AI data center regulation is drawing fresh attention as lawmakers prepare proposals addressing electricity costs, outside safety testing and AI developer liability in 2027. The debate involves technology companies, utilities and other power customers. Previous congressional votes and new federal energy estimates explain why the measures could matter to U.S. businesses.
Key Takeaways
- New AI and data center bills are planned for January 2027
- Proposals address electricity costs, independent AI testing and developer liability
- A 417-3 House vote preceded a failed 57-43 Senate procedural vote
- Federal researchers estimate data centers could use 9.5% to 15.3% of U.S. electricity by 2030, depending on the scenario
Reuters reported November 6 that lawmakers plan bills on data center electricity costs and AI safety for the Congress convening in January 2027. Reporting described the measures as plans rather than filed legislation, leaving their exact scope, possible exemptions and implementation dates uncertain.
No final bill text was reported, and the proposals are not enforceable rules. Utilities and developers remain subject to existing contracts, rates and state requirements.
AI Data Center Regulation Puts Electricity Costs in Focus
A proposal under discussion would protect electricity customers from costs associated with new data centers. Large computing campuses may require additional generation, transmission lines and substations. Allocation of those costs depends on local utility rules.
A June 2026 Lawrence Berkeley National Laboratory report estimated that data centers could use 11.8% of U.S. electricity in 2030 under its reference scenario, with other scenarios ranging from 9.5% to 15.3%.
The forecasts use assumptions about equipment shipments, energy consumption and cooling. AMD’s data center revenue growth indicates rising demand for AI hardware, although chip sales do not directly measure electricity consumption.
A September Congressional Research Service report found limited effects on electricity prices in many areas but warned of greater pressure where power supplies are constrained. Infrastructure upgrades, fuel prices and severe weather also affect bills.
A 417-3 House Vote Shows Why Power Cost Rules Remain Unsettled
On September 16, the House passed the Ratepayer Protection Act 417-3. It would require state regulators and certain utilities to consider a standard for recovering infrastructure costs attributable to large computing customers.
The standard would cover qualifying sites with peak demand of at least 100 megawatts and consider costs remaining if a customer ended service. It would not automatically impose one national electricity rate.
On September 30, a Senate motion to advance the bill failed 57-43, short of the required 60 votes. No new federal cost-allocation standard resulted. The vote showed that broad House support did not guarantee Senate agreement on the proposed approach to recovering power infrastructure costs.
Some jurisdictions already consider specialized rates and financial assurances for large users. Contract terms affect upgrade expenses and financial exposure if projects are delayed or canceled.
Independent AI Testing and Developer Liability Enter the Debate
The planned measures also extend beyond physical infrastructure. Reuters reported that Representative Lori Trahan of Massachusetts intends to pursue legislation for independent experts to evaluate leading AI models against safety standards. A separate effort would address when companies may be held legally responsible for harmful behavior involving their AI systems.
Neither proposal has an established nationwide testing schedule or a confirmed list of covered models. Details such as evaluation criteria, enforcement responsibilities and liability thresholds would depend on the final language of any bill. Independent review requirements for software should not be confused with electricity and construction requirements for data centers.
Congressional oversight is already taking place. In October, lawmakers asked OpenAI, Anthropic, Meta, Google and SpaceX for information about security practices and instances in which AI tools took actions their designers did not intend, Reuters reported. Information requests are distinct from binding safety standards.
For AI developers, the discussion raises questions about model documentation and how safety assessments could be conducted. For data center operators, the more immediate requirements remain power availability, connection terms, local approvals and cooling systems. Those physical needs also extend to AI data center cooling, an area that uses different equipment and materials depending on the technology selected.
Business Costs Depend on Existing Utility and State Requirements
For now, companies proposing new computing campuses must work within existing energy and permitting processes. Utility negotiations may involve the cost of new connections, required capacity, payment guarantees and the possibility of service delays. Conditions vary between regions, particularly where transmission networks have little room for additional demand.
The Congressional Research Service noted that data center development could raise costs in constrained markets, while additional electricity sales may help distribute existing utility expenses in other areas. Those different outcomes make it inaccurate to treat every new facility as the source of higher local power prices.
No reported proposal establishes an automatic electricity surcharge or a fixed compliance bill for U.S. businesses. AI data center regulation remains a developing legislative issue, while actual power charges continue to depend on approved rates, project agreements and state-level decisions.
Frequently Asked Questions
What is Congress proposing for AI data centers?
Lawmakers are preparing bills that could address electricity costs associated with large computing facilities, outside safety assessments and AI developer liability. As of November 6, 2026, the measures described in recent reporting had not become law.
Will AI data center regulation raise electricity bills?
AI data center regulation has not established a new nationwide charge. Whether data center growth affects local bills depends on grid capacity, utility costs and how regulators allocate expenses.
How much electricity could U.S. data centers use by 2030?
Lawrence Berkeley National Laboratory estimates a range of 9.5% to 15.3% of total U.S. electricity use in 2030, depending on the scenario. Its reference estimate is 11.8%, and all three figures are forecasts.
Did the 2026 Ratepayer Protection Act become law?
No. The House passed the bill on September 16, but a Senate procedural vote on September 30 fell short of the threshold needed to advance it.
