Abstract visualization representing AI infrastructure and data center energy consumption

The AI infrastructure boom is reshaping America's power grid — and now facing its first major legislative challenge.

Credit: Pawel Czerwinski / Unsplash

TechnologyPoliticsAI Policy6 min read

Sanders and AOC Unveil "AI Data Center Moratorium Act" to Pause Infrastructure Boom

The legislation seeks an immediate federal freeze on AI data center construction until Congress establishes comprehensive safeguards for ratepayers, the environment, and workers

JS

Politics & Investigations

The Bill

Senator Bernie Sanders (I-VT) and Representative Alexandria Ocasio-Cortez (D-NY) introduced the Artificial Intelligence (AI) Data Center Moratorium Act on Wednesday, March 25, 2026 — a direct legislative strike at the trillion-dollar infrastructure boom transforming the American power grid.

The bill seeks an immediate, nationwide halt on the construction or upgrading of AI data centers until Congress establishes comprehensive federal safeguards. In a joint statement, Sanders and Ocasio-Cortez framed the current wave of AI infrastructure spending as an "unregulated frenzy" that is being conducted at the direct expense of residential ratepayers, local water supplies, and workers facing displacement.

"The American people did not vote to subsidize the electricity and water bills of trillion-dollar corporations building data centers," Sanders said. "We are calling for a reasonable pause so that democratic oversight can catch up with technological advancement."

The Infrastructure Boom They're Targeting

The legislation is a direct response to a wave of AI infrastructure commitments that has accelerated dramatically since the release of ChatGPT in late 2022 and has reached a crescendo in the first quarter of 2026.

Microsoft, Google, Amazon, Meta, and a consortium of investors behind OpenAI's Project Stargate have collectively committed more than a trillion dollars in AI infrastructure spending over the next five years. New data center campuses are under construction or in planning across Virginia, Texas, Georgia, Nevada, and the Midwest — each requiring hundreds of megawatts of power and millions of gallons of water for cooling systems.

The Department of Energy projects that AI data centers could account for as much as 9% of total US electricity consumption by 2030, up from approximately 3% in 2023. In several states, utilities have already raised rate projections for residential customers, citing data center load growth as a primary driver.

stats35B

Meta's AI infrastructure spend (2025 guidance)

$500B

Project Stargate commitment (OpenAI / SoftBank / Oracle)

3%

Share of US electricity AI data centers consumed in 2023

9%

Projected share of US electricity AI will consume by 2030 (DOE estimate)

Why Now

The timing is not accidental. Meta announced this week that it is beginning a new round of layoffs — framed internally as a strategic pivot away from Reality Labs and toward its $135 billion AI infrastructure build-out. The announcement crystallized, for many progressive lawmakers, the central paradox of the current AI moment: companies are simultaneously cutting workers and demanding billions in infrastructure investment that flows through public utilities.

Sanders and Ocasio-Cortez have been consistent critics of what they call "corporate AI welfare" — a framing that positions data center tax incentives, below-cost power purchase agreements, and favorable zoning approvals as subsidies flowing from public to private hands without public consent or accountability.

Ocasio-Cortez's track record on tech regulation has been notably consistent. Earlier this month she raised concerns about the normalization of pervasive gambling platforms in professional sports , arguing that the same pattern of rapid commercial expansion without regulatory guardrails that characterized online gambling is now playing out in AI infrastructure.

The Export Ban Provision

The bill's most internationally consequential clause would prohibit the export of high-end AI accelerator chips — specifically calling out NVIDIA H100 and H200-class hardware — to any country that does not have equivalent federal AI safeguards on the books. This provision would effectively extend the moratorium's conditions into a foreign policy instrument, potentially creating a two-tier global AI infrastructure regime.

The provision echoes existing Biden-era export controls that restricted NVIDIA chip sales to China, but goes further by applying the standard to all countries regardless of geopolitical alignment — including allies in Europe and Southeast Asia that currently lack comprehensive national AI regulation.

Industry Response

The bill landed predictably on the tech industry. The Consumer Technology Association released a statement arguing that a moratorium would "cede AI leadership to China and other adversaries" and cost tens of thousands of data center construction jobs. Several venture capital firms circulated memos describing the legislation as "economically illiterate."

NVIDIA did not respond to a request for comment by publication time. Meta, Microsoft, and Google declined to comment on pending legislation.

More nuanced pushback came from moderate Democrats, several of whom acknowledged the underlying concerns about grid stability and environmental impact while arguing that a blanket moratorium was too blunt an instrument. "I share the concern about ratepayer protection and water use," one House Democrat told reporters on background. "I don't think the answer is to stop building."

Prospects and Path Forward

The bill faces substantial headwinds in a Congress where both parties have generally supported AI infrastructure investment as an economic and national security priority. Republicans are broadly opposed to federal moratoriums on private construction, and the legislation is unlikely to receive a floor vote in the current session without significant bipartisan support it does not currently have.

However, lawmakers and political observers note that the bill's primary political function may not be passage — it may be agenda-setting. By codifying a specific set of conditions (ratepayer protection, renewable energy mandates, union labor, chip export controls), the legislation establishes a progressive negotiating floor for any future federal AI framework that does move forward.

More immediately, the bill gives state-level regulators and local governments a rhetorical and political tool to use when negotiating with data center developers over power purchase agreements, zoning variances, and tax incentives. Several mayors and governors have already cited federal inaction on AI infrastructure as justification for more aggressive local terms.

WHAT TO WATCH
  • Whether the bill attracts co-sponsors beyond the progressive caucus — any crossover signals broader legislative momentum.
  • Industry lobbying response: NVIDIA, Microsoft, and Google are expected to brief sympathetic committee members within days.
  • State-level activity: California, New York, and Texas are each considering their own data center regulation frameworks, which may move faster than federal legislation.
  • The chip export provision will draw scrutiny from the national security community — watch for pushback from the Department of Commerce.

Discussion

Sign in to join the conversation

Your comments appear live in our Discord server — every post grows the community.

Every comment appears live in our Discord server.

Join to see the full conversation, get notified on new articles, and connect with the community.

Join ObjectWire Discord

Comments sync to our ObjectWire Discord · Sanders and AOC Unveil "AI Data Center Moratorium Act" to Pause Infrastructure Boom.