Senator Bernie Sanders and Rep. Alexandria Ocasio-Cortez are introducing legislation today that would ban construction of any new data centers with peak power loads exceeding 20 megawatts, according to TechCrunch AI. The moratorium would stay in place until Congress passes comprehensive AI regulation.
This is a significant escalation in the political pushback against AI infrastructure. We’ve seen local communities fight individual data center projects before, but a federal construction ban from two of the most prominent progressive lawmakers is a different animal entirely.
What the bill actually calls for
- Review and certify AI models before they’re released to the public
- Protect workers against AI-driven job displacement
- Cap the environmental footprint of data infrastructure
- Require union labor for data center construction
- Ban exports of advanced chips to countries without comparable AI regulations
That last point is particularly aggressive. Right now, almost no country has the kind of regulatory framework this bill envisions, which would effectively create a global chip export restriction far broader than current U.S. policy.
The political backdrop
Sanders’ office is leaning on quotes from tech leaders themselves to justify the move. They cite Elon Musk’s “AI is far more dangerous than nukes” remark, along with concerns raised by Google DeepMind’s Demis Hassabis, Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and AI pioneer Geoffrey Hinton.
Public sentiment is on their side, at least in broad strokes. A March Pew Research poll found that a majority of Americans are more concerned than excited about AI. Only 10% said their excitement outweighed their concern.
But political reality cuts the other way. AI companies have been spending heavily in Washington, and the competitive pressure from China gives hawks on both sides of the aisle reason to resist anything that could slow down American AI development.
Why this matters for the industry
The 20-megawatt threshold is key. That’s not a small number for a residential neighborhood, but for modern AI training facilities, it’s relatively modest. Most hyperscale data centers planned by companies like Microsoft, Google, and Amazon blow past that limit easily. If this bill somehow passed as written, it would freeze virtually all major AI infrastructure expansion in the U.S.
It won’t pass. Not in its current form, anyway. The political math doesn’t support it, and bipartisan resistance from members who see AI as critical to national competitiveness would be fierce.
What stands out here is the framing. This bill is best understood as an opening bid, a marker for what the progressive wing of Congress thinks AI regulation should look like. It sets the ceiling for negotiations. When a more moderate AI bill eventually moves through Congress, Sanders and AOC will have established a reference point that makes lesser regulation look like a compromise.
For AI companies and data center developers, this is a signal that regulatory pressure isn’t going away. Even if this specific bill goes nowhere, it reflects a growing political constituency that wants real limits on AI infrastructure expansion.
The full details of the legislation are available through the original TechCrunch AI report.