How Do Large Loads Affect Grid Reliability? – What NERC Is Watching
Load used to be the easy part of running a grid. Millions of homes and businesses added up to a smooth, predictable curve. A single data center campus drawing hundreds of megawatts is different. It can disappear in a fraction of a second, swing quickly with computing work and grow faster than new supply can be built. That puts large loads' grid reliability effects near the top of every planner's list. So how do large loads affect grid reliability, and what is the North American Electric Reliability Corporation (NERC) watching?
Large loads affect grid reliability in two main ways: they can strain supply and transmission planning by growing faster than new resources, and they can disturb real-time operations by dropping offline together during voltage dips or swinging their power use quickly. NERC has documented events where a single transmission fault caused hundreds or even thousands of megawatts of data center load to switch to backup power at once.
In response, NERC formed a task force, issued its strongest category of alert in 2026 and is working with the Federal Energy Regulatory Commission (FERC) on how to bring large loads under reliability rules.
Why Are Large Loads a Grid Reliability Issue Now?
Large loads have become a grid reliability issue because of their size, their concentration and their electronics. A single campus can draw as much as a mid-sized city, and many campuses cluster in the same few areas, connected to the same transmission lines.
Inside, almost everything sits behind power electronics: uninterruptible power supplies (UPS), rectifiers and server power supplies. Those devices protect the servers well. But they also make the load respond to grid disturbances in ways traditional motors and lighting never did, and often all at the same moment.
Growth adds to the pressure. NERC's September 2026 Reliability Insights report noted that its 2025 Long-Term Reliability Assessment projected more than 224 GW of summer peak demand growth over ten years. Supply, transmission and planning models all have to catch up.
What Happens When Data Centers Drop Off the Grid at Once?
When many data centers drop off at once, the grid suddenly has more generation than load, and frequency and voltage spike upward until operators and plant controls rebalance. The load doesn't stop computing. It switches to backup power. From the grid's point of view, though, hundreds of megawatts simply vanish.
NERC's incident review, published in January 2025, described a July 10, 2024 event in which a fault on a 230 kV transmission line caused about 1,500 MW of data center load to transfer to backup power. The fault itself was ordinary. The response was not. Its later Reliability Insights report described more events in 2026, including a 500 MW event in April and a load reduction of about 4 GW in Virginia in July.
To see why that matters, take an illustrative area serving 15,000 MW of total load. Losing 1,500 MW is 1,500 ÷ 15,000 = 10% of demand gone in seconds. Generators have to cut output quickly to avoid overvoltage and overfrequency, and if the load comes back all at once minutes later, they have to ramp up just as fast.
Why Do Data Centers Disconnect During Voltage Dips?
Data centers disconnect during voltage dips because their protection is set to keep servers safe, not to support the grid. A UPS watches incoming voltage closely. If it sags below a set threshold for longer than a set time, the UPS switches to battery and the facility may start its generators.
That's good design for the facility in isolation. The problem is scale and sameness. When dozens of sites with similar settings see the same dip from the same fault, they all transfer together. Then they return to grid power on their own timers, which can create a second disturbance.
The fix is ride-through: settings and equipment that let a facility stay connected through brief, shallow dips while still protecting servers. Generators have long had ride-through requirements. Large loads mostly haven't, and that's changing.
How Do Large Loads Affect Voltage on Weak Parts of the Grid?
Large loads can pull voltage down on weaker parts of the grid, especially at sites far from big generators or strong substations. Voltage support depends on reactive power, which nearby generators, capacitor banks and other equipment provide. A remote 300 MW load at the end of a long line can leave little margin before voltage sags during a contingency.
That's why load studies look at more than whether lines overheat. Planners check how voltage behaves with the new load during outages, and they may require capacitor banks, static VAR compensators or other voltage support equipment as a condition of connection. Sites in remote areas, chosen for cheap land or nearby gas, often run into this first.
Can Fast Load Swings Hurt the Grid?
Fast load swings can hurt the grid if they're large, repeated and land at frequencies that excite oscillations in nearby generators or controls. Large AI training jobs are the prime example. A 2025 Microsoft Research paper by authors from Microsoft, OpenAI and NVIDIA described how synchronized training swings between computation and communication phases, causing large power oscillations, and proposed software, hardware and storage-based fixes.
Grid operators are writing limits. ERCOT, for example, presented a draft rule in September 2026 that would cap large-load power variation at 10 MW per 5 seconds. Batteries on site can smooth these swings before they reach the grid, which is one reason storage keeps showing up in large data center designs.
Is There Enough Supply for All the New Large Loads?
Not everywhere, and not on the current schedule. Resource adequacy, meaning enough generation and transmission to meet demand with a safety margin, is the slower-moving side of large loads grid reliability. When demand forecasts jump by tens of gigawatts in a few years, building supply to match is hard, given turbine lead times, interconnection queues and transmission timelines.
Markets show the strain. PJM's capacity auction for 2027/2028 cleared at its price cap and still fell short of its reliability requirement, with most forecast load growth tied to data centers. Other regions face similar math, even if it shows up through utility plans rather than auction prices.
What Is NERC Doing About Large Loads?
NERC is moving from study to action. The main steps so far:
- In 2024 NERC formed a Large Loads Task Force to study how data centers and other emerging loads affect reliability.
- It published incident reviews of large load loss events, starting with the July 2024 event.
- In May 2026 it issued a Level 3 Essential Action alert, its highest alert category, with seven required actions covering modeling, commissioning and protection of computational loads, according to Renewable Energy World.
- Its September 2026 report noted that FERC wants criteria for registering large loads under reliability rules by the end of 2026.
Registration is the big one. NERC standards traditionally apply to generators, transmission owners and operators, not to customers. Bringing large loads into that framework would mean enforceable requirements for ride-through, modeling and data sharing.
How Are Grid Operators Responding?
Grid operators are adding reliability requirements to their large load connection processes. Common elements include detailed equipment models, ride-through performance, ramp-rate limits, commissioning tests and the ability to curtail during emergencies.
The Southwest Power Pool, for example, now has a dedicated process for SPP large load interconnection that FERC accepted in early 2026. ERCOT has moved on ramp limits and emergency curtailment, backed by Texas Senate Bill 6. The specifics differ, but the direction is the same: large loads must behave predictably on the grid.
What Can Data Center Operators Do?
Data center operators can reduce their grid reliability footprint with design and settings choices that cost little compared with a campus:
- Coordinate UPS and protection settings with the utility so the site rides through brief dips.
- Stagger and control the return to grid power after a transfer.
- Provide accurate dynamic models of the facility's electrical behavior.
- Use batteries or controls to smooth fast power swings.
- Commission and test grid behavior before full energization, not after an event.
Expect these to move from best practice to requirement. Work with your utility and qualified power systems engineers, since protection settings involve real trade-offs between equipment safety and grid support.
Large Loads and Grid Reliability Are Now One Conversation
So how do large loads affect grid reliability? They can outgrow supply and transmission planning, and they can disturb the grid by dropping offline together or swinging their power use quickly. NERC has documented those events, issued its strongest alert and is working with FERC on bringing large loads under enforceable standards.
The good news is that most of these problems have engineering solutions: ride-through settings, coordinated reconnection, accurate models and on-site storage. If you're building or operating a large load, treat grid behavior as a design requirement from day one, because regulators increasingly will.