Test report DSG-9535 · Rev F · tested November 18, 2026

AI Datacenter InfrastructureDevice under test

One Hour Off the Grid Saves a Texas Data Center $33.4 Million

NRG Energy says one hour off the grid per year spares a large Texas data center $33.4 million in transmission costs. Virginia testimony shows households paid 48% of one project.

Read
3 min
Words
666
Node
14nm
Operator
Amara Osei

Spec summary

  1. NRG Energy: a large Texas data center can avoid $33.4 million in transmission costs by going off-grid one hour per year.
  2. Peak shaving is legal and encouraged under U.S. power-market rules and state tariffs.
  3. Virginia SCC lawyer Andrew Major testified in July that residential customers paid 48% of a transmission project serving a data center.
  4. One hour equals 0.011% of a year's 8,760 hours; the avoided cost is roughly $9,300 per second.
  5. Will Baker of Everyday Electric: residential customers generally lack the same options to cut grid-related charges.
Data Centers Cut $33.4M in Grid Costs With 1 Hour Off
Fig. AData Centers Cut $33.4M in Grid Costs With 1 Hour Off — Akela999 / pixabay

A large data center in Texas can avoid $33.4 million in transmission costs by staying off the grid for one hour per year, according to NRG Energy. The claim, reported on Nov. 17, 2026, has become the top trending technology topic in the U.S., with coverage from Reuters, Ars Technica, Electric Vehicles and TechRadar.

The practice behind the number is peak shaving: briefly reducing grid electricity use during the few hours when system-wide demand peaks. It is legal and encouraged under U.S. power-market rules and state tariffs, because lower peak demand eases strain on the grid. The controversy concerns distribution, not legality. When one customer class shrinks its share of grid costs, the remaining total lands somewhere else — likely with ordinary households.

Spread over a single hour, $33.4 million works out to roughly $9,300 for every second of reduced draw. That arithmetic is ours, not NRG's, and it shows how lopsided the incentive is. One hour is 0.011% of the 8,760 hours in a non-leap year; the facility operates normally for the other 99.989%.

How does peak-based cost allocation work?

Grid costs are not spread evenly across the year. Regulators assign each customer a share of fixed transmission costs partly according to its use during the system's highest-demand hours. Customers who drive the peak pay more of the cost of building capacity for it.

Data centers fit the rule well. They run enormous, predictable loads, employ energy teams that watch the grid, and many operate backup generators. NRG identifies two routes off the grid: reduce consumption by pausing deferrable jobs or lowering cooling and processing loads, or run backup generation. Both count for the cost calculation.

Utilities pay the transmission costs and pass them to electricity customers. No source quantifies how much moved between customer classes.

Two cautions apply to the $33.4 million figure. It describes what a large Texas facility can avoid, not a documented saving by any named operator — no company names, sites or realized reductions are confirmed. And avoided transmission costs are not a total bill reduction; energy charges and capacity payments follow separate rules the sources do not break out.

Why can't households do the same?

Will Baker, head of market development at Everyday Electric, said residential customers generally do not have the same opportunities as large electricity users to reduce grid-related charges. A household cannot run a generator that carries a building, rarely knows which hour will peak, and pays a rate built on averages rather than peak response.

The strongest evidence in the reporting comes from Virginia. Andrew Major, a lawyer at the Virginia State Corporation Commission, testified in July that residential customers paid 48% — the largest share — of a transmission project serving a data center. The testimony supplies only the percentage and customer class; the project cost, facility identity and payment period are unknown.

What could regulators do?

The reporting describes no pending rule change, but the options on the table include:

  • Basing cost shares on a longer demand window, blunting the one-hour reward
  • Setting minimum contributions for very large customers
  • Extending demand-response programs to households via smart thermostats, batteries or time-based rates
  • Requiring utilities to publish class-by-class cost splits before projects are approved

A counterargument exists: if peak shaving genuinely lowers the system peak, less new capacity may be needed, which could reduce costs for everyone. No source quantifies that benefit.

What does this mean for cloud buyers?

For hyperscalers, power is a gating cost, and pricing that rewards flexibility makes deferrable workloads — batch analytics, training runs, storage maintenance — more valuable. If regulators tighten large-load tariffs, the savings now supporting data center economics could shrink. The context is already political: local fights have blocked 75 data center projects worth $130 billion, and AI infrastructure spending estimates approach $700 billion.

Peak shaving works as designed. The design now faces customers it was never built for.

via shattered.io (Source)

Filed under

  • peak-shaving
  • demand-response
  • transmission-costs
  • utility-rates
  • data-center-power
Share this article:

More from Amara Osei

Amara Osei

Show full bio

Staff writer covering money, funding and market shifts around hardware layer of artificial intelligence at Die Signal.

698 articles

Same lot · LOT-362E

« Previous articleNext article »