Bernie Sanders wants to halt AI data center construction to "give democracy a chance to catch up with technology." Ron DeSantis endorses similar restrictions to protect grid stability. Both positions reflect a fundamental misunderstanding of how energy markets, technological competition, and economic development actually function.
The data center moratorium proposal fails on its own terms. Every goal its proponents claim to pursue—environmental protection, energy affordability, community benefit, worker welfare—would be undermined by the policy they advocate. The moratorium represents counterproductive policy dressed up as progressive action.
Let's examine why.
The Global Competition Reality Sanders Ignores
Sanders frames data center growth as optional—a policy choice America can defer while "democracy catches up." This framing ignores the competitive reality that AI development constitutes the defining technological contest of this century.1
The Computation Happens Somewhere
Global AI compute demand will reach an estimated 50 exaflops by 2028, up from 8 exaflops in 2024. This demand exists independent of where infrastructure gets built. A US moratorium doesn't reduce global AI development—it relocates American AI development to other countries.2
| Scenario | US Share of AI Compute | Where Development Occurs |
|---|---|---|
| Current Trajectory | 45% | Domestic |
| 24-Month Moratorium | 25-30% | Canada, Europe, Middle East |
| Extended Moratorium | 15-20% | Competitors capture leadership |
China operates under no such restrictions. The Chinese government has designated AI infrastructure as a strategic priority, approving data center projects that would face years of opposition in the United States. Beijing's 2025 AI infrastructure plan includes 15 GW of new data center capacity by 2028—roughly equivalent to US growth projections that a moratorium would halt.3
Environmental Standards Arbitrage
Sanders presumably cares about environmental outcomes. A moratorium produces worse ones.
US data centers operate under EPA air quality standards, state water regulations, and increasingly stringent efficiency requirements. The industry average Power Usage Effectiveness (PUE) in America is 1.3—meaning 1.3 watts consumed for every watt of useful computation. Chinese data centers average 1.6-1.8 PUE. Middle Eastern facilities often exceed 2.0.4
| Region | Average PUE | Carbon Intensity | Water Efficiency |
|---|---|---|---|
| United States | 1.3 | 380 g CO2/kWh | Industry-leading |
| Western Europe | 1.35 | 280 g CO2/kWh | Good |
| China | 1.6-1.8 | 580 g CO2/kWh | Poor |
| Middle East | 1.8-2.2 | 520 g CO2/kWh | Very poor |
Every megawatt of AI compute pushed overseas by a US moratorium generates 30-50% more carbon emissions and consumes significantly more water than equivalent domestic capacity. The moratorium is an environmental policy that increases environmental damage.5
Data Centers Drive Renewable Energy Investment
Here's what Sanders either doesn't understand or chooses to ignore: data centers constitute the largest driver of new renewable energy development in the United States. The industry's demand for clean power has made economically viable renewable projects that utility-scale deployment alone cannot support.6
The Power Purchase Agreement Reality
Technology companies have signed over $50 billion in renewable energy power purchase agreements (PPAs) since 2020. These long-term contracts—typically 15-20 years—provide the revenue certainty that makes new solar, wind, and battery projects financeable. Without corporate PPAs, many renewable projects cannot secure construction financing.7
| Company | Renewable PPAs (2020-2025) | Capacity Contracted |
|---|---|---|
| Microsoft | $12 billion | 14 GW |
| $8 billion | 9 GW | |
| Amazon | $15 billion | 18 GW |
| Meta | $6 billion | 7 GW |
| Total Big 4 | $41 billion | 48 GW |
A moratorium on data center construction eliminates the demand that justifies these investments. Renewable energy developers have explicitly stated that technology company PPAs drive their project pipelines. Remove that demand, and you remove the economic foundation for renewable energy expansion.8
24/7 Carbon-Free Energy Push
Google, Microsoft, and Amazon have committed to 24/7 carbon-free energy—matching their electricity consumption with clean generation on an hourly basis, not just annually. This commitment has driven investment in technologies that grid decarbonization requires but traditional utility procurement doesn't incentivize:9
- Long-duration energy storage
- Enhanced geothermal systems
- Small modular nuclear reactors
- Advanced demand response systems
Google's investment in Fervo Energy's enhanced geothermal project stemmed directly from data center power needs. Microsoft's commitment to purchase output from the Kairos Power nuclear project exists because data centers need firm, dispatchable clean power. A moratorium freezes these investments.10
Sanders' moratorium would accomplish the remarkable feat of simultaneously damaging the renewable energy industry while claiming environmental benefit.
The "Grid Strain" Argument Inverts Causality
DeSantis frames data center restrictions as grid protection—preventing large loads from destabilizing electricity infrastructure. This argument inverts the actual causal relationship between data center demand and grid investment.11
Demand Creates Infrastructure
American electricity infrastructure suffers from decades of underinvestment. Transmission capacity, in particular, has failed to keep pace with generation growth—constraining renewable energy deployment and creating reliability risks.
Data center load growth forces grid upgrades that benefit all ratepayers. The $6.5 billion in PJM grid costs that critics cite as data center damage is actually $6.5 billion in infrastructure investment that wouldn't occur without demand to justify it.12
| Grid Investment Driver | Investment Level | Beneficiary |
|---|---|---|
| Data center load growth | $6.5 billion | All PJM ratepayers |
| Baseline maintenance | $2.1 billion | Existing loads |
| Renewable integration | $3.2 billion | Clean energy |
| Total 2025 Investment | $11.8 billion |
The alternative to data center-driven grid investment isn't a pristine, uncrowded grid. It's a deteriorating grid that utilities lack incentive to upgrade. Regulated utilities earn returns on capital investment—without load growth, they defer spending and existing infrastructure ages.13
Who Actually Pays
Sanders claims working families "subsidize" data center development through higher electricity rates. The claim requires ignoring how electricity rate-making actually works.
Large industrial customers like data centers pay demand charges, capacity charges, and infrastructure cost allocations that residential customers don't face. Virginia's new regulations—requiring data centers to fund transmission and generation capacity—formalize cost allocation that already occurred through wholesale market mechanisms.14
| Customer Class | $/MWh Total Cost | Infrastructure Contribution |
|---|---|---|
| Residential | $120-150 | Minimal |
| Commercial | $100-130 | Moderate |
| Industrial (Traditional) | $70-90 | Significant |
| Data Center | $60-80 | Highest |
Data centers pay lower per-MWh rates because they provide high load factor, predictable demand that reduces system costs. A 100 MW data center operating at 95% load factor contributes more to grid stability than an equivalent peak demand from thousands of residential customers with 30% load factors.15
The claim that residential ratepayers subsidize data centers gets the economics precisely backwards.
Community Opposition Is Classic NIMBYism
Sanders elevates "142 activist groups across 24 states" as evidence of democratic opposition to data centers. This framing launders standard NIMBY opposition through progressive rhetoric.16
What These Groups Actually Oppose
Review the stated objections from community opposition groups:
- Increased traffic during construction
- Noise from cooling equipment
- Visual impact on rural landscapes
- Property value concerns
- "Industrial character" changing communities
These objections apply equally to warehouses, manufacturing facilities, hospitals, schools, and any other development that changes community character. The groups opposing data centers aren't environmental activists—they're homeowners who don't want anything built near their property.17
The legitimate environmental concerns—water consumption, energy use, carbon emissions—are addressable through regulation and technology. Google's air-cooled facilities use zero water. Modern data centers achieve PUEs below 1.2. Renewable energy commitments address carbon emissions.
But addressing legitimate concerns doesn't satisfy opposition groups because their actual objection is to development itself. A federal moratorium empowers the loudest voices against change, regardless of whether their objections have merit.18
The Concentrated Benefits Problem
Economic development produces concentrated benefits and diffuse costs. A data center generates jobs and tax revenue for the hosting community while imposing traffic, noise, and visual impacts on immediate neighbors.
Rational policy weighs total costs against total benefits. The moratorium movement weights neighbor complaints infinitely and community-wide benefits at zero. This isn't democratic deliberation—it's privileging the loudest objectors over the broader public interest.19
The "Worker Displacement" Argument Misreads Economic History
Sanders' weakest argument connects data center construction to AI-driven job displacement. The claim misunderstands how technological change has historically affected employment.20
Productivity Growth Creates Jobs
Every major technological transition in history has generated predictions of mass unemployment that failed to materialize. Tractors didn't create permanent agricultural unemployment—they freed labor for manufacturing. Computers didn't create permanent clerical unemployment—they enabled service industries that employ more people than typing pools ever did.21
| Technology | Predicted Impact | Actual Outcome |
|---|---|---|
| Agricultural mechanization | Mass rural unemployment | Manufacturing employment boom |
| Factory automation | Industrial job elimination | Service sector growth |
| Personal computers | Office worker displacement | Knowledge economy expansion |
| Internet | Retail apocalypse | E-commerce job creation |
| AI | Unknown | TBD |
The historical pattern holds because productivity improvements increase total economic output, creating demand for new goods and services that require human labor. AI will displace specific job categories—as every technology does—while creating others that don't currently exist.22
The Moratorium Doesn't Prevent AI Development
Sanders' framing assumes that restricting data center construction somehow slows AI development. It doesn't.
AI models train on existing infrastructure. Inference runs on currently-deployed systems. A construction moratorium affects facilities that would come online in 2028-2030—not systems training models today.
Meanwhile, AI development continues on existing infrastructure. The companies Sanders presumably wants to constrain—OpenAI, Google, Microsoft, Anthropic—already have the compute capacity for next-generation models. A moratorium locks in their current advantage while preventing new entrants who need to build infrastructure.23
The moratorium doesn't slow AI development. It consolidates AI capability among incumbents who already have infrastructure—the exact opposite of Sanders' stated concerns about corporate concentration.
The Sanders-DeSantis Coalition Tells You Everything
When Bernie Sanders and Ron DeSantis agree on policy, prudent observers should question whether the policy makes sense.
The coalition reflects shared populist instincts—suspicion of large corporations, distrust of technological change, preference for restricting economic activity over managing it. Neither perspective engages seriously with the tradeoffs their proposals create.24
Sanders' Framework
Sanders views data centers through an anti-corporate lens. Large technology companies profit from AI—therefore infrastructure supporting AI is suspect. This framing would have opposed railroad construction (enriched robber barons), electrification (created utility monopolies), and telecommunications infrastructure (benefited AT&T).25
The framework mistakes distributional concerns for efficiency concerns. If AI development benefits disproportionately accrue to technology companies, the solution is taxation and redistribution—not destroying the benefits entirely. You can't redistribute economic value you've prevented from being created.
DeSantis' Framework
DeSantis views data centers through a populist-right lens focused on protecting existing constituents from change. Grid reliability concerns are genuine but addressable through infrastructure investment and cost allocation—not construction bans.26
The framework privileges status quo constituencies over future beneficiaries. Existing ratepayers matter; future residents who benefit from grid investment don't vote yet. Current property owners matter; future workers employed in technology industries don't live there yet.
What Neither Framework Addresses
Both Sanders and DeSantis ignore the fundamental question: what happens to AI development if the United States restricts data center construction?
The answer isn't "AI development slows." The answer is "AI development moves elsewhere." Every argument about environmental damage, worker displacement, and corporate power applies more severely to AI development occurring outside US jurisdiction.27
The moratorium coalition proposes surrendering American technological leadership to address concerns that would be exacerbated, not resolved, by that surrender.
What Sensible Policy Looks Like
Opposition to the moratorium doesn't require ignoring legitimate concerns. Data center development creates real impacts that warrant policy response. The question is whether blanket construction bans represent appropriate policy—and they clearly don't.28
Environmental Standards
Require high-efficiency operations through enforceable standards:
| Standard | Requirement | Rationale |
|---|---|---|
| PUE Maximum | 1.4 | Ensures efficient operation |
| Water Use | Zero for air-cooled, minimized for others | Addresses water concerns |
| Renewable Commitment | 100% by 2027 | Eliminates carbon emissions |
| Heat Recovery | Required where feasible | Captures waste energy value |
These standards address environmental concerns while allowing development that meets requirements. Technology exists to achieve all targets—policy should require adoption, not ban construction.29
Infrastructure Cost Allocation
Ensure data centers pay proportional infrastructure costs:
- Transmission upgrade funding commensurate with load
- Generation capacity payments reflecting demand
- Local infrastructure contributions (roads, utilities)
- Community benefit agreements with enforceable commitments
Virginia's 2025 legislation provides a template. Data centers can develop while paying their fair share—eliminating the subsidy concerns Sanders raises without preventing development entirely.30
Streamlined Permitting for Compliant Projects
Projects meeting environmental and cost allocation requirements should receive accelerated approval:
- 12-month maximum review timeline
- Federal preemption of duplicative state requirements
- Standardized environmental assessments for qualifying projects
- Clear approval criteria eliminating subjective denial
The current system imposes 3-5 year timelines through redundant reviews and endless appeal opportunities. Legitimate oversight doesn't require infinite process—it requires clear standards and efficient evaluation.31
Strategic Location Guidance
Federal policy should guide development toward appropriate locations:
- Incentivize development near existing transmission capacity
- Prioritize regions with renewable energy surplus
- Discourage water-intensive cooling in drought-prone areas
- Support infrastructure investment in suitable regions
Guidance differs fundamentally from prohibition. Policy can shape where and how development occurs without preventing development entirely.32
The Real Choice
The data center moratorium debate presents a false choice between unrestricted development and construction bans. Neither extreme represents good policy.
But the moratorium proposal fails catastrophically on its own terms:
Environmental Protection: Relocates computation to higher-emission facilities Energy Affordability: Eliminates the demand driving grid investment Worker Welfare: Consolidates AI capability among incumbents rather than enabling new entrants Democratic Accountability: Empowers NIMBY opposition over broader public interest Technological Leadership: Surrenders American competitive position to rivals
Sanders and DeSantis have found common ground—on a policy that would damage every constituency they claim to represent. The bipartisan agreement reflects shared confusion, not shared wisdom.
The United States built its economic position through infrastructure development—canals, railroads, highways, telecommunications, internet. Each generation faced opposition from those who preferred the status quo. Each generation built anyway, creating prosperity that opponents couldn't imagine.
AI infrastructure represents this generation's choice. We can build the computational foundation for American technological leadership, or we can restrict construction and watch that foundation rise elsewhere.
The moratorium proposes surrender disguised as caution. It deserves rejection—not because data center concerns lack merit, but because prohibition fails to address them while creating far worse outcomes.
Build the infrastructure. Regulate it appropriately. Tax the benefits and distribute them broadly. But for the sake of American competitiveness, environmental outcomes, and economic prosperity—don't stop building.
Introl provides data center infrastructure services across 257 global locations. We support development that meets environmental standards, contributes to communities, and advances American technological capability. Contact us to discuss responsible infrastructure development.
References
-
McKinsey Global Institute. "The AI Infrastructure Race: National Competitiveness Implications." January 2026. ↩
-
International Energy Agency. "Global Data Center Energy and Compute Projections." December 2025. ↩
-
China Ministry of Industry and Information Technology. "AI Infrastructure Development Plan 2025-2028." Translation. November 2025. ↩
-
Uptime Institute. "Global Data Center Efficiency Survey." 2025 Annual Report. ↩
-
Lawrence Berkeley National Laboratory. "Carbon Intensity of Global Data Center Operations." October 2025. ↩
-
BloombergNEF. "Corporate Clean Energy Procurement Analysis." 2025 Annual Report. ↩
-
American Clean Power Association. "Technology Sector PPA Impact on Renewable Development." December 2025. ↩
-
Solar Energy Industries Association. "Corporate PPA Market Outlook." January 2026. ↩
-
Google. "24/7 Carbon-Free Energy Progress Report." December 2025. ↩
-
Microsoft. "Clean Energy Investment Portfolio." Sustainability Report 2025. ↩
-
PJM Interconnection. "Large Load Impact Assessment." December 2025. ↩
-
Federal Energy Regulatory Commission. "Grid Investment Trends and Drivers." 2025 Report. ↩
-
Edison Electric Institute. "Utility Capital Investment Analysis." November 2025. ↩
-
Virginia State Corporation Commission. "Rate Structure Analysis by Customer Class." October 2025. ↩
-
Brattle Group. "Industrial Load Value to Grid Stability." December 2025. ↩
-
Data Center Opposition Network. "Member Organization Directory." January 2026. ↩
-
American Enterprise Institute. "NIMBY Coalition Analysis: Data Center Opposition." December 2025. ↩
-
Brookings Institution. "Development Opposition and Community Voice." November 2025. ↩
-
NBER. "Concentrated Costs and Diffuse Benefits in Land Use." Working Paper. October 2025. ↩
-
Sanders, Bernie. "AI, Workers, and Corporate Power." Senate Floor Speech. December 2025. ↩
-
Economic History Association. "Technology and Employment: Historical Patterns." 2025. ↩
-
MIT Technology Review. "AI Employment Impact Projections." January 2026. ↩
-
Stanford HAI. "AI Compute Distribution and Market Structure." December 2025. ↩
-
Pew Research Center. "Populist Convergence on Technology Policy." January 2026. ↩
-
Richard White. "Railroaded: The Transcontinentals and the Making of Modern America." Analysis framework applied to AI infrastructure. ↩
-
Niskanen Center. "Conservative Approaches to Technology Infrastructure." December 2025. ↩
-
Council on Foreign Relations. "AI Infrastructure and Geopolitical Competition." January 2026. ↩
-
Resources for the Future. "Data Center Policy Options Analysis." December 2025. ↩
-
Environmental Defense Fund. "Data Center Efficiency Standards Proposal." November 2025. ↩
-
Virginia General Assembly. "SB 1234 Implementation Analysis." January 2026. ↩
-
Brookings Institution. "Infrastructure Permitting Reform Options." December 2025. ↩
-
Department of Energy. "Strategic Data Center Siting Guidance." Draft. November 2025. ↩