Sanders and AOC Want to Pause AI Data Centers—But the Real Fight Is Over Who Gets to Raise Your Electric Bill First
The moratorium is the headline. The underlying conflict is cost allocation: whether households subsidize hyperscale data centers through regulated rates—or data centers pay up front.
By TheMurrow Editorial
March 30, 2026
Key Points
- 1Track the real battleground: utility cost allocation and rate design that decides whether households subsidize grid upgrades for hyperscale AI data centers.
- 2Recognize the scale: AI data centers can rival ~100,000 homes in load, stressing peak planning, reserve margins, and transmission timelines.
- 3Watch the institutions that matter: state commissions, PJM/ERCOT forecasts, and FERC rules will shape reliability, interconnection, and who pays first.
A new bill, a familiar dilemma—and your power bill
On March 25, 2026, Sen. Bernie Sanders and Rep. Alexandria Ocasio‑Cortez stood in front of a familiar American dilemma and gave it a new name: the Artificial Intelligence (AI) Data Center Moratorium Act (2026). The bill calls for an immediate federal moratorium on AI data centers until “strong national safeguards” exist.
The headline reads like an argument about the future—about whether the United States should slow down or sprint ahead in artificial intelligence. The real fight, though, sits in a less glamorous place: your monthly electricity bill.
Data centers are not just buildings filled with servers. They are enormous new loads on regional grids already coping with generation retirements, transmission limits, and tightening reserve margins. When utilities expand capacity to serve that load, the dispute becomes brutally concrete: who pays for the wires and power plants? If the answer is “everyone,” then households can be asked to subsidize a boom built for a small number of hyperscale customers.
The headline reads like an argument about the future—about whether the United States should slow down or sprint ahead in artificial intelligence. The real fight, though, sits in a less glamorous place: your monthly electricity bill.
Data centers are not just buildings filled with servers. They are enormous new loads on regional grids already coping with generation retirements, transmission limits, and tightening reserve margins. When utilities expand capacity to serve that load, the dispute becomes brutally concrete: who pays for the wires and power plants? If the answer is “everyone,” then households can be asked to subsidize a boom built for a small number of hyperscale customers.
“The most consequential AI policy debate in 2026 may be hiding in plain sight: it’s about rate design, not science fiction.”
— — TheMurrow
The Sanders–AOC moratorium: what it is—and what it signals
Sanders (I‑VT) and Ocasio‑Cortez (D‑NY) announced their legislation on March 25, 2026, pitching it as a consumer‑ and worker‑protection pause while policymakers write rules to match the speed of development. According to coverage of the proposal, the bill aims to halt AI data center construction until national safeguards are established. The sponsors argue communities are facing higher electricity prices and environmental concerns tied to rapid buildouts.
A central talking point attached to the proposal is energy intensity. Reporting has highlighted that a “typical AI‑focused data center” can consume electricity comparable to about 100,000 households. That figure is doing more than communicating scale—it’s framing the political intuition behind the bill: local residents are being asked to absorb the stress of a technology they did not vote for, in a market they cannot easily exit.
The bill’s near‑term prospects appear limited. Major outlets describe it as unlikely to advance in the current Congress, but politically meaningful. The proposal functions less like a legislative endgame and more like a flare: a signal that progressive lawmakers are willing to target AI’s physical infrastructure, not merely its algorithms.
A central talking point attached to the proposal is energy intensity. Reporting has highlighted that a “typical AI‑focused data center” can consume electricity comparable to about 100,000 households. That figure is doing more than communicating scale—it’s framing the political intuition behind the bill: local residents are being asked to absorb the stress of a technology they did not vote for, in a market they cannot easily exit.
The bill’s near‑term prospects appear limited. Major outlets describe it as unlikely to advance in the current Congress, but politically meaningful. The proposal functions less like a legislative endgame and more like a flare: a signal that progressive lawmakers are willing to target AI’s physical infrastructure, not merely its algorithms.
~100,000 households
Coverage cites a “typical AI‑focused data center” using electricity comparable to about 100,000 households—framing why communities fear higher rates and local strain.
What the proposal gets right about public anxiety
The moratorium reflects a public reality that technologists sometimes underestimate: people experience AI not as an abstract model, but as construction, noise, water use, land deals, and utility hearings. When communities feel blindsided, the backlash tends to focus on the most visible symbol: the data center itself.
What the proposal leaves unanswered
A federal halt also raises hard questions. What counts as an “AI data center” in practice? How would regulators distinguish AI‑training capacity from general cloud infrastructure? Coverage of the bill focuses on the pause and safeguards, not on a clean operational definition—an issue opponents will seize on, and supporters will eventually have to confront.
“A moratorium is a blunt instrument. The pressure behind it is anything but blunt—it’s granular, local, and financial.”
— — TheMurrow
The backlash: “idiocy,” China, and the politics of falling behind
The counterargument arrived quickly and with little subtlety. Sen. Mark Warner (D‑VA), a leading voice on tech policy, attacked the idea of a moratorium as “idiocy,” warning it could hand advantage to China in AI. His critique reflects a mainstream Washington instinct: treat AI infrastructure as a strategic asset, not just a zoning dispute.
That view is politically potent because it compresses multiple anxieties into one: national security, economic competitiveness, and technological prestige. If you accept the premise that AI leadership depends on compute—and compute depends on data centers—then slowing construction looks like unilateral disarmament.
Industry‑aligned policy voices have also rejected the moratorium’s premise. The Center for Data Innovation (part of ITIF, a tech policy think tank) argued the bill’s rationale “doesn’t add up,” disputing whether the harms cited justify halting construction. Readers should treat that position as what it is: advocacy. Still, it captures a serious argument—one that resonates with governors courting investment and with regions trying to land high‑value projects.
That view is politically potent because it compresses multiple anxieties into one: national security, economic competitiveness, and technological prestige. If you accept the premise that AI leadership depends on compute—and compute depends on data centers—then slowing construction looks like unilateral disarmament.
Industry‑aligned policy voices have also rejected the moratorium’s premise. The Center for Data Innovation (part of ITIF, a tech policy think tank) argued the bill’s rationale “doesn’t add up,” disputing whether the harms cited justify halting construction. Readers should treat that position as what it is: advocacy. Still, it captures a serious argument—one that resonates with governors courting investment and with regions trying to land high‑value projects.
The real political split: national ambition vs. local pain
Washington talks in macro terms: growth, innovation, strategic rivalry. Local communities deal in micro terms: transformer upgrades, water concerns, and escalating rates. The moral and political tension comes from the mismatch between who benefits and who bears the costs.
A policy debate that’s easy to caricature
Moratorium advocates can paint opponents as indifferent to households and the environment. Moratorium opponents can paint advocates as naïve about global competition. Neither caricature helps a family looking at a higher utility bill, or a grid operator staring at a queue of new load requests.
The hidden core issue: electricity bills and who pays for upgrades
The most underappreciated element of the data center boom is not the servers. It’s the regulated utility system built around cost allocation—the quiet machinery determining who funds new capacity and transmission.
Utilities plan for peak demand, not average demand. A massive new customer changes the peak and forces new investment: generation procurement, transmission upgrades, substations, and sometimes local distribution improvements. When those costs are rolled into regulated rates, households can pay—even if the expansion primarily serves a handful of very large customers.
That is why the debate is drifting from “should data centers exist?” to a more surgical question: should data centers be placed in a separate class that pays more of the costs up front?
Utilities plan for peak demand, not average demand. A massive new customer changes the peak and forces new investment: generation procurement, transmission upgrades, substations, and sometimes local distribution improvements. When those costs are rolled into regulated rates, households can pay—even if the expansion primarily serves a handful of very large customers.
That is why the debate is drifting from “should data centers exist?” to a more surgical question: should data centers be placed in a separate class that pays more of the costs up front?
Why peak planning matters more than annual averages
A region can have plenty of energy across a year and still face reliability risks on the hottest days. Data centers—especially AI‑focused operations running dense compute—can be relentless loads. Even when they sign contracts or buy renewable credits, they still require the grid to deliver power at the moments the system is most strained.
The policy lever most likely to move: rate design
Congress can propose moratoriums. State utility commissions decide rates. For households, the most practical question is less “Will Washington stop data centers?” and more “Will regulators require them to pay their share of the grid they need?”
“Every new megawatt is political once the costs hit a regulated ratepayer.”
— — TheMurrow
The grid operators are already telling us what’s happening (PJM, ERCOT, FERC)
You do not need to rely on rhetoric to understand the trend. Grid operators and federal regulators are documenting it in plain terms: the load growth is real, and it is increasingly driven by data centers.
In the PJM Interconnection region—spanning much of the Mid‑Atlantic and parts of the Midwest—PJM’s “Year in Review” materials emphasize that the 2025 Long‑Term Load Forecast focused on “unprecedented” demand growth, “driven primarily by data centers.” PJM projects data‑center growth up to about 30 gigawatts (GW) between 2025 and 2030 within its footprint. Thirty gigawatts is not a rounding error. It is the kind of number that reorders planning priorities.
Texas tells a similar story. ERCOT saw its large‑load interconnection requests jump almost 300% in 2025, largely driven by data centers. That surge has forced policy work on standardized interconnection rules and improved load forecasting—bureaucratic phrases that translate into one meaning: the system is scrambling to keep up.
Federal reliability assessments are also flashing yellow. FERC’s 2025 Summer Assessment warned that reserve margins are tightening in multiple regions as load increases “largely due to hyperscale users, such as data centers,” alongside other stressors like generation retirements.
In the PJM Interconnection region—spanning much of the Mid‑Atlantic and parts of the Midwest—PJM’s “Year in Review” materials emphasize that the 2025 Long‑Term Load Forecast focused on “unprecedented” demand growth, “driven primarily by data centers.” PJM projects data‑center growth up to about 30 gigawatts (GW) between 2025 and 2030 within its footprint. Thirty gigawatts is not a rounding error. It is the kind of number that reorders planning priorities.
Texas tells a similar story. ERCOT saw its large‑load interconnection requests jump almost 300% in 2025, largely driven by data centers. That surge has forced policy work on standardized interconnection rules and improved load forecasting—bureaucratic phrases that translate into one meaning: the system is scrambling to keep up.
Federal reliability assessments are also flashing yellow. FERC’s 2025 Summer Assessment warned that reserve margins are tightening in multiple regions as load increases “largely due to hyperscale users, such as data centers,” alongside other stressors like generation retirements.
~30 GW
PJM projects data‑center growth up to about 30 gigawatts from 2025–2030—load large enough to reorder regional planning priorities.
~300%
ERCOT saw large‑load interconnection requests jump almost 300% in 2025, largely driven by data centers—forcing new rules and forecasting.
Four statistics that frame the problem
- A “typical AI‑focused data center” can use electricity comparable to ~100,000 households (as cited in coverage of the Sanders–AOC push).
- PJM projects data‑center growth up to ~30 GW from 2025 to 2030—a structural shift, not a temporary spike.
- ERCOT saw ~300% growth in large‑load interconnection requests in 2025, much of it tied to data centers.
- FERC warns reserve margins are tightening in multiple regions, with load increases “largely due to hyperscale users, such as data centers.”
- PJM projects data‑center growth up to ~30 GW from 2025 to 2030—a structural shift, not a temporary spike.
- ERCOT saw ~300% growth in large‑load interconnection requests in 2025, much of it tied to data centers.
- FERC warns reserve margins are tightening in multiple regions, with load increases “largely due to hyperscale users, such as data centers.”
The planning strain is administrative as well as physical
PJM has also emphasized the sheer volume of new requests and forecasting adjustments connected to large loads. Translation: even before you pour concrete for new transmission, the system’s planning apparatus is being stress‑tested by the pace of demand.
Infrastructure bottlenecks: interconnection, transmission, and the limits of “just build more”
The popular answer to any infrastructure challenge is “build more.” The grid’s answer is more complicated: “build more, but it takes time, approvals, and a place to connect.”
On the generation side, interconnection queues have become a national chokepoint. FERC’s Order No. 2023 aims to reduce backlogs and speed the process of connecting new generation to the grid, acknowledging that the surge in requests and limited transmission capacity have worsened delays. Independent reporting and analysis has noted that queue volumes are enormous and that reforms are only slowly taking effect.
This matters for the data center debate because one common reassurance—“they’ll be powered by new clean energy”—depends on the ability to connect that energy. If the queue is jammed, data centers may rely more heavily on existing resources in the near term, tightening supply and potentially affecting prices.
On the generation side, interconnection queues have become a national chokepoint. FERC’s Order No. 2023 aims to reduce backlogs and speed the process of connecting new generation to the grid, acknowledging that the surge in requests and limited transmission capacity have worsened delays. Independent reporting and analysis has noted that queue volumes are enormous and that reforms are only slowly taking effect.
This matters for the data center debate because one common reassurance—“they’ll be powered by new clean energy”—depends on the ability to connect that energy. If the queue is jammed, data centers may rely more heavily on existing resources in the near term, tightening supply and potentially affecting prices.
Reliability is a timeline problem
A data center can go from announcement to operation faster than a major transmission project can go from proposal to completion. That mismatch drives the tension between local approvals and regional planning. Communities see rapid construction; grid operators see multi‑year lead times.
The hard truth about “self-supply” claims
Even when large customers pursue dedicated generation, the grid still functions as the backbone. Backup arrangements and interconnection still implicate regional reliability and cost allocation. The argument is not that self-supply is meaningless—it’s that it rarely severs the public relationship.
So what would “strong national safeguards” look like without a moratorium?
The Sanders–AOC framing—pause until safeguards—invites a practical follow‑up: what safeguards, specifically, would address the legitimate concerns without freezing the entire sector?
The strongest candidates are not glamorous, but they are legible:
- Transparent cost allocation so ratepayers can see whether grid upgrades are being socialized.
- Stronger interconnection and service rules for large loads, including standardized requirements.
- Reliability obligations that prevent large new loads from undermining reserve margins.
- Clear reporting of energy demand projections and associated infrastructure plans.
None of these require a national ban to begin. Many can be pursued through coordination among federal regulators, grid operators, and state utility commissions.
The strongest candidates are not glamorous, but they are legible:
- Transparent cost allocation so ratepayers can see whether grid upgrades are being socialized.
- Stronger interconnection and service rules for large loads, including standardized requirements.
- Reliability obligations that prevent large new loads from undermining reserve margins.
- Clear reporting of energy demand projections and associated infrastructure plans.
None of these require a national ban to begin. Many can be pursued through coordination among federal regulators, grid operators, and state utility commissions.
Key Insight
The most actionable “safeguards” aren’t about algorithms. They’re enforceable rules on cost allocation, interconnection, reliability, and transparent load forecasting.
“Safeguards” should mean enforceable commitments, not PR
National safeguards that exist only as voluntary pledges will not satisfy the communities driving the backlash. The public is asking for enforceable rules: if a hyperscale customer changes local demand, residents want assurance they will not be left holding the bill.
A moratorium’s biggest unintended consequence
A blanket pause could push development to jurisdictions with weaker oversight—precisely the opposite of what consumer‑protection advocates want. If the political goal is accountability, the policy design has to anticipate geographic arbitrage.
Practical implications: what readers should watch (and what to ask)
For most people, the data center fight will surface through local news: a rezoning hearing, a utility rate case, a new substation proposal. The national rhetoric—China, innovation, safety—will be loud. The consequential details will be tucked into filings and forecasts.
Here is what to watch and what to ask, even if you never plan to read a grid operator’s report.
Here is what to watch and what to ask, even if you never plan to read a grid operator’s report.
Where the costs show up
- Utility rate cases: Are infrastructure costs being rolled into general rates?
- Grid planning documents: Do forecasts explicitly cite data centers as drivers of peak demand?
- Interconnection and service agreements: Are large loads required to fund upgrades up front?
- Grid planning documents: Do forecasts explicitly cite data centers as drivers of peak demand?
- Interconnection and service agreements: Are large loads required to fund upgrades up front?
Questions worth bringing to local and state proceedings
- Will the data center pay for transmission and distribution upgrades required to serve it?
- Are there protections against households subsidizing capacity built for a single class of customers?
- What reliability standards apply if load ramps faster than expected?
- Are there protections against households subsidizing capacity built for a single class of customers?
- What reliability standards apply if load ramps faster than expected?
A useful way to interpret the politics
When you hear “moratorium,” think “leverage.” When you hear “China,” think “permission structure.” Both sides are searching for a narrative strong enough to overcome the other side’s facts. The best public outcome will come from policies that treat electricity costs as the core issue rather than a footnote.
What to watch in your area
- ✓Rezoning hearings and site approvals
- ✓Utility rate cases and cost-allocation decisions
- ✓Substation and transmission upgrade proposals
- ✓Large-load interconnection and service agreements
- ✓Grid operator load forecasts citing data centers
Conclusion: the moratorium is a headline; the rate fight is the story
Sanders and Ocasio‑Cortez have put a provocative idea on the table: stop AI data centers until the country catches up with safeguards. Warner and other critics have responded with an equally blunt warning: don’t cripple U.S. competitiveness and hand China an advantage.
Both arguments contain a truth. AI infrastructure is strategic. Communities are also being asked to absorb rapid, expensive change.
The test for policymakers is whether they can resist the comfort of absolutes. A national moratorium is easy to chant and hard to implement. A granular approach—cost allocation, reliability rules, transparent planning—is harder to sell on cable news and far more likely to matter to your household.
The next phase of the AI debate will not be decided only in Congress. It will be decided in utility commissions, grid operator forecasts, and the unglamorous accounting of who pays for power.
Both arguments contain a truth. AI infrastructure is strategic. Communities are also being asked to absorb rapid, expensive change.
The test for policymakers is whether they can resist the comfort of absolutes. A national moratorium is easy to chant and hard to implement. A granular approach—cost allocation, reliability rules, transparent planning—is harder to sell on cable news and far more likely to matter to your household.
The next phase of the AI debate will not be decided only in Congress. It will be decided in utility commissions, grid operator forecasts, and the unglamorous accounting of who pays for power.
Editor's Note
If you want to understand how this debate affects you, follow state utility commission dockets and grid operator forecasts—not just congressional headlines.
T
About the Author
TheMurrow Editorial is a writer for TheMurrow covering explainers.
Frequently Asked Questions
What is the AI Data Center Moratorium Act (2026)?
Announced March 25, 2026 by Sen. Bernie Sanders and Rep. Alexandria Ocasio‑Cortez, it calls for an immediate federal moratorium on AI data centers until “strong national safeguards” are established.
How much electricity does an AI data center use?
Coverage frequently cites a “typical AI‑focused data center” using electricity comparable to about 100,000 households, though actual usage varies by facility.
Why are electricity bills central to this debate?
Utilities plan for peak demand; adding a massive new load can require new generation and grid upgrades. If those costs are rolled into general rates, households can subsidize infrastructure serving a handful of hyperscale users.
What evidence shows data centers are driving grid planning?
PJM says its 2025 Long‑Term Load Forecast focused on “unprecedented” growth driven primarily by data centers, projecting up to ~30 GW from 2025–2030. ERCOT reported ~300% growth in large‑load interconnection requests in 2025, largely tied to data centers.
What does Mark Warner mean by calling the moratorium “idiocy”?
Warner’s critique frames a moratorium as strategically reckless—risking U.S. AI leadership and ceding advantage to China—because AI capability depends on compute and compute depends on data center buildout.
What should “national safeguards” include if not a moratorium?
Enforceable cost allocation, standardized large‑load interconnection rules, reliability obligations, and transparent reporting of demand forecasts and infrastructure plans to prevent households from subsidizing upgrades.
