects electricity demand to grow 5.5 % per year through 2035, with peak load doubling by 2039—growth driven almost entirely by data centers (11).

Excluding data centers, Dominion's annual peak demand growth is only 0.5%. Instead, the utility confronts “the largest growth in power demand since the years immediately following World War II.” Meeting that need will require 21 GW of new clean generation, 5.9 GW of natural gas capacity, and extensive transmission and distribution upgrades.

The price tag is staggering: $78 billion to $103 billion over the next 15 years. For Virginia households, like Bob and Linda, whose average monthly electric bill is $142, that translates into $214–$315 by 2039.

Serving Dominion’s projected 26.9 GW of new capacity could yield over 200 TWh of annual generation, assuming high utilization typical of AI workloads. At a blended levelized cost of energy (LCOE) of $65–$85/MWh, this equates to $13–$17 billion per year in electricity costs (13). These are structural additions to the rate base; without reform, households and small businesses will continue to fund infrastructure benefiting only the world’s most profitable firms.

Meanwhile, Virginia’s hyperscalers seek behind-the-meter solutions to prioritize faster connections (14). The irony is stark: ordinary customers will pay for infrastructure that large users may ultimately bypass or abandon.

Texas: The Queue Explosion

Texas faces even greater uncertainty than Virginia. In 2024, ERCOT approved 5,496 MW of new hyperscale load, while more than 99,000 MW—nearly equal to the state’s current peak demand—remains in the interconnection queue awaiting study. Planning systems built for incremental growth are overwhelmed (15).

Utilities must choose between costly speculative infrastructure or risk losing digital investments to other locations. The unprecedented queue forces a stark choice: gamble on demand that may not materialize or risk saddling smaller ratepayers with the cost of unused assets. Credit agencies are beginning to notice: utility bond spreads have widened as investors question whether rate bases built on speculative AI demand can maintain investment-grade stability.

The Ohio Pushback

Columbus has emerged as an unexpected hub for hyperscale data centers. Firms such as Vantage have announced more than $2 billion in digital infrastructure projects, and the region’s data center load has surged from 100 MW in 2020 to 600 MW in 2024 (16). Projections point to 5 GW by 2030, meaning central Ohio’s total electricity demand will more than double within six years.

AEP, previously focused on coal retirements, now builds Silicon Valley-scale infrastructure in the Rust Belt. Aware of the risk, Ohio regulators introduced a cost-allocation reform in October 2024: any new data center larger than 25 MW must pay for at least 85% of its contracted monthly energy, even if actual usage is lower. This “minimum-take” rule tackles a core inequity: data centers ramp up slowly, leaving ratepayers to finance unused capacity for years (17).

AEP Ohio’s $95 million Green Chapel Station—a 500 MW substation sized to power 500,000 homes—now sits largely idle after Intel postponed its $100 billion semiconductor plant from 2027-28 to 2030-32. Intel and AEP have reached a payment arrangement, yet much of the cost still falls on customers. Stranded infrastructure can result when large industrial loads prompt utility investment, only to delay or abandon planned consumption (18).

Whether for AI training or chip fabrication, the result is the same: utilities invest billions anticipating demand that may arrive late, or never. Ohio's minimum-take rule offers one solution to this problem within state borders, but it can't address a larger inequity: the way hyperscale growth in one state forces neighboring states to pay for the infrastructure needed to support it.

The Interstate Subsidy Machine

While Ohio’s minimum-take rule limits internal cost shifts, hyperscale growth creates cross-state subsidies that no single regulator can control. Regional transmission organizations, such as PJM Interconnection, allocate infrastructure costs across multiple states using formulas designed for gradual, widely distributed growth, not the concentrated, hyper-scale buildouts now underway. As a result, states that court data centers with aggressive incentives can offload a significant portion of the associated electric grid expenses onto their neighbors.

A recent dispute highlights the stakes: on April 8, 2024, the Federal Energy Regulatory Commission (FERC) approved PJM’s cost-allocation plan for approximately $5.1 billion in transmission upgrades. Maryland ratepayers were assigned nearly $800 million of that total, even though the projects chiefly support Virginia’s data center build-out and could become stranded if those facilities migrate behind the meter (19). Maryland People’s Counsel David Lapp called the outcome “fundamentally unfair,” arguing that “Virginia—not other states—should pay for transmission projects driven by its data-center incentive policy.”These cross-subsidies distort regional economics, allowing one state to benefit while shifting grid costs onto neighbors (20).

A Parallel Grid Emerges

Hyperscalers strain public grids with unprecedented demand while simultaneously building independent power networks. What began as a workaround for lengthy interconnection delays is evolving into a permanent strategy that positions these firms to compete with, rather than complement, traditional utilities.

The scale is impressive. Crusoe Energy has secured 4.5 GW of dedicated natural gas generation to serve AI data centers, entirely bypassing the interconnection queue—a capacity roughly equal to New Hampshire's entire power fleet yet unbound by utility oversight or cost-allocation rules (22). Microsoft is investing $80 billion in renewables, paired with dedicated transmission, and Meta is constructing Arizona's largest independent solar-plus-storage project, capable of islanded operation during grid outages (23).

Three imperatives drive this shift: utility planning cycles cannot keep pace with AI's breakneck expansion, AI workloads demand reliability that public grids cannot always provide, and at scale, self-generation is often cheaper than grid power. These companies are no longer just voracious power consumers; they are becoming quasi-utilities that control generation, transmission, and load within self-contained private power islands.

Federal demand amplifies the trend. The U.S. government plans to spend $8.3 billion on cloud services in FY 2025, with civilian agencies accounting for $4.5 billion, according to the Federal IT Dashboard (24). Defense and intelligence agencies are awarding multibillion-dollar contracts—including the Department of Defense's $9 billion JWCC agreement—to Amazon, Microsoft, and other hyperscalers. As these platforms fold dedicated generation into their networks, they are becoming vertically integrated digital utilities, central to both U.S. cloud infrastructure and national energy strategy. This dual role complicates regulatory oversight and may insulate these companies from aggressive cost-allocation reforms (25).

The dynamic creates a perverse cycle: hyperscalers use the grid to establish a footprint, trigger major upgrades, while also positioning to pivot to private generation. These moves deepen utility risk: public investment in transmission is greenlit under assumptions of long-term service, but when hyperscalers go private, those assets remain stranded and ratepayer-financed.

Strategic Pivot: Amazon and the Nuclear Hedge

Amazon’s nuclear strategy highlights the growing regulatory pressure on hyperscalers’ power procurement. The company initially sought a behind-the-meter deal with Talen Energy’s Susquehanna plant that would have delivered power directly, avoiding standard transmission charges and effectively subsidizing Amazon at the expense of other customers. In November 2024, FERC rejected the arrangement, citing concerns that it would shift grid costs onto smaller ratepayers (26).

Faced with that ruling, Amazon recalibrated its approach. In June 2025, it expanded its partnership with Talen, securing 1.9 GW of carbon-free nuclear capacity through a conventional, front-of-the-meter contract that includes full transmission fees. The move is as much damage control as strategy: Amazon preserves its carbon-free power ambitions while signaling regulatory compliance, even as FERC continues to scrutinize hyperscaler procurement tactics. The episode demonstrates that consistent regulatory pressure can shape hyperscaler behavior if applied firmly (27).

The Decisions Begin Now

The hyperscale boom is the most disruptive force in U.S. electricity since the advent of rural electrification. The challenge is institutional as much as financial: growth is outpacing the regulatory frameworks built for the 20th century. Yet regulators still have potent tools—updated cost-allocation rules, hyperscaler-specific tariffs, and mandatory infrastructure contributions—to restore equity. Those choices must be made now.

This chapter focused on generation and transmission, but the real burden for households will be felt at the distribution level, where aging neighborhood circuits must be rebuilt to serve data-center loads. Chapter 2 explores pressure in detail.

Chapter 2: The Distribution Squeeze

Where Costs Hit Home

Consider Crystal, a substitute teacher in Macon, Georgia, who already stretches every paycheck to cover rent and groceries. When two neighbors bought electric cars this spring, the 25-year-old transformer on her street began to hum at night and trip on humid afternoons. The utility's remedy—a $10,000 pad-mount replacement—will be folded into the following rate case, adding roughly $11 to her $126 monthly bill.

Meanwhile, an AI data center campus seventy miles away now draws more power than downtown Atlanta, prompting Georgia Power to delay coal retirements and double down on gas peaker plants. She is being asked to finance both ends of the grid: the neighborhood transformer that keeps her lights on and the bulk-system build-out that keeps GPUs spinning.

Chapter 1 covered generation pressures, but they're only half the issue. The same hyperscale demand that strains bulk power systems is also crushing the local wires and transformers that serve ordinary households. What was intended as a managed transition to beneficial electrification has become a scramble where ratepayers fund a double build-out (28).

Utilities must expand bulk generation for AI data centers while simultaneously rebuilding distribution networks for the electric vehicles and heat pumps that actually decarbonize homes. The result: households pay twice—once for hyperscale infrastructure that may crowd out their clean-energy goals, and again for the neighborhood upgrades still needed to achieve them.

The National Distribution Gap

Distribution networks face an impossible squeeze: the same local circuits meant to handle millions of new EV chargers and heat pumps must also feed power-hungry AI data centers. It's a direct clash at the neighborhood level, with stark math behind it.

Electric vehicle adoption alone is accelerating as fast as the grid can adapt. The U.S. had 4.4 million registered EVs in 2024 and is projected to have 40 million by 2030, according to the IEA (29). Each home charger draws about 7 kW—nearly matching an entire household's 8 kW peak demand. Meanwhile, NREL estimates that distribution-transformer capacity must rise by 160-260% by 2050 just to meet electrification across all sectors, a bill running into hundreds of billions (30).

The breaking point is immediate. A typical neighborhood transformer operates at 64% capacity while serving six homes. Adding just two Level 2 EV chargers exceeds capacity, risking failure. As EV adoption clusters in neighborhoods, many circuits will hit thermal limits years before utilities can upgrade them.

But AI demand has already claimed much of the grid capacity earmarked for electrification. Climate-focused expansion plans have morphed into emergency upgrades serving both data centers and homes simultaneously. The result: households fund a dual expansion that may actually slow their own clean-energy transition.

State Spotlights

Georgia: When Forecasts Collapse

An Integrated Resource Plan is a utility's long-term playbook—typically spanning 15-20 years and approved by state regulators—for how it will meet growing electricity demand through the development of new power plants, grid upgrades, and efficiency programs. Georgia Power's 2022 IRP envisioned a gradual increase: peak demand rising 0.7% annually, with approximately 200 MW of new load added each year. That incremental growth could be covered by routine substation upgrades and orderly coal retirements. Then AI arrived and shattered every assumption (31).

The utility's January 2025 IRP filing projects 8,200 MW of load growth over six years—27 times what it had forecast for the entire decade. To grasp the scale: Georgia now needs nearly four Plant Vogtle-sized nuclear reactors, or 16,400 MW of gas turbines, just to meet this surge. The utility reports a pipeline of potential customers approaching 23 GW, with individual data centers consuming more electricity than entire metropolitan areas.

This demand explosion forces a devastating policy reversal. Georgia Power now forecasts a winter shortfall of 17-18 GW by 2030—twice previous estimates. Coal units scheduled for retirement must now operate indefinitely. The utility acknowledges it will "lean on fossil fuel sources like gas-fired power plants" because AI consumed the grid headroom meant for home electrification (32).

The clean energy transition has been inverted: what began as a decarbonization roadmap has become a fossil fuel expansion plan, with artificial intelligence consuming the capacity meant for genuine emission reductions.

California: The PG&E Distribution Challenge

Pacific Gas & Electric faces a similar squeeze in Northern California. Meeting EV and building electrification demand will require upgrading 1,679 feeders and 17 substations by 2030, with distribution investments climbing to $5 billion by 2050, according to 2023 filings (33).

But data center demand is accelerating even faster. PG&E's pipeline expanded to 8.7 GW over the next decade—a 60% jump from 5.5 GW at the end of 2024. The utility has 18 projects in final engineering (1.4 GW) plus 21 new applications from this spring alone (4.1 GW).

PG&E’s May 2025 Investor release pitches this as good news, arguing that every 1 GW of data center demand could cut monthly bills by 1-2% by spreading grid costs across more customers. Yet the same filing reveals a dangerous sequencing problem: the utility must first fast-track high-capacity infrastructure for data centers, then retrofit neighborhood circuits for EVs and heat pumps later, risking "building certain parts of the grid twice."

The numbers tell the real story. PG&E's 2027 rate case requests $10.9 billion in capital spending, raising typical household bills by $11 monthly. "Growth & new-business" distribution work alone tops $4 billion over five years—dwarfing residential electrification programs and signaling that data centers, not households, are driving investment priorities (34).

Arizona: The Reliability Dilemma

Arizona Public Service (APS) faces an acute planning crisis, epitomizing the grid's escalating pressures. In November 2024, APS Senior Vice-President José Esparza revealed the utility had nearly 10 GW of pending data-center interconnection requests—a staggering figure that already surpasses its entire 2024 summer peak demand of 8.2 GW. Accepting this full queue, Esparza cautioned, "would put existing customers at risk of poor reliability" (35).

The core issue is a stark timing mismatch. While a U.S. Department of Energy review indicates new high-voltage transmission lines require roughly 10 years to permit and build, hyperscale developers often expect service within 18–24 months. This forces APS into an unenviable choice: prioritize rapid data center deployment or safeguard the reliable service of today's households and businesses (36).

This trade-off is reflected in capital plans. Esparza informed regulators that APS anticipates investing approximately $2 billion annually "for the foreseeable future," largely into substations and high-capacity feeders dedicated to data centers. At this rate, spending could exceed $15 billion by 2035. While APS has proposed a special tariff to shift these data center-specific costs from households to hyperscalers—a plan currently under review by the Arizona Corporation Commission—the immediate challenge of balancing demand against capacity remains.

Reliability concerns are also reshaping Arizona's energy mix. APS's projections show peak demand soaring from 8.2 GW in 2024 to approximately 11.35 GW by 2027—a 3.1 GW leap in just three summers. To bridge this gap, APS plans to add about 3.5 GW of renewables by 2031, alongside new gas-fired peaker plants. While this strategy aims to preserve reliability, it simultaneously slows the pace of decarbonization. Executives acknowledge that, absent significant federal aid or aggressive demand-side management, Arizona may be forced to choose between its climate goals and AI-driven growth; achieving both on the current timeline appears increasingly impossible (37).

Supply-Chain Bottleneck: The Physical Choke Point

Beyond infrastructure capacity, the grid faces a fundamental physical choke point: distribution transformers. This essential hardware is creating a hard stop; unless resolved, AI expansion will stall regardless of how much new generation is built.

The problem starts with aging equipment. The U.S. operates 60–80 million distribution transformers, over half of which are more than 33 years old and nearing the end of their design life, according to NREL's 2024 study. Simply replacing these units, combined with meeting new demand from data centers, EVs, and heat pumps, will require 1.5–2.4 million new transformers annually through 2050 (38).

But manufacturing can't keep pace. Evidence from DOE's transformer-rulemaking docket reveals severe constraints. Domestic plants, though faster, are limited, while import-dependent supply chains face longer lead times due to overseas sourcing and customs delays (39). The 2024 NIAC report on transformer shortages cites median deliveries of 32–64 weeks for pad-mount units – a drastic increase from historical norms, forcing utilities to stockpile inventory and delay projects (40).

The severity of this market failure prompted extraordinary federal intervention. In June 2022, President Biden invoked the Defense Production Act to accelerate domestic manufacturing of transformers and grid components. This allows DOE to issue grants and loan guarantees for new steel lines and assembly plants – a clear signal of the market's inability to simultaneously meet demands from electrification and hyperscale growth (41).

Utilities are feeling the direct pressure. Georgia Power's January 2025 IRP explicitly warns that serving its massive new load queue will necessitate "significant incremental orders of distribution transformers." With current lead times exceeding one year for key sizes, the company is "pre-purchasing critical inventory" to avoid delaying data-center connections, a cost ultimately borne by ratepayers (31). This pre-purchasing strategy epitomizes the broader affordability crisis: utilities stockpile equipment for loads that may never materialize, while ratepayers guarantee recovery regardless of whether the demand justifying these investments ever arrives.

Double-Build Arithmetic: The Ratepayer's Burden

Today's grid crisis centers on a brutal "double-build trap." Households must fund two simultaneous infrastructure expansions: local distribution upgrades for their own electrification—such as EVs, heat pumps, and rooftop solar—and massive system-wide buildouts serving the demand of hyperscale AI.

Georgia Power's 2025 integrated resource plan reveals the arithmetic: data centers will account for 80% of new load, requiring nearly 9 GW of additional capacity. State regulators have warned of overbuild risks, cautioning that if these AI projects "fail to materialize, all of that additional cost would need to be paid" by ratepayers (42).

The timing is perverse. AI's economic disruption is shrinking the household incomes needed to recover these investments, while utilities pivot to new gas plants—not to advance decarbonization, but because AI has devoured the grid capacity once reserved for clean residential electrification.

Ratepayers face a stark paradox: permanent rate increases to finance the very technologies destroying their ability to pay those rates.

A Call for Fair-Share Policy

Crystal, the substitute teacher in Macon epitomizes this unfair burden. She will never set foot in the GPU hall 70 miles away, yet she pays for it twice: first through higher base rates funding the new generation, then through distribution surcharges for her aging neighborhood transformer. Regulators, however, possess critical tools to mitigate this impact: imposing minimum-take tariffs on hyperscalers, fast-tracking managed-charging programs, and targeting federal transformer grants to the hardest-hit feeders. Early, decisive action could keep her bill near the low end of projections. Delay, however, risks leaving her funding stranded assets she never needed, while the AI campus quietly slips behind its own meter, its full costs not transparently borne.

Chapter 3 explores an even sharper, more personal irony: The very AI systems driving her rising electricity bill—the ones she's being forced to finance—could soon threaten her livelihood.

Chapter 3: The Wage-less Load Crisis

When Income Shrinks and Bills Rise

Maria is a software developer in Atlanta, earning $85,000. Her $180 electric bill was routine—until Georgia Power’s 15 % hike to fund data-center expansion added another $27. Then Salesforce froze hiring and “rightsized” her team after AI tools lifted productivity by about 30 %. CEO Marc Benioff’s parting words: “We are the last generation to manage only humans.”

Unable to find work in a tightening tech market, Maria now clocks 18 hours a week at a nearby McDonald’s, where AI kiosks take most orders and scheduling algorithms cut her shifts. Her monthly income has fallen from $7,100 to roughly $2,400, yet her power bill continues to rise to cover the very AI infrastructure that ended her career.

Maria’s story reveals a fatal contradiction. As AI accelerates electricity demand, utilities raise rates to build new generation and modernize their distribution systems. At the same time, those same AI systems erode household wages, stripping customers of the means to pay the rising bills.

The numbers echo her experience. Georgia Power projects a 27-fold jump in AI-driven load by 2030. The month the forecast was filed, Salesforce halted software engineer hiring, citing 30% productivity gains from AI (43). Three months later, McDonald’s announced plans to equip its 43,000 restaurants with AI-powered kitchens and drive-throughs (44). AI thus forces a double build-out of the grid while undercutting the incomes that finance it.

The Broken Promise

For more than a century, America’s electric grid rested on a straightforward bargain: more power meant more prosperity. When the Tennessee Valley Authority electrified aluminum smelters in the 1940s, those energy-hungry plants created thousands of jobs. Workers’ paychecks covered rising rates, and that revenue financed still more wires and generators. Steel mills, auto factories, and chemical refineries repeated the pattern—massive loads that paid their own way precisely because they put people to work, forming a virtuous cycle in which the people funding grid expansions directly shared in the benefits.

AI severs that bargain. A single hyperscale data center now consumes as much electricity as 50,000 homes, yet employs fewer than 50 people. It extracts grid value while contributing almost nothing to local payrolls—and the algorithms it hosts are already erasing jobs across entire service territories, from call-center agents to long-haul drivers.

The timing is unforgiving. Infrastructure financed between 2024 and 2030 will be recovered over 20–30 years, but the entry-level hiring freezes spreading through the tech sector today are the early tremors of structural unemployment that could solidify within the next decade—well before those 30-year grid costs are paid off. Utilities will spend three decades trying to collect on assets whose very purpose helped wipe out the wages needed to fund them—much like taking a 30-year mortgage on a salary that disappears after year five.

When Smart Machines Target Smart Jobs

Past automation waves offered escape routes. When factories closed, workers could retrain for office jobs. However, today's AI revolution targets cognitive work directly, including legal assistants, financial analysts, content creators, and customer-service representatives. AI is climbing the economic ladder, eliminating precisely the middle-class jobs that once supported steady rate increases.

New roles will eventually emerge, but the transition won't be seamless. Recall the #learntocode refrain from the 2010s and ask how many coal miners actually made it to tech accelerators. The timing and skills gaps are real—most displaced workers won't vault overnight from processing TPS reports to engineering AI prompts. But their utility bills will continue arriving every month.

The cruel irony? AI excels first at tasks paying $65,000-$150,000 annually, precisely the income bracket capable of absorbing 15-20% rate hikes. Leading AI models now score 88-90% on tests covering 57 academic subjects, outperforming most college graduates (45). GPT-4 scored in the 88th percentile on the bar exam in 2023, surpassing most law school graduates (46). Similar dominance appears in medical licensing, CPA certification, and STEM assessments.

These aren't party tricks—they're job elimination roadmaps. ChatGPT-5, expected mid-2025, will likely match or exceed average human expertise across an even wider range of tasks. We're approaching a post-labor economy where the marginal value of cognitive work approaches zero. Goldman Sachs forecasts that automation could eliminate 300 million jobs globally (47). The World Economic Forum expects 92 million positions—22% of today's workforce—to disappear by 2030 (48). These aren't distant projections—they represent the systematic erosion of the very income base that must service three decades of AI infrastructure debt, creating a structural mismatch between cost recovery timelines and employment disruption.

The Crisis Is Already Here—and Deepening Fast

The energy affordability crisis isn't looming on the horizon; it's already ravaging American households today. NEADA data reveals a staggering reality: national utility debt has exploded from approximately $12 billion before the pandemic to $21.1 billion by September 2024—a devastating 76% surge. Nearly one in six U.S. households, representing roughly 21 million families, are drowning in overdue energy bills (49). And this catastrophe unfolds before the next wave of rate increases or the full arrival of enterprise artificial intelligence, reshaping the workforce.

The state-by-state devastation tells an equally grim story. In New York, 1.3 million households struggle under $1.8 billion in debt—a dramatic escalation from 929,000 customers owing $742 million in 2019 (50). Massachusetts presents a similarly dire picture, with 825,000 households carrying $794 million in unpaid balances (51).

The COVID Dress Rehearsal

History offers a chilling preview of what lies ahead. When economically stressed customers cannot pay their bills, the consequences cascade rapidly. As unemployment rocketed to 14.8% in April 2020, utility arrearages ballooned from $12 billion to $32 billion by year's end (52). Shut-off moratoria provided temporary relief while unpaid balances accumulated like storm clouds. Regulators allowed utilities to classify this debt as "regulatory assets," guaranteeing recovery—with compounding interest—through inevitable future rate increases.

Yet COVID-19 represented a temporary disruption measured in months. The approaching AI-driven employment transformation will prove structural and enduring, systematically eroding household incomes far longer than any moratorium could possibly sustain. That spike shows why the industry needs a simple yard-stick—call it Utility Arrearage Elasticity (UAE)—to track how every uptick in unemployment cascades into unpaid bills

Some may point to COVID's managed response as proof that policy solutions exist, but this misses the critical difference in scale and permanence. Just as utilities file for record-breaking rate increases to power hyperscale data centers, the federal programs that once cushioned vulnerable customers are being dismantled. Chief among them is LIHEAP—the Low-Income Home Energy Assistance Program—which, until recently, provided heating and cooling support to approximately 5.9 million households burdened by energy costs consuming 8.6% to 17% of their income. The FY 2026 federal budget proposes eliminating LIHEAP entirely, and in April 2025, the program’s entire administrative staff was laid off—abandoning millions of Americans to navigate energy poverty without their last line of defense (53).

The perfect storm approaches: soaring energy costs driven by AI infrastructure demands, structural unemployment from automation, and the deliberate destruction of assistance programs. What we witness today is merely the tremor before the earthquake.

The Business Model Breakdown

Residential customers are a utility’s financial backbone. Their steady, non-negotiated payments anchor investment-grade credit and fund long-lived assets. When that foundation cracks, the whole structure wobbles.

Household electricity bills have jumped nearly 30 % since 2021, while natural-gas costs are up ≈ 40 % since 2019—roughly twice the rise in overall consumer-price inflation. Hardship is spreading: PowerLines estimates that about 80 million Americans now struggle to pay utility bills (54). A Smart Energy Consumer Collaborative survey puts the figure at 31% of adults—up from 25% in 2023—with nearly half of them earning more than $50,000. The same crisis, viewed through two lenses (55).

Credit agencies see a double threat. Moody’s warns that utilities are piling on debt to meet data-center demand even as they fear overbuilding if that load never materialises. Ten-year utility-bond spreads have widened roughly 60 basis points over Treasuries since early 2023, and retail prices keep climbing—an unusual trio that signals stress in the business model (56).

The irony would be comic if the stakes weren’t so high: companies are borrowing billions to power AI systems poised to erode the very paycheques customers need to cover those higher rates. Unless regulators rethink who pays for the next grid, an investment-grade industry could drift toward junk status in real time.

Conclusion: Affordability Is the New Reliability

Neither regulators nor utilities currently maintain formal indices linking unemployment to residential arrearages. This blind spot made sense when job losses were cyclical and brief, with retail collections reliably bouncing back before arrears could threaten balance-sheet stability. The post-labor AI transition completely rewrites this equation. When middle-income households face structural under- or unemployment, arrearages evolve beyond a "low-income program" concern, hollowing out the very customer class underpinning utility credit ratings.

The Missing Early-Warning System

COVID-19 revealed the severity of the feedback loop: unemployment spiking to 14.8% led to an explosion in national utility debt, which rose from approximately $12 billion to $32 billion within months. Yet no agency translated this shock into durable planning parameters. Establishing a Utility Arrearage Elasticity (UAE) metric—measuring percentage change in arrears per percentage-point change in unemployment—would provide commissions and rating agencies with a forward indicator as fundamental as reserve-margin targets. Without this yardstick, resource planners continue treating "ability to pay" as someone else's responsibility. The COVID debt explosion was a temporary shock; the AI employment transition represents permanent structural change requiring fundamentally different regulatory tools.

Two Destructive Feedback Loops

These loops are the financial mirror of the double-build trap detailed in Chapter 2: each turn of the cycle piles new costs on households already underwriting both ends of the grid.

Rate-basing bad debt creates a vicious cycle. When delinquencies surge, regulators typically permit utilities to securitize or rate-base shortfalls. Each recovery tranche pushes average rates higher, forcing more households into arrears and triggering additional financing petitions. Left unchecked, this mirrors the classic load-defection death spiral—driven by non-payment rather than rooftop solar.

Capital-flight contagion offers no better alternative. If commissions deny full recovery, utilities face dividend cuts and potential downgrades. Higher debt costs inflate revenue requirements, ultimately feeding back into rates regardless. Whether socializing bad debt or starving equity returns, both pathways drive weighted-average capital costs toward punitive levels precisely when the grid requires record investment for AI loads and climate resilience.

Private Power Islands: The Final Pressure Point

Hyperscale companies, insulated within private microgrids, observe this affordability crisis from behind their firewalls. Their departure erodes volumetric sales exactly when fixed-cost recovery becomes most challenging, amplifying the spiral while transferring additional cost risk onto shrinking residential demand.

Essential Actions

Integrate UAE stress testing into every Integrated Resource Plan. Planners must model scenarios where sustained unemployment rates of 7-10% coincide with AI-driven load growth, quantifying default risk alongside traditional reserve margins.

Mandate hyperscaler fair-share tariffs and minimum-take provisions, ensuring that the most significant new loads underwrite the capacity they trigger, whether remaining grid-connected or transitioning to private supply.

Establish credit-watch thresholds tied to arrearage data. State commissions and rating agencies should establish agreed trigger points—such as when arrears exceed 5% of annual retail revenue—beyond which new capital projects require affordability mitigation plans.

The Bottom Line

Whether it’s Bob and Linda in Loudoun or Crystal in Macon, the math is stark and unforgiving. Utilities are borrowing billions to power AI systems that simultaneously erode the paychecks customers need to service that debt. Maria, the displaced software developer now working part-time at McDonald's, represents millions who will face this impossible equation: monthly bills rising to fund the very infrastructure that eliminated their careers.

Affordability must achieve the same statutory priority as reliability. The current trajectory leads to systemic failure—investment-grade utilities drifting toward junk status as their customer base loses the capacity to pay for AI-driven infrastructure expansion. Until regulators systematically track and price the employment-arrearage relationship, today's trillion-dollar build-out risks becoming tomorrow's stranded debt, financed by households whose AI-disrupted incomes cannot sustain the burden.

The choice is binary: engineer equity into the next-generation grid now, or inherit a collapsing two-tier system where hyperscalers retreat to private power islands. We are financing tomorrow's stranded assets with today's disappearing paychecks—a 30-year mortgage on infrastructure that eliminates the jobs needed to pay for it within the first decade. Half-measures and delayed action will not suffice when the timeline for infrastructure cost recovery extends decades beyond the employment disruption already underway.

Work Cited

1.) U.S. Energy Information Administration. (2024, May). Table 2.15: Average retail price of electricity to ultimate customers [Data set]. U.S. Department of Energy. https://www.eia.gov/totalenergy/data/monthly/pdf/sec2_15.pdf

2.) North American Electric Reliability Corporation. (2024). 2024 Long-Term Reliability Assessment. https://www.nerc.com/pa/RAPA/ra/Reliability%20Assessments%20DL/NERC_Long%20Term%20Reliability%20Assessment_2024.pdf

3.) Bloom Energy. (2025). 2025 Data Center Power Report. https://resources.bloomenergy.com/data-center-power-report

4.)Liu, Z. (2023). Powering intelligence: Electricity demand and carbon implications of generative AI. University of California. https://escholarship.org/uc/item/32d6m0d1

5.) Electric Power Research Institute. (2023). Powering intelligence: Understanding the energy implications of AI and data center growth (Product ID 000000003002028905)

6.) Maloney, P. (2024, April 8). Data center load growth could boost US electricity bills 1% annually through 2032: Bain. Utility Dive. https://www.utilitydive.com/news/data-center-load-growth-us-electricity-bills-bain/730691/

7.) U.S. Department of Energy. (2023). National transmission needs study: Final report (Publication No. DOE/EP-0006). https://www.energy.gov/sites/default/files/2023-12/National%20Transmission%20Needs%20Study%20-%20Final_2023.12.1.pdf

8.) Shehabi, A.; Newkirk, A.; Smith, S.; Hubbard, A.; Lei, N.; Siddik, M., et al. (2024). 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory. Report #: LBNL-2001637. http://dx.doi.org/10.71468/P1WC7Q Retrieved from https://escholarship.org/uc/item/32d6m0d1

9.) IBM. (n.d.). Hyperscale data center. IBM. https://www.ibm.com/think/topics/hyperscale-data-center

10.) McKinsey & Company. (2025, February). The cost of compute: A $7 trillion race to scale data centers. McKinsey & Company. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers

11.) Dominion Energy. (2024). 2024 Integrated Resource Plan [PDF]. https://cdn-dominionenergy-prd-001.azureedge.net/-/media/content/about/our-company/irp/pdfs/2024-irp-w_o-appendices.pdf?rev=5b28b014e4814135bb2fcec470dcc92b

12.) Virginia Economic Development Partnership. (n.d.). Data centers. Virginia Economic Development Partnership. https://www.vedp.org/industry/data-centers

13.) Wikipedia contributors. (n.d.). Levelized cost of electricity. Wikipedia. Retrieved June 24, 2025, from https://en.wikipedia.org/wiki/Levelized_cost_of_electricity

14.) Hoes, R. (2024, May 6). Onsite power for data centers: How to choose the right solution. Data Center Frontier. https://www.datacenterfrontier.com/sponsored/article/55276009/onsite-power-for-data-centers-how-to-choose-the-right-solution

15.) Electric Reliability Council of Texas. (2025, April 1). System planning and weatherization update [Presentation]. https://www.ercot.com/files/docs/2025/04/01/14.2-System-Planning-and-Weatherization-Update.pdf

16.) Vantage Data Centers. (2024, February 13). Vantage Data Centers breaks ground on 192MW campus, enters growing Ohio data center market with over $2 billion investment in hyperscale digital infrastructure. https://vantage-dc.com/news/vantage-data-centers-breaks-ground-on-192mw-campus-enters-growing-ohio-data-center-market-with-over-2-billion-investment-in-hyperscale-digital-infrastructure/

17.) American Electric Power. (2024, October 19). AEP Ohio, PUCO staff, Ohio Consumers’ Counsel, Ohio Energy Group and others agree on how to address growing data center power needs. https://www.aep.com/news/stories/view/9829/AEP-Ohio-PUCO-Staff-Ohio-Consumers-Counsel-Ohio-Energy-Group-and-Others-Agree-on-How-to-Address-Growing-Data-Center-Power-Needs/

18.) Shilov, A. (2024, May 3). Power utility built $95 million, 500-megawatt power substation for Intel’s $100 billion Ohio fab, but six-year delay leaves substation capable of powering 500,000 homes idle. Tom's Hardware. https://www.tomshardware.com/tech-industry/semiconductors/power-utility-built-usd95-million-500-megawatt-power-substation-for-intels-usd100-billion-ohio-fab-but-six-year-delay-leaves-substation-capable-of-powering-500-000-homes-idle

19.) Federal Energy Regulatory Commission. (2024, April 8). Order on transmission planning and cost allocation—PJM Interconnection, L.L.C. (Accession No. 20240408-3047). https://elibrary.ferc.gov/eLibrary/filelist?accession_number=20240408-3047&optimized=false

20.) Sinclair Broadcast Group. (2024, April 18). ‘Unjust and unreasonable’: People’s Counsel speaks out over transmission costs. Fox Baltimore. https://foxbaltimore.com/news/local/unjust-and-unreasonable-peoples-counsel-speaks-out-over-transmission-costs

21.) Data Center Dynamics. (2024, April 17). Crusoe secures 4.5GW of gas for AI data centers: Report. https://www.datacenterdynamics.com/en/news/crusoe-secures-45gw-of-a-gas-for-ai-data-centers-report/

22.) Construction Today. (2024, May 6). Microsoft bets $80 billion on AI, data centers, and sustainability. https://construction-today.com/news/microsoft-bets-80-billion-on-ai-data-centers-and-sustainability/

23.) Heynes, G. (2024, October 14). Arizona’s biggest battery storage system goes online to feed Meta data centre demand. Energy-Storage.news. https://www.energy-storage.news/arizonas-biggest-battery-storage-system-goes-online-to-feed-meta-data-centre-demand/

24.) Cyber Defense Magazine. (2024, February 12). A cloud reality check for federal agencies. https://www.cyberdefensemagazine.com/a-cloud-reality-check-for-federal-agencies/

25.) Kindervag, J. (2022, December 9). The DoD announces the winners for its $9B JWCC contract. Forrester. https://www.forrester.com/blogs/the-dod-announces-the-winners-for-its-9b-jwcc-contract/

26.) Wheeler, T. (2024, May 9). FERC rejects interconnection plan for Amazon data center, Talen’s Susquehanna nuclear plant. Utility Dive. https://www.utilitydive.com/news/ferc-interconnection-isa-talen-amazon-data-center-susquehanna-exelon/731841/

27.) Talen Energy. (2025, June 5). Talen Energy expands nuclear energy relationship with Amazon. https://ir.talenenergy.com/news-releases/news-release-details/talen-energy-expands-nuclear-energy-relationship-amazon

28.) Environmental and Energy Study Institute. (n.d.). Beneficial electrification. Retrieved June 25, 2025, from https://www.eesi.org/electrification/be

29.) International Energy Agency. (2024, April). Global EV Outlook 2024: Trends in electric cars. International Energy Agency. Retrieved June 25, 2025, from https://www.iea.org/reports/global-ev-outlook-2024/trends-in-electric-cars

30.) National Renewable Energy Laboratory. (2024, May 28). How many transformers will the U.S. distribution grid need by 2050? Retrieved June 25, 2025, from https://www.nrel.gov/news/detail/program/2024/how-many-transformers-will-the-us-distribution-grid-need-by-2050#:~:text=This%20analysis%20is%20based%20on%20estimating%20the

31.) Georgia Public Service Commission. (2022, January 31). Georgia Power Company: 2022 integrated resource plan (Document No. 188519). Retrieved June 25, 2025, from https://psc.ga.gov/search/facts-document/?documentId=188519

32.) Georgia Power. (2025, January 31). Georgia Power files 2025 IRP plan to meet energy needs. Georgia Power News Hub. https://www.georgiapower.com/news-hub/press-releases/georgia-power-files-2025-irp-plan-to-meet-energy-needs.html

33.) Public Advocates Office, California Public Utilities Commission. (2023, August). Distribution Grid Electrification Model – Study and Report [PDF]. Retrieved June 25, 2025, from https://www.publicadvocates.cpuc.ca.gov/-/media/cal-advocates-website/files/press-room/reports-and-analyses/230824-public-advocates-distribution-grid-electrification-model-study-and-report.pdf

34.) Pacific Gas and Electric Company. (2025, May 27). Surging data center growth to help lower energy costs for PG&E customers [Press release]. Pacific Gas and Electric Company. Retrieved June 25, 2025, from https://investor.pgecorp.com/news-events/press-releases/press-release-details/2025/Surging-Data-Center-Growth-to-Help-Lower-Energy-Costs-for-PGE-Customers/default.aspx

35.) Howland, E. (2024, November 11). Rising data center loads pose grid reliability, residential cost risks: APS executive. Utility Dive. https://www.utilitydive.com/news/data-center-grid-reliability-residential-cost-aps-load-growth/732480/

36.) U.S. Department of Energy, Office of Policy. (2024, October). Transmission impact assessment: Power sector infrastructure deployment to reduce costs, improve reliability, and lower pollution. https://www.energy.gov/sites/default/files/2024-10/DOE_OP_2024_Report-Transmission_Impact_Assessment.pdf

37.) Arizona Public Service. (2023, November 1). Looking ahead: New APS plan addresses growing customer energy needs. APS Newsroom. https://www.aps.com/en/About/Our-Company/Newsroom/Articles/New-APS-Plan-Addresses-Growing-Customer-Energy-Needs

38.) McKenna, K., Abraham, S. A., Wang, W., & Poudyal, A. (2024, November). Distribution transformer demand: Understanding demand segmentation, drivers, and management through 2050 (NREL/FS-6A40-92076). National Renewable Energy Laboratory. https://docs.nrel.gov/docs/fy25osti/92076.pdf

39.) U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy. (2022, December). Energy conservation program: Energy conservation standards for distribution transformers; Notice of proposed rulemaking. https://www.energy.gov/sites/default/files/2022-12/dt-ecs-nopr.pdf

40.) National Infrastructure Advisory Council. (2024, June 11). Addressing the critical shortage of power transformers to ensure reliability of the U.S. grid. Cybersecurity and Infrastructure Security Agency. https://www.cisa.gov/sites/default/files/2024-09/NIAC_Addressing%20the%20Critical%20Shortage%20of%20Power%20Transformers%20to%20Ensure%20Reliability%20of%20the%20U.S.%20Grid_Report_06112024_508c_pdf_0.pdf

41.) U.S. Department of Energy. (2022, June 6). President Biden invokes Defense Production Act to accelerate domestic manufacturing of clean energy. https://www.energy.gov/articles/president-biden-invokes-defense-production-act-accelerate-domestic-manufacturing-clean

42.) Dunlap, S. (2025, June 1). Georgia Power’s plan to support potential data center surge with fossil fuel energy faces scrutiny. Yahoo News. https://www.yahoo.com/news/georgia-power-plan-support-potential-070016649.htm

43.) Munster, G., & Baker, B. (2024, December 6). Salesforce’s AI message is resonating with investors. Deepwater Asset Management. https://deepwatermgmt.com/salesforces-ai-message-is-resonating-with-investors/

44.) Steinberg, B. (2025, March 6). McDonald’s to employ AI at 43K locations to speed up service: “Technology solutions will alleviate the stress.” New York Post. https://nypost.com/2025/03/06/lifestyle/mcdonalds-to-employ-ai-at-43k-locations-to-speed-up-service

45.) TIGER-Lab. (n.d.). MMLU-Pro [Interactive model card]. Hugging Face. Retrieved June 25, 2025, from https://huggingface.co/spaces/TIGER-Lab/MMLU-Pro

46.) Weiss, D. C. (2023, March 16). Latest version of ChatGPT aces the bar exam with score in 90th percentile. ABA Journal. https://www.abajournal.com/web/article/latest-version-of-chatgpt-aces-the-bar-exam-with-score-in-90th-percentile

47.) Briggs, J., & Kodnani, D. (2023, March 27). Generative AI could raise global GDP by 7%. Goldman Sachs Economics Research. https://www.gspublishing.com/content/research/en/reports/2023/03/27/d64e052b-0f6e-45d7-967b-d7be35fabd16.html

48.) World Economic Forum. (2025). The Future of Jobs Report 2025. https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf

49.) National Energy Assistance Directors Association. (n.d.). NEAD toolkit: Data [Dataset]. Retrieved June 25, 2025, from https://neada.org/neadtoolkit/data/

50.) Donaldson, I., & Wheelock, L. (2025, January 25). Written testimony for the SFY-2026 New York State Environmental Conservation budget hearing [Testimony]. Public Utility Law Project of New York. https://www.nysenate.gov/sites/default/files/admin/structure/media/manage/filefile/a/2025-01/public-utility-law-project.pdf

51.) Kowanko, A., & Harak, C. (2021, November). More can and must be done to prevent utility consumers from losing service due to mounting COVID-driven arrearages in Massachusetts and other states (Report). National Consumer Law Center. https://www.nclc.org/wp-content/uploads/2022/10/Rpt_More_Covid_Util_Arrearage_Svc.pdf

52.) U.S. Bureau of Labor Statistics. (2020, May 8). The employment situation — April 2020 (News Release No. USDL-20-0832). https://www.bls.gov/news.release/archives/empsit_05082020.pdf

53.) LeMoult, C. (2025, May 15). Trump budget proposal would end energy assistance program for low-income Americans. NPR. https://www.npr.org/2025/05/15/nx-s1-5389917/trump-budget-proposal-would-end-energy-assistance-program-for-low-income-americans

54.) Hua, C. (2025, April). Utility bills are rising: An analysis of how utility bills are impacting American energy consumers and who determines them. PowerLines. https://powerlines.org/wp-content/uploads/2025/04/PowerLines_Utility-Bills-Are-Rising_2025-1.pd

55.) Smart Energy Consumer Collaborative. (2024, June 27). New survey: Addressing energy affordability. Smart Energy Consumer Collaborative. https://smartenergycc.org/new-survey-addressing-energy-affordability/

56.) Moody’s Investors Service. (2025, May 5). Data centers – US: Regulated utilities move to limit risk amid still-surging data-center power demand [Outlook report]. https://www.moodys.com/web/en/us/insights/credit-risk/data-centers.html

📘 About Load to Ruin

Load to Ruin is the first installment in a three-part white paper series exploring the collapse of the legacy investor-owned utility (IOU) business model under the combined weight of hyperscale growth, AI-driven job displacement, and infrastructure cost misalignment.

This paper is available in three formats:

• Full Article: Published in its entirety here for open access. No email gating EVER!

• Mobile Reading Version: Read the optimized edition on Carrd—ideal for immersive, distraction-free reading on mobile phones

• PDF Download: Available at the bottom of this page for offline access or traditional formatting.

Over the coming months, I’ll be publishing:

• Two additional white papers, covering:

  1. Scenario planning for post-labor grid finance models.

  2. Policy interventions that could prevent collapse or accelerate transition.

• A companion blog series, designed for non-industry leaders—investors, public officials, journalists, and engaged citizens—who want to understand how today’s utility economics are being quietly rewritten by forces they didn’t see coming.

If you're a policymaker, regulator, energy investor, or just grid-curious, I hope this series offers a clearer view into what’s really happening beneath rising utility bills—and who pays when the system breaks.

→ Ready to dive in? Read the full paper here!

Subscribe now

Prefer a printable version? Download the PDF!

6.42KB ∙ PDF file

Download

Download