HN Debrief

Texas grid flags risks as data centers, crypto sites fail voltage tests

  • Infrastructure
  • Energy
  • AI
  • Regulation

The Reuters piece says Texas grid operator ERCOT is warning that some big new power users, especially data centers and crypto mines, failed voltage ride-through tests. The issue is not simply that these sites use huge amounts of electricity. It is that many are designed to protect their own equipment by instantly disconnecting and switching to on-site batteries or generators when the grid wobbles. When hundreds of megawatts or gigawatts disappear at once, generation suddenly has nowhere to go, frequency and voltage can swing, and protective systems can start tripping in a cascade. Because ERCOT is only weakly tied to the rest of the US grid, it has less room to absorb those shocks.

If you run or finance a large load in Texas, interconnection is no longer just a capacity question. Expect regulators and utilities to demand ride-through, load smoothing, or on-site fast-response storage instead of allowing facilities to dump load onto the grid during faults.

Discussion mood

Concerned and annoyed. Most commenters saw the risk as real and fixable, but were frustrated that ERCOT and large loads are pushing coordination and mitigation costs onto the wider grid, especially in a state already viewed as fragile after the 2021 outages.

Key insights

  1. 01

    The danger is synchronized load loss

    What breaks the system is not that a single large facility trips offline. The real hazard is many sites using similar protection logic, seeing the same voltage sag, and disconnecting at nearly the same moment. That turns a manageable event into a cascade because each shutdown worsens the disturbance for the next facility and for grid protection equipment.

    When you assess large-load risk, model correlated trips, not just N-1 style single-facility failures. Shared control settings across many sites can create systemwide failure modes.

      Attribution:
    • manarth #1
    • erincandescent #1
    • Aurornis #1
  2. 02

    Data centers could be part of the fix

    Large modern facilities already sit behind big three-phase converters, UPS systems, and battery banks. Several commenters argued that with the right architecture, those same power electronics can supply reactive power, smooth real-power swings, and even sell fast grid services instead of acting as dumb loads that vanish on faults. That makes data centers a plausible host for mitigation, not just the source of the problem.

    If you are building a new campus, treat grid support as a product requirement. Designing power systems for ride-through and ancillary services up front is likely cheaper than retrofitting after interconnection rules tighten.

      Attribution:
    • namibj #1 #2
    • AnthonyMouse #1
  3. 03

    Grid-forming batteries are already real

    The useful correction here was that inverter-based resources are not stuck as passive followers anymore. Grid-forming inverters and synthetic inertia let batteries and other converter-based systems actively support frequency and voltage during disturbances. That does not make the problem easy, but it means the technical path is known and already deployed in some places.

    Do not plan around a false choice between old spinning machines and unstable inverter fleets. Ask vendors and utilities specifically whether a project includes grid-forming capability and what performance it has under faults.

      Attribution:
    • jillesvangurp #1
    • peterus #1
    • bryanlarsen #1
  4. 04

    An external master clock would not solve it

    Several people floated GPS, TAI, or a separate pilot signal as a way to force everything to the same phase. The implicit rebuttal across the technical comments is that grid stability is not mainly a timekeeping problem. It is a power-flow and control problem inside a coupled physical system, where voltage, phase, and frequency must stay coherent under disturbances and line constraints. A perfect clock does not remove the need for inertia, damping, or coordination.

    Be skeptical of control ideas that sound like synchronization in software. Grid behavior is constrained by physics first, then communications and control.

      Attribution:
    • digitalsushi #1
    • chaz6 #1
    • tenthirtyam #1
  5. 05

    Texas trades faster buildout for messier operations

    ERCOT's "connect and manage" approach came through as a deliberate policy choice. It helps Texas add generation, batteries, and large loads faster than many other regions, but it also means more operational complexity lands on the grid operator later. That is tolerable until summer peaks, weather stress, or clustered data center growth expose the thin margin for mistakes.

    If you want projects energized quickly in Texas, price in future compliance churn. Fast interconnection does not mean stable long-term operating rules.

      Attribution:
    • klysm #1
    • Schiendelman #1
    • brycewray #1

Against the grain

  1. 01

    Big data centers can help renewable-heavy grids

    Instead of treating large data centers as pure parasites, this view sees them as unusually flexible infrastructure. Because they already carry UPS batteries and backup generation, they can shift load, charge or discharge storage, and even run their own generation when prices spike. That can make a renewable-heavy grid easier to manage if operators are required to participate correctly.

    Do not assume every large load is net-negative for grid reliability. The commercial model can be structured so these sites provide resilience services the grid already needs.

      Attribution:
    • AnthonyMouse #1
    • jeffbee #1
  2. 02

    Grid expansion itself is overdue and useful

    Some pushed back on the instinct to blame data centers for every transmission upgrade. The broader point is that many US regions underbuilt generation and transmission for years, and new industrial demand may finally force investment that should have happened anyway. Even if AI demand cools later, the system benefits from more interconnection and more headroom.

    Separate legitimate complaints about cost shifting from blanket opposition to new power infrastructure. Some of the buildout triggered by data centers may still be worth having after the boom passes.

      Attribution:
    • phil21 #1
    • bob1029 #1
    • jeffbee #1

In plain english

ERCOT ↩
Electric Reliability Council of Texas, the operator of most of the Texas electric grid and power market.
frequency ↩
The rate at which alternating current oscillates on the grid, 60 hertz in most of the United States.
inertia ↩
In power systems, the stabilizing effect of spinning machines that resist sudden changes in grid frequency.
phase ↩
The timing relationship between alternating current waveforms on different parts of the grid.
reactive power ↩
Power that moves back and forth in an AC system to support voltage, without doing net useful work like running a motor over time.
synthetic inertia ↩
Software and power electronics behavior that imitates the stabilizing effect of physical spinning mass on the grid.
TAI ↩
International Atomic Time, a high-precision time standard based on atomic clocks.
UPS ↩
Uninterruptible Power Supply, a battery-backed system that keeps equipment running through short power disturbances.
voltage ride-through ↩
A requirement that connected equipment stay online through short voltage dips or disturbances instead of disconnecting immediately.

Reference links

Technical background on grid stability

Australia and battery deployment

Explainers and videos

Grid operations and pricing references

Data center and private generation examples