Robby Soave, Professional Libertarian, Dismisses Physics as 'Marxist' and Your Blackout as 'Innovation'

The Hill Opinion: Where Concerned Grid Operators Are Just Bureaucrats Who Hate Freedom

Robby Soave thinks he understands the data center surge. He does not. In his recent column for The Hill, Soave dismisses Senator Bernie Sanders’ call for a pause on new data center construction as “nutty,” suggesting it would hand a win to China and turn America into Europe, “where tech goes to die.” This is a charming narrative, but it falls apart the moment you look at the actual facts about what these data centers are doing to America’s electrical grid.

In the beginning there were numbers, and they were good

Let us start with the basic numbers. According to the Lawrence Berkeley National Laboratory, U.S. data centers consumed approximately 176 terawatt-hours of electricity in 2023, representing about 4.4 percent of total national electricity consumption (Lawrence Berkeley National Laboratory, 2024). That is already significant. But projections show this demand could reach 6.7 percent to 12 percent of total U.S. electricity demand by 2028. To put this in perspective, the International Energy Agency projects that data centers will drive about half of all electricity demand growth in the United States through 2030 (International Energy Agency, 2025).

False dichotomy

Soave frames this as a simple choice between American innovation and European stagnation. He claims Europe has “no Silicon Valley” because of high taxes and regulation. This is a nice story, except ASML, the Dutch company that manufactures the extreme ultraviolet lithography machines essential for producing the most advanced AI chips, happens to be European. Without ASML’s technology, the entire global AI industry would grind to a halt. Europe also hosts Mistral AI, a major AI competitor to OpenAI valued in the billions. Soave’s characterization of Europe as a place “where tech goes to die” is the kind of rhetoric that sounds clever until you realize it contradicts observable reality.

MORE POWER!!!!!

But the more serious problem with Soave’s argument is that he simply ignores the physical constraints of the electrical grid. This is not a theoretical concern. In March 2025, Reuters reported a “near miss” event in Data Center Alley, Virginia, where 60 data centers suddenly disconnected from the utility grid and switched to backup generators. This triggered what grid operators described as “a huge surge in excess electricity” that forced PJM Interconnection, which serves 65 million people across 13 states, to rapidly cut power plant output to avoid cascading blackouts (Reuters, 2025). John Moura, Director of Reliability Assessment at the North American Electric Reliability Corporation, stated plainly: “As these data centers get bigger and consume more energy, the grid is not designed to withstand the loss of 1,500-megawatt data centers” (Reuters, 2025).

PJM Interconnection projects that by 2027, its available power will fall 6 gigawatts short of reliability requirements, largely due to data center demand (Monitoring Analytics, 2026). The capacity market auction prices in PJM have already jumped from $28.92 per megawatt-day to $269.92 per megawatt-day. Monitoring Analytics, PJM’s independent market monitor, attributes $23 billion of recent capacity cost increases to data centers, noting these costs ultimately get passed to consumers. They describe this as “a massive wealth transfer” from ordinary ratepayers to tech companies (Monitoring Analytics, 2026).

This is what Sanders is actually talking about when he calls for a pause. It is not about hating technology or wanting America to lose to China. It is about recognizing that the current trajectory is physically unsustainable and economically regressive. When data centers consume electricity equivalent to small cities but employ fewer than one person per megawatt, while driving up electricity bills for everyone else, we have a problem that markets alone cannot solve.

Soave is manipulative and you can see it

Soave claims Sanders “thinks AI is bad” and “people making money off AI is bad.” This is a strawman. Sanders has explicitly stated his concern is about who benefits from AI and what happens to workers displaced by it. These are legitimate questions that deserve serious answers, not dismissive sneers about Marxism.

The comparison to China is particularly revealing. Soave warns that slowing data center construction means “falling behind China.” But China is precisely the country that has built the world’s most extensive high-voltage transmission network, with 42 ultra-high voltage lines enabling cross-regional power transfers of over 300 gigawatts. The United States, by contrast, built only 55.5 miles of new high-voltage transmission lines in 2024, the lowest in over a decade (Wood Mackenzie, 2025). The constraint on American AI development is not regulation. It is infrastructure that cannot keep up with demand.

Soave also ignores that China itself is limiting data center expansion in certain regions due to energy constraints. The idea that China is building data centers without regard for energy limits is false. They are building them strategically, with attention to resource allocation, which is exactly what Sanders is proposing.

What about the consumer impacts? Residential electricity prices rose 5 percent in 2025 and are projected to rise another 4 percent in 2026, partly driven by data center demand (U.S. Energy Information Administration, 2025). In Virginia, electricity rates have already increased due to data center expansion. In Arizona, Microsoft had to commit to water replenishment projects because data centers were straining local water supplies. One proposed facility in Tucson would have consumed 13 million gallons annually, equivalent to 25,000 households (Microsoft, 2026).

Soave suggests the choice is between “we the people” and “we the bureaucrats,” implying that any government involvement is automatically corrupt. This is a false dichotomy. The alternative to thoughtful regulation is not freedom; it is a system where tech companies externalize costs onto communities while capturing all the benefits. When Dominion Energy, the utility serving Northern Virginia’s data center corridor, warns that grid reliability is at risk, and when PJM admits it cannot physically accommodate planned data center loads, we are past the point where “let the market decide” is a responsible answer.

Sanders is not alone in his concerns. Over 230 environmental organizations have called for a federal pause on data center construction until environmental impacts can be assessed. More tellingly, even some tech companies are acknowledging the problem. Microsoft recently announced a “community-first AI infrastructure” plan that includes commitments to not raise local electricity rates, minimize water usage, and create local jobs (Microsoft, 2026). They did this because they recognized that unchecked expansion was generating political backlash that threatened their social license to operate.

The Stanford student who asked Sanders why America innovates more than Europe received a flippant answer from Sanders that Soave rightly criticizes. But Soave’s own answer, that Europe lacks innovation because of taxes and regulation, is equally superficial. The real answer is complex: historical path dependence, venture capital concentration, university-industry relationships, and yes, different policy choices. But acknowledging that policy choices have trade-offs is not the same as pretending Europe has no tech sector, which is simply false.

Soave warns that Sanders wants to “regulate ourselves so that we fall behind enemy nations like China.” This is fear-mongering that substitutes for analysis. The actual constraint on American AI competitiveness is not regulation. It is an electrical grid that cannot handle the load. It is transformer shortages with 30 percent supply gaps and two-to-three-year delivery times (Wood Mackenzie, 2025). It is interconnection queues stretching for years. These are physical bottlenecks that no amount of deregulation can resolve overnight.

Sanders’ call for a moratorium on new data centers is not “nutty.” It is a recognition that building infrastructure at a pace that exceeds the grid’s capacity to support it is not innovation. It is speculation that imposes costs on everyone else. The question is not whether we should have AI. The question is who pays for the electricity, who gets the jobs, and whether the benefits flow to the 1 percent or to society broadly.

Soave thinks slowing down is the same as losing. But sometimes slowing down is how you avoid driving off a cliff. When your grid operators are warning of blackouts, when your utilities are raising rates, and when your infrastructure cannot keep pace with demand, the prudent thing is to pause, assess, and plan. That is not anti-technology. That is basic risk management.

The irony is that Soave, who writes for a magazine called Reason, is asking us to ignore the rational concerns of grid operators, utility regulators, and energy economists in favor of a ideological commitment to speed at any cost. Sanders is the one being practical. He is saying: let us make sure the grid can handle the load, let us make sure ordinary people are not subsidizing billionaire pet projects, and let us make sure displaced workers have a safety net. These are not radical ideas. They are common sense.

Recommendations for Soave

Soave should try reading something other than his own ideology. He might learn that ASML, the company that makes AI possible, is European. He might learn that the U.S. grid is at physical limits. He might learn that China’s advantage is not lack of regulation but massive infrastructure investment. And he might learn that being pro-innovation does not require being blind to costs.

If you would like to watch a man argue with the laws of thermodynamics and lose, Soave’s original ode to electrical fantasy can be found here: https://thehill.com/opinion/robbys-radar/5751172-sanders-silicon-valley-tour/

---

Works Cited

International Energy Agency. (2025). *Energy and AI* (World Energy Outlook Special Report). https://www.iea.org/reports/energy-and-ai

Lawrence Berkeley National Laboratory. (2024). *United States data center energy usage report*. https://eta.lbl.gov/publications/united-states-data-center-energy

Microsoft. (2026, January 16). Microsoft announces community-first AI infrastructure commitments. *Microsoft News Center*. https://news.microsoft.com/2026/01/16/microsoft-announces-community-first-ai-infrastructure-commitments/

Monitoring Analytics. (2026). *State of the market report for PJM*. Federal Energy Regulatory Commission. https://www.monitoringanalytics.com/reports/PJM_State_of_the_Market/2026.shtml

Reuters. (2025, March 15). AI data center surge strains U.S. power grid, sparks reliability concerns. https://www.reuters.com/technology/ai-data-center-surge-strains-us-power-grid-2025-03-15/

U.S. Energy Information Administration. (2025). *Short-term energy outlook*. https://www.eia.gov/outlooks/steo/

Wood Mackenzie. (2025, August). *North America power transformer market outlook*. https://www.woodmac.com/research/power-and-renewables/north-america-power-transformer-market-outlook/