Transmission system operators (TSOs) in several regions are preparing rules to address what they see as an emerging threat: simultaneous actions by large data centers that can destabilize the grid. The new grid codes will require data centers to ride out transient variations in the supply. This follows a series of incidents where data centers, behaving in unison, nearly caused blackouts.

The problem occurs when a grid power disturbance, such as a voltage change, prompts many facilities to switch simultaneously to on-site batteries and/or generators. On grids where data centers consume a large share of electric power, this sudden drop in demand can destabilize the system. When the fault is cleared, a secondary problem can occur: as data centers return to grid power to recharge batteries alongside their normal demand, they can cause a sharp increase in the overall load.

Uptime Institute is aware of at least six TSOs in the US and Europe that plan to issue new codes requiring large loads connected to the transmission network to remain on the grid during fluctuations in voltage and frequency, riding through the event until the fault is cleared, rather than shifting to on-site power.

TSOs already routinely require generating sources — such as wind farms, solar plants and battery energy storage systems — to ride through faults lasting up to several seconds.

The new rules are designed to address grid stability and are in addition to recently announced measures by some TSOs on curtailing demand (such as requiring on-site power generation during shortfalls of generation). Meeting these new requirements could be costly, however, as many operators will have to redesign and reconfigure their power systems, add equipment and revise operating procedures.

Two near misses linked to data centers

The two incidents in which simultaneous action by data centers nearly destabilized an electrical grid occurred in grid regions with a high concentration of facilities. In each case, several data centers automatically dropped off the grid in response to fluctuations, causing further distortion.

The TSOs responded quickly before other generating and demand units tripped off the net for their own safety, which could have led to a blackout.

• Northern Virginia. On July 10, 2024, a failed lightning arrestor caused a series of short (around 50 ms) fluctuations, according to NERC (North American Electric Reliability Corporation). In response, 60 data centers — totaling 1,500 MW — simultaneously dropped off the transmission grid. This group represents a substantial portion of the region’s 300-plus data centers, which account for around 26% of electricity demand. Dominion Power and PJM Interconnection controlled the situation, preventing a blackout.
• Ireland. On December 13, 2022, a dip in voltage on the 220 kV transmission grid prompted 200 MW of data center capacity to switch to battery power before swiftly returning to the grid. The country hosts over 82 data centers, consuming more than 20% of the grid power demand. The TSO EirGrid controlled the resulting frequency rise, avoiding a potential region-wide blackout.

Uptime Intelligence believes that other near-miss incidents may have occurred.

TSOs maintain frequency by adding or removing sources and loads, while reactive compensators prevent changes in voltage. If a distortion overwhelms these measures, it can cause a local, isolated power loss on a sub-grid or cascade through an entire grid region.

On April 28, 2025, such a cascade caused national blackouts in Spain and Portugal — however, it is important to stress that data centers are not believed to have been the cause of this incident.

TSOs will demand fault ride through

The issue was reported to the International Energy Agency in October 2024. Now, multiple TSOs are planning new grid codes that will require data centers and other large loads to remain connected to the transmission grid unless fluctuations exceed certain thresholds. These facilities must also reconnect in a staggered fashion.

Fingrid (Finland), EirGrid (Ireland), Energinet (Denmark), RTE (France), AEMO (Australia), ERCOT (Texas) and ENTSO-E (Europe) are preparing requirements — including fault ride through (FRT) and post fault active power recovery (PFAPR) — to govern how loads return to the grid. ENSO-E, RTE, ERCOT, EirGrid and Energinet have already issued proposals.

Despite the near miss in Virginia, PJM and Dominion have not yet issued any comments on FRT requirements for loads, although they do require generators to ride through.

Figure 1, prepared by electrical infrastructure equipment provider Eaton, compares the available FRT proposals with the established “ITIC curve” — the accepted AC voltage boundary that most information technology equipment can tolerate or ride through without malfunctioning.

Figure 1 Proposed fault ride through requirements (source: Eaton)

Note, this curve, established in the 1970s, is maintained by the Information Technology Industry Council (ITIC) and is sometimes referred to by the name of ITIC’s predecessor, CBEMA.

Operators will have to modify UPS provision

Data center operators will have to modify their power infrastructure to comply with FRT requirements — and this could be costly. Uptime Intelligence expects these requirements will only apply to large facilities connected to transmission networks. Smaller facilities and those connected to distribution networks will not be affected. Depending on how TSOs apply the rules, some data centers may require upgrades.

Operators may also need to protect mechanical and electrical (M&E) equipment, which is susceptible to grid fluctuations but is not normally exposed to them.

During a grid fluctuation, the entire facility normally trips offline. In most facilities, the UPS system provides battery power only to the IT equipment, while the M&E equipment receives no power until generators switch on or the facility returns to the grid. Some facilities, including all Tier IV facilities, power their cooling with a UPS system, but most lower-tier, air-cooled data centers are designed to tolerate a brief pause in cooling when the facility switches to on-site power.

If a data center is required to ride through a fluctuation, both the IT and the M&E equipment must be protected from transient changes in voltage. Operators should either:

• Increase UPS capacity to provide power conditioning for both M&E and IT equipment, or
• Install a power conditioner (a double-conversion UPS system without batteries) upstream of the data center, sized to match the total facility load (IT plus M&E).

Operators may have to reconfigure UPS systems with more granularity so thresholds vary between different parts of the facility. Rules from the TSO will likely manage dropping off the grid.

These changes will have significant implications for colocation contracts. Some TSOs may demand coordinated manual connections and disconnection procedures.

Other related reports published by Uptime Institute include: 
Data centers weather grid failures — but utilities want change 
Grid demand will require active participation from data centers 
Data center operators will face more grid disturbances