The Grid Act Is the Wrong Way to Protect Consumers from Price Spikes
Senators Hawley (R-MO) and Blumenthal (D-CT) introduced the Guaranteeing Rate Insulation from Data Centers Act (GRID Act) last week in response to concerns that rapid AI-driven data center growth could raise household electricity bills. Some fear that large new loads from data centers will require costly grid upgrades and push demand high enough that more expensive generators must run, raising the market price of electricity for everyday consumers. Rather than modernizing how the grid manages large-scale demand to integrate new growth and protect consumers through better market design, the GRID Act imposes a punitive tax-and-subsidize regime exclusively for data centers.
The GRID Act would require data centers with a demand of 20 megawatts or more to either fully supply their own electricity through on-site generation or pay substantial “rate effect credits” designed to offset any measured increase in residential electricity rates attributable to their presence on the grid. The amount of those credits would be determined through an annual study conducted by the Department of Energy (DOE), assessing whether, and to what extent, a facility’s load raised local retail rates. If the study finds that a data center increased residential rates, it must pay credits equal to that estimated increase, with the funds directed to utilities or state entities to subsidize household bills. If a facility does not build its own power and also fails to pay the required credits, it faces civil penalties of up to $1 million per day until it complies.
The bill’s logic rests on a misunderstanding of how AI demand operates in electricity markets, assuming AI demand is inelastic—which it isn’t—and that hyperscalers knowingly push costs onto households, rather than responding to pricing structures that fail to reward load shifting and instead allow costs to be socialized after the fact.
First, to see how the bill’s logic depends on the flawed assumption that AI demand is inelastic, it is useful to return to the classic “fish market” analogy from economics. Imagine a local fish market early in the morning. The fishermen have already returned to port; the day’s catch is fixed and cannot be increased in the short run. Supply is perfectly inelastic. If a sudden influx of new buyers arrives, prices rise sharply because everyone is competing for the same fixed quantity of fish.
Now consider the two types of fish buyers. The first has an inflexible demand: they need fish for dinner that night and have no substitute. Even as prices climb, they remain in the market and pay whatever the going rate is. The second type has flexible demand. They see the higher price and either defer their purchase to another day or switch to a substitute, such as chicken. It is this second type of buyer that prevents prices from spiraling. Their ability to adjust limits whether, and to what extent, the most expensive suppliers set the price for everyone else.
Many households and critical services resemble the first type of buyer. When electricity prices rise, they often cannot meaningfully reduce consumption in the moment, or doing so would come at a cost. Electricity for heating, cooling, refrigeration, and medical devices is not discretionary.
But many AI workloads do not behave this way. Large portions of AI training demand are inherently flexible. These tasks can pause, slow down, shift to off-peak hours, or even relocate geographically without losing progress. In that sense, data centers resemble the second type of buyer in the fish market; they have flexible demand. This flexibility limits whether scarcity translates into sustained price increases for everyone else. Price hikes from AI demand become inevitable only when that demand is rigid. Where demand can adjust, price pressure is not a foregone conclusion.
Second, the bill rests on a narrative that casts data centers as a kind of deliberate burden on society. Senator Blumenthal describes households as being “forced to bankroll Big Tech’s electricity and infrastructure costs” and warns of an AI-driven “drain” on family pocketbooks. That framing implies hyperscalers are gaming the system, taking advantage of pricing rules that allow them to expand while others pick up the tab.
But this argument mistakes a failure of market design for a failure of corporate character. The behavior that critics say drives price increases is, in many cases, the only rational and incentivized response under current electricity pricing rules.
Returning to the fish analogy, imagine the fishmonger tells all buyers at the start of the season that fish will cost $5 each, a price that remains fixed for months. If several large buyers suddenly increase their orders, they face no higher cost, so they continue buying at the $5 rate. As supply tightens, the fishmonger is forced to procure additional fish at a much higher cost, absorbing the loss in the short term. At the end of the season, he raises the base price for everyone to recover his margins.
The large, flexible buyers might have been willing to switch to an alternative or delay their purchases if they had faced real-time price increases—or if they had been offered a discount to stay away during a shortage. Their behavior isn’t a greedy drain; it is a perfectly rational response to the rigid pricing structure they are given. In a functioning market, it is clearly harmful to the ecosystem to shut out these high-volume buyers just because the seller failed to design a way to integrate them. The onus is on regulators and the seller to offer the tools and terms that turn a massive new customer into a benefit for the whole market, rather than simply blaming the customer for showing up.
Similarly, treating the presence of data centers in the electricity market as inherently extractive confuses poor incentives with malintent. If policymakers want to protect households from price spikes, they should focus on supporting markets that activate and reward demand flexibility rather than forcing large users off the grid.
States like Texas are already moving in this direction. Under Texas Senate Bill 6 (2025), access to the grid is conditioned on flexibility. For new large loads above 75 megawatts, SB 6 requires data centers to install automated curtailment technology as a condition of connection. This is a set of digital protocols that allow the data center to surgically ramp down its grid consumption during an emergency. Crucially, this does not mean the data center necessarily pauses its work. Instead, most operators use this requirement as a catalyst to build on-site microgrids, using massive batteries or natural gas generation, to pick up the slack. During a curtailment event, the data center simply moves its load off the grid and runs on its own power, providing up to 100 percent reduction in grid demand without losing a single second of uptime for its AI workloads.
In exchange for this commitment, data centers are given faster interconnection priority. In a traditional system, a utility might tell a data center to wait 10 years while they build $500M in new transmission lines. But if a data center can move off-grid during times of the year when the system is stressed, it is not such a risk to other customers. This allows the utility to plug them in years faster because they don’t have to wait for those massive infrastructure upgrades.
Moreover, the Texas law establishes a voluntary program where the grid operator (ERCOT) compensates these flexible loads through capacity-style payments. These are essentially retainer fees paid to the data center just for remaining available as a controllable safety valve. This creates a powerful financial incentive to reduce grid draw precisely when demand and prices would otherwise surge.
Congress should encourage regional markets to adopt a flexibility-first model, one that treats large AI loads as active partners in grid stability rather than as passive cost drivers. By treating data centers as sophisticated partners, we can move past the false choice of choosing between innovation and affordability.
Image credit for social media preview: Don Sniegowski/Flickr
