AI Isn't Always the Priority: DOE Limits Data Center Power | Derui Electric
For years, data centers have been treated as one of the power grid's highest-priority customers.
They consume enormous amounts of electricity, operate around the clock, and support everything from cloud computing to artificial intelligence.
But during the July 2026 heatwave, something remarkable happened.
The U.S. Department of Energy (DOE) issued an emergency order allowing PJM Interconnection to temporarily reduce electricity supplied to certain large electricity users—including eligible data centers—to protect the stability of the power grid.
It wasn't because AI demand had disappeared. It was because the grid couldn't safely serve everyone at once.
For perhaps the first time, one of the world's fastest-growing industries was asked to step aside so hospitals, emergency services, and millions of households could keep the lights on.
That decision signals something much bigger than one heatwave. It signals that grid capacity is becoming a strategic constraint on AI growth.
Key Takeaways
- ✓ The U.S. DOE authorized PJM to reduce electricity supplied to certain large electricity users during emergency grid conditions—three times in less than three weeks.
- ✓ The decision reflects growing pressure from AI data centers, electrification, and extreme weather.
- ✓ Grid reliability is becoming a limiting factor for AI expansion.
- ✓ Future investment will increasingly focus on substations, transformers, and distribution infrastructure—not just new generation.
- ✓ Utilities that strengthen their networks today will be better prepared for tomorrow's demand.
PJM powers 67 million people across 13 states.
During the July heatwave, PJM forecast peak load could top 166,000 MW—above the 2006 record. The strain pushed the grid to its limit, forcing the DOE to act.
What Happened?
During late June and early July, an extreme heatwave swept across much of the eastern United States. Air-conditioning demand surged. Electricity consumption reached levels rarely seen outside winter storms.
At the same time, several power plants were unavailable because of maintenance and unexpected outages.
PJM—the largest regional transmission organization in North America—forecast demand approaching historic records. On Thursday, July 2, demand surged to roughly 163 GW. PJM had forecast the day's load could top 166,000 MW, which would have broken the 2006 record of 165,563 MW.
To prevent wider instability, the DOE issued an emergency authorization allowing PJM to implement extraordinary operating measures. Among them was the ability to curtail electricity consumption from certain large industrial and data center customers when necessary to maintain system reliability.
The order was issued not once, but three times in less than three weeks. The first, on June 30, lasted through July 3. The second, from July 14 through July 21. The third, from July 15 through July 21.
The objective wasn't to punish AI companies. It was to protect the grid.
The DOE Order: What It Actually Did
Signed by U.S. Secretary of Energy Chris Wright, the order allowed PJM to tap into diesel backup generators, battery arrays, and other onsite power sources as a "last resort" to prevent power shut-offs.
Specifically, the order applied to any customer with at least 50 MW of peak load. Within 15 minutes of an emergency signal, they could be directed to switch to onsite backup power, freeing capacity for homes and hospitals.
"Currently, there are tens of gigawatts of readily available backup generation that have remained largely untapped. Deployment of backup generation resources … can prevent avoidable blackouts, thereby saving lives and reducing costs to the American people."
— Secretary Chris Wright, DOE
The DOE estimates more than 35 GW of backup generation remains unused across the United States—enough to power roughly 26 million homes. The order also waived environmental permitting restrictions for certain power generators to maintain grid reliability.
PJM's Chief Operating Officer Frederick S. Bresler warned that without such measures, "there is a risk that residences, hospitals, military facilities, water treatment plants, and other critical facilities may lose electric service".
Why This Matters
For decades, utilities have managed emergencies by increasing supply. Start another generator. Import electricity. Activate reserve capacity.
But today's power system is changing.
AI data centers can require 50 MW, 100 MW, or even several hundred megawatts at a single location. One hyperscale campus can consume as much electricity as a medium-sized city.
When demand grows faster than generation and transmission infrastructure, utilities eventually face a difficult choice. They can no longer increase supply fast enough. Instead, they must temporarily reduce demand.
That is exactly what happened.
The cost of inaction is staggering. PJM's capacity auction for the 2026/2027 delivery year cleared at $329.17 per megawatt-day—up from just $28.92 two years earlier, a roughly tenfold increase. Day-ahead power contracts topped $2,000 per megawatt-hour in parts of the system, while the Western Hub benchmark settled at $1,222.75/MWh—nearly triple where it stood during a comparable peak the previous summer. In some areas, spot wholesale electricity prices surged beyond $2,500 per MWh.
PJM projects peak demand will grow by 32 GW between 2024 and 2030—and according to the grid operator's own forecasts, all but 2 GW of that increase will come from data centers.
A New Reality for AI Infrastructure
Only two years ago, the biggest question was: Can utilities supply enough electricity for AI?
Today, another question is emerging. What happens when they can't?
Governments are beginning to answer that question. New York paused permits for large AI data centers. Australia announced that hyperscale facilities must help finance new power generation. Now the DOE has demonstrated that, during emergencies, even data centers may be asked to reduce consumption to keep the wider grid operating.
Different policies. Same direction.
Electricity is becoming the limiting factor in AI expansion.
Grid equipment is now the frontline of AI reliability.
transformers, switchgear, and substations must handle faster load swings, more frequent switching, and tighter voltage tolerance than ever before.
What This Means for Utilities and EPC Contractors
The message is becoming increasingly clear. Future grid planning is no longer just about adding more transformers or building larger substations. It is about creating enough flexibility to absorb rapidly growing AI demand while maintaining reliability for every other customer.
Utilities will need:
- Faster substation expansion
- Higher-capacity transformers
- Smarter distribution automation
- More resilient switchgear
- Better demand-side management
Grid infrastructure is becoming just as critical as computing infrastructure.
What This Means for Equipment Manufacturers
For transformer and switchgear manufacturers, the implication is equally significant. AI investment is not slowing. Grid investment is accelerating. Every new transmission upgrade, substation expansion, and distribution reinforcement ultimately depends on reliable electrical equipment.
The challenge is no longer simply delivering more capacity. It is delivering equipment that utilities can deploy quickly as grid expansion timelines become increasingly compressed.
Derui Electric's Perspective
At Derui Electric, we view the DOE's emergency order as another reminder that the future of AI depends on the future of the grid.
Reliable transformers, switchgear, and distribution equipment are no longer supporting infrastructure—they are becoming strategic assets.
As utilities invest to strengthen networks for AI, electrification, and extreme weather, dependable equipment will play an increasingly important role in keeping power flowing where it is needed most.
Conclusion: The Grid Is Changing Faster Than We Expected
The July 2026 emergency order was never really about limiting AI. It was about protecting the power system that AI depends on.
As electricity demand continues to accelerate, the question facing utilities is no longer whether they can generate enough power. It is whether the grid itself can keep pace.
In the years ahead, the winners won't simply be those building larger data centers. They will be the ones building stronger grids.
How does your current equipment perform under volatile grid conditions?
If you are planning a new substation or upgrading an existing one, the question is no longer just about capacity—it's about response time, voltage stability, and protection coordination.
Need transformers and switchgear designed for tomorrow's grid?
Send us your project specifications. Our engineering team will review your requirements and provide a tailored solution with full IEC certification and loss guarantees.
Get a Technical Consultation →IEC 60076 certified · Full type test reports · Written loss guarantees
Sources: U.S. Department of Energy, PJM Interconnection, Monitoring Analytics, Canary Media. Event occurred June–July 2026.
Frequently Asked Questions (FAQ)
Q1: What did the DOE order actually do?
A: The order authorized PJM to direct large electricity users (50 MW+) to switch to onsite backup generators during emergencies, freeing grid capacity for critical users like hospitals and households.
Q2: Was this a one-time event?
A: No. The order was issued three times in less than three weeks. It is becoming a pattern, not an anomaly.
Q3: Why are data centers being targeted?
A: Data centers are the fastest-growing load on the grid. PJM projects 32 GW of demand growth by 2030—with data centers accounting for 30 GW of that increase.
Q4: What does this mean for transformer and switchgear demand?
A: Grid investment is accelerating. Utilities are upgrading substations, transmission lines, and distribution networks—all of which require transformers, switchgear, and protection systems.
Q5: Will similar policies emerge in other regions?
A: Likely. New York has paused AI data center permits. Australia is requiring hyperscalers to fund new generation. The trend is global.