Both True · edition
AI-powered data centers are at the heart of modern innovation, but their appetite for energy is pushing the U.S. electrical grid to its limits. Today, we’re looking at PJM Interconnection’s decision to manage this strain with temporary power cuts.
This isn’t just a story about power. It’s about trade-offs: how far we’ll stretch infrastructure for AI and who bears the cost when the grid buckles under pressure. Let’s weigh it.
The lead
STRAIGHT TALK
PJM Interconnection, the largest U.S. electrical grid operator, will begin cutting power to large data centers during shortages starting in June 2027. Data centers, projected to quadruple their energy use by 2035, are being blamed for spiking wholesale electricity prices. Compensated under demand response programs, affected data centers may rely on diesel generators, raising pollution concerns. (Source: TechCrunch) →
The case for
The case for temporary power cuts to data centers hinges on grid resilience and fairness. PJM Interconnection serves 67 million customers, and blackouts would have catastrophic impacts on households, hospitals, and businesses. By curtailing power to 50-megawatt-plus data centers during peak demand, PJM can prioritize essential services while maintaining system stability.
This move also incentivizes innovation. Data centers—critical to AI, cloud computing, and digital infrastructure—will likely accelerate investments in energy efficiency, renewable on-site power, and smarter load management as a result. Demand response programs, which compensate facilities for reducing usage during peak periods, have been successful in other sectors like manufacturing. By extending this model to data centers, PJM creates a market-driven mechanism that aligns financial incentives with grid health. If managed well, the cuts could steer the industry toward a more sustainable energy future without stifling AI innovation outright.
The cost
The cost of this strategy is layered. First, the near-term reliance on diesel generators introduces significant environmental and public health risks. Diesel emissions can harm nearby communities, and Vantage Data Centers has already faced backlash for attempting to downplay these impacts. Tens of millions in potential health damages are not theoretical—they’re lived realities for people near these facilities.
Second, wholesale electricity prices in PJM’s territory have nearly doubled in the past year, with data centers taking much of the blame. These costs ripple through the economy, affecting consumers and small businesses alike. Finally, the infrastructure gap remains a glaring issue. PJM’s recent auction for new generating capacity fell short, and while another auction is planned, the timeline is tight. If these gaps aren’t bridged, temporary cuts could become a long-term crutch, undermining trust in the grid’s reliability and fairness.
Terms, plainly
Context
PJM Interconnection’s struggles reflect a broader trend: energy infrastructure lagging behind the rapid growth of AI and cloud computing. Data centers have doubled their electricity use globally over the past decade, and the push for generative AI only accelerates this demand. Other regions facing similar challenges are experimenting with renewable microgrids, virtual power plants, and stricter energy efficiency mandates for new facilities.
In the U.S., regulatory frameworks around diesel emissions and demand response programs will play a critical role. Watch for how quickly PJM can add new generating capacity and whether federal or state incentives emerge to support greener data center operations. The stakes are high, not just for the grid but for the communities and industries that depend on it.
Both true
PJM’s plan is a stopgap solution to a sprawling problem. Yes, it buys time for the grid, but at the expense of local air quality and rising costs for everyone. AI’s energy hunger is undeniable, and the gap between our digital aspirations and physical infrastructure is widening. Both true. The long-term fix? Smarter data centers and a smarter grid. For now, we’re running on borrowed time—and borrowed fumes.
Also today
TECH & PLANET
The Energy Department will convert a shuttered Cold War-era uranium enrichment facility into a data center campus and gas plants. This repurposing reflects a growing push to meet AI's infrastructure demands. (Source: The New York Times) →
Why it mattersRepurposing old facilities for AI aligns with resource efficiency, but raises questions about energy priorities.
FRONTIER
An AI system successfully planned and adapted observations using the 570-megapixel Dark Energy Camera in Chile. This marks the first use of AI to autonomously schedule a national telescope’s observations. (Source: Northwestern Now News) →
Why it mattersAI’s role in astronomy signals a leap in how we explore and understand the universe.
HUMAN IMPACT
A healthcare workers' union alleges Kaiser Permanente uses an automated algorithm to screen mental health patients without clinician oversight, violating state law. Kaiser skipped a public hearing on the complaint for the second year. (Source: CalMatters) →
Why it mattersWhen algorithms replace clinicians, ethical and legal questions about patient care take center stage.
My analysis
The bigger picture
AI’s hunger for energy, data, and decision-making power is reshaping industries and infrastructures alike. From the grid to telescopes to healthcare, the strain and the promise are everywhere.
What it means for you
AI is forcing us to rethink the systems that power our world. What breaks first?