ERCOT is raising alarms over a new Texas grid reliability risk: hyperscale data center and crypto loads that may abruptly disconnect during ordinary voltage disturbances. In recent testing, four groups of proposed large customers failed voltage ride-through requirements, and models showed each group could drop more than 5,000 MW of demand in a single event.
That scale matters. A sudden loss of load equivalent to a major U.S. city can push grid frequency higher, disrupt generators, and turn a localized disturbance into a broader system event. For investors tracking utilities, independent power producers, data center developers, and grid equipment suppliers, the issue is no longer theoretical.
ERCOT has already recorded at least 26 disconnection events involving data centers or crypto operations since 2023. With roughly 20 GW of large-customer applications under review and several gigawatts expected to energize before July, the Texas market is becoming a test case for how AI-driven power demand collides with grid physics.
Key Facts
- ERCOT disclosed on May 21 that four clusters of proposed data centers and crypto facilities failed voltage ride-through testing.
- Each modeled disconnection event could remove more than 5,000 MW of demand from the Texas grid at once.
- ERCOT has logged at least 26 disconnection events tied to data centers or crypto operations since 2023.
- The grid operator is reviewing about 20 GW of large-load applications, including several gigawatts planned to come online before July.
- The Iberian blackout on April 28, 2025 intensified attention on voltage control, frequency response, and the grid effects of inverter-based resources and sensitive electronic loads.
ERCOT data center load risk
The core issue is counterintuitive: large electricity users are usually viewed as demand sources, not grid-stability threats. But hyperscale data centers and crypto facilities rely on sensitive power electronics and protection systems designed to shield servers and mining equipment from abnormal voltage conditions. If those systems trip offline during faults, capacitor switching, or equipment disturbances, the grid can suddenly lose thousands of megawatts of demand.
That abrupt drop creates a generation surplus. Frequency can rise quickly, leaving power plants and other assets exposed to over-frequency conditions or forced operating changes. In a system with tight reserves, heavy summer demand, or limited operational flexibility, the shock can propagate well beyond the original disturbance. The risk is especially relevant in Texas, where load growth from AI infrastructure, industrial expansion, and electrification is arriving faster than many networks were designed to absorb.
Who is affected extends well beyond utilities. Data center developers may face stricter interconnection standards, additional engineering costs, and longer timelines before energization. Power producers with flexible, dispatchable fleets could see stronger demand signals and improved capacity economics. Transmission owners, substation equipment makers, and suppliers of voltage control, protection, and reactive power technologies may also benefit as grid operators seek to harden systems against both supply-side and demand-side instability.
The new challenge for Texas is not just how to serve massive data center demand, but how to keep that demand from becoming a grid shock when voltage conditions move outside narrow tolerances.
Why the comparison to Spain matters
The April 28, 2025 blackout in Spain and Portugal sharpened focus on how voltage and reactive power problems can cascade across modern power systems. That event highlighted the limits of simplistic explanations and instead underscored a broader engineering problem: when generators and large electronic loads do not respond to disturbances in ways the grid expects, routine faults can escalate quickly.
ERCOT’s concern is the mirror image of that episode. In Iberia, the discussion centered on the supply side, including generator responses and dynamic voltage support. In Texas, the current warning emphasizes the demand side, where hyperscale loads may disconnect instead of riding through a disturbance. In both cases, the market lesson is similar: fast-changing grids need more than nameplate megawatts. They need inertia, frequency response, reactive power support, and operational resilience.
Implications for Investors
For investors, ERCOT’s findings reinforce a major theme in U.S. power markets: not all megawatts are equal. As AI-related power demand expands, assets that provide flexible generation, voltage support, and rapid balancing capability may become more valuable. Companies with exposure to natural gas peakers, combined-cycle plants, nuclear generation, grid-forming inverter technologies, synchronous condensers, transformers, and advanced protection systems could benefit from tighter reliability standards and higher infrastructure spending.
There are also clear risks. Data center developers and landlords may encounter higher capital expenditures as interconnection studies increasingly examine voltage ride-through performance, on-site backup integration, and power quality controls. Projects once modeled on simple access to large power blocks may require more bespoke engineering, potentially affecting return assumptions, energization schedules, and contract pricing. Utilities serving fast-growing digital infrastructure corridors could face political and regulatory pressure if reliability concerns rise alongside customer bills.
Investors should also watch the policy response. If grid operators conclude that large electronic loads need revised protection settings, mandatory ride-through standards, or dedicated mitigation equipment, that would reshape project economics across the AI infrastructure chain. At the same time, the episode strengthens the case for retaining and adding dispatchable resources where reliability margins are thinning. In markets already confronting reserve constraints, that may support improved revenue visibility for certain thermal and nuclear assets even as renewable deployment continues.
The next phase will be defined by interconnection decisions, summer reliability conditions, and how quickly large-load customers adapt their designs to meet grid needs. Texas is likely to remain the leading indicator for whether AI-era electricity demand becomes a growth engine for infrastructure investment or a stress test for market stability.
VIP Algorithmic Setups
