The largest power grid in the United States, operated by PJM Interconnection, is considering a controversial measure: temporarily cutting power to data centers during peak demand to prevent widespread blackouts. According to a recent report from TechCrunch (July 28, 2026), PJM is exploring rules that would allow it to curtail electricity supply to data centers — facilities that have become some of the most energy-intensive consumers on the grid. This move highlights the growing tension between the explosion of AI-driven computing and the aging infrastructure of America’s electricity network.
What Is PJM and Why Does This Matter?
PJM Interconnection manages the high-voltage electricity grid serving 13 states (including Illinois, Ohio, Pennsylvania, New Jersey, Virginia) and the District of Columbia — home to more than 65 million people. It is the largest competitive wholesale electricity market in the world. As demand for cloud computing, cryptocurrency mining, and especially AI training surges, data centers have become a primary driver of new load growth in PJM’s territory. The grid operator estimates that data center capacity requests have more than doubled in the past two years, putting unprecedented strain on transmission and generation resources.
The reported proposal would classify data centers as “interruptible load” — meaning they could be required to reduce consumption, potentially to zero, during emergency grid conditions. This is not a theoretical exercise: PJM has already faced repeated near-miss events where reserve margins dipped dangerously low. In summer 2025, the grid came within minutes of forced rolling blackouts during a heat wave.
How the Power Cut Mechanism Would Work
According to the TechCrunch report, the new rules (still under review by PJM’s stakeholder committee) would allow the grid operator to request or mandate load reductions from large data centers — those drawing 25 megawatts or more — when the system faces a capacity shortage. In exchange, data centers would receive either lower electricity rates (similar to existing demand-response programs) or direct compensation. However, the key difference is that participation could shift from voluntary to mandatory during emergencies.
Data centers would be expected to have backup power (diesel generators, battery storage) to carry critical loads during the interruption. The proposal discusses “fast-acting curtailment” — reducing power within minutes — which is feasible for facilities with adequate battery reserves. Non-critical computing tasks, such as batch processing or model training, could be paused, while essential services (like emergency response systems) would remain powered.
| Scenario | Current Practice | Proposed Rule Change |
|---|---|---|
| Grid stress (e.g., heatwave) | Voluntary load reduction with incentives | Mandatory curtailment for data centers >25 MW |
| Emergency (reserve margin < 1.5%) | Rolling blackouts for all customers | Targeted data center cuts to avoid wider outages |
| Compensation | Market-based demand response payments | Fixed tariff or capacity payment for curtailment right |
Real-World Precedents and Challenges
Temporary power cuts for large industrial users are not new. PJM already runs a Demand Response program that pays factories and commercial buildings to reduce usage. But data centers present unique challenges. Uptime is sacred in the industry: SLAs often guarantee 99.999% availability. Forcing a cut could violate contracts and cause financial penalties or even permanent data loss.
However, the grid operator argues that data centers are often better equipped to handle short outages than hospitals or residential neighborhoods. Many modern facilities have on-site backup power that can run for hours. The proposal would require data centers to demonstrate at least 15 minutes of battery backup before they can qualify for the reduced rate — enough time for generators to start or for the grid to stabilize.
A case in point: in Northern Virginia — the world’s largest data center market — Dominion Energy already has agreements with some hyperscalers to curtail load during peak events. Google, for instance, has publicly shared that it uses battery storage at its data centers to shift power consumption and participate in demand response programs. The PJM proposal would formalize and expand this practice across the entire footprint.
Impact on Data Center Operators and the Cloud Industry
If implemented, the rule could reshape how data centers are designed and operated. Key implications include:
- Battery storage becomes essential: Facilities without substantial on-site battery backup may face higher rates or be unable to sign up new customers in PJM territory.
- Geographic diversification: Companies may accelerate building data centers in regions with more stable or cheaper power (e.g., MISO, ERCOT) to reduce exposure to PJM interruptions.
- Operational flexibility: Training workloads that are interruptible by nature (e.g., batch AI model training) could be scheduled around grid alerts, while latency-sensitive apps (streaming, gaming) may need redundant routing.
- Financial models: Data center operators might offer two tiers of service: “guaranteed power” (higher cost) and “interruptible” (lower cost) — similar to how cloud providers offer spot instances.
The TechCrunch report quotes industry analysts warning that the move could slow data center buildout in PJM states, potentially affecting tech hubs like Ashburn, Virginia. However, grid reliability advocates see it as a necessary step to avoid catastrophic blackouts that would affect everyone, including data centers themselves.
Grid Reliability as a Shared Responsibility
The fundamental issue is that data centers are no longer a negligible fraction of electricity demand. In some PJM zones, they account for over 30% of new load growth. Building new transmission lines takes a decade or more, and new power plants face regulatory hurdles. Therefore, managing demand — including temporary cuts to large users — becomes a practical tool for maintaining grid stability.
The PJM proposal is still in stakeholder review; it has not been approved by the Federal Energy Regulatory Commission (FERC). But the direction is clear: data centers will be expected to be more active participants in grid reliability, not just passive consumers. The era of guaranteed, uninterruptible power for every server rack may be ending in the most congested areas.
Conclusion
Data centers have become the backbone of the digital economy, but their energy appetite is testing the limits of the largest US grid. PJM’s potential rule to temporarily cut power to these facilities represents a significant shift: from viewing data centers as critical infrastructure to seeing them as flexible resources that can be tapped during emergencies. For operators, the message is to invest in backup power, load management software, and operational agility. For the rest of us, it’s a reminder that the cloud runs on a physical grid that has real constraints. The path forward lies in smarter energy use — and perhaps a bit of old-fashioned sacrifice during peak hours.
Comments