OpenAI Tells Energy Dept: Fast Permits and Federal Backing for AI Supercomputer Hubs

OpenAI has formally asked the US Department of Energy to speed up permitting, offer favorable electricity tariffs and tax incentives, and derisk private investment for AI supercomputer hubs built on federal land. The company laid out its position in a public response to the DOE's Request for Information, titled "Development of a Public Solicitation of Private-Sector Proposals for AI Infrastructure Construction, Operation, Maintenance, and Decommissioning on Federal Land," and published the full submission on May 7, 2025.

The filing arrives as the federal government weighs how to open public land to private AI data center construction. The outcome matters beyond OpenAI: the terms the DOE sets will shape how quickly hyperscale computing capacity — and the power generation to feed it — gets built in the United States, and who bears the financial risk if projects stall.

A track record with the national labs

OpenAI grounds its pitch in existing work with the US National Laboratories, which it describes as "deep, mission-driven collaborations." In January 2025, the company launched a multi-year partnership with Los Alamos National Laboratory that it says is designed to "supercharge national science."

LANL Director Thom Mason has gone further in his own characterization, calling the partnership "a watershed moment for mission science" — a quote OpenAI repeats in its submission to underscore the seriousness of the arrangement.

The company's argument is straightforward: pairing national scientific talent and government supercomputing infrastructure with OpenAI's models has already, in its telling, unlocked new avenues of inquiry. In the RFI response, OpenAI writes that associating industry supercomputer infrastructure with "the talent and unparalleled scientific resources of our national labs" could expand those synergies, with three claimed payoffs — catalyzing American innovation, strengthening national security, and delivering durable economic growth in American communities.

That framing positions OpenAI not simply as a vendor seeking cheap land and power, but as an infrastructure partner to the national security and scientific research establishment. The DOE's solicitation process will test whether the federal government sees the relationship the same way.

The permitting ask

The most concrete section of OpenAI's response concerns permitting, where the company requests a suite of administrative accelerants. These apply, OpenAI says, whether projects sit on public or private land.

The specific mechanisms OpenAI advocates include:

• Categorical exclusions and programmatic reviews for repeatable deployments
• Emergency exemptions "where warranted and appropriate"
• Shot clocks for assessments and other required processes
• Surge staffing and prioritization of AI-powered permitting tools across key workflows

Each item targets a known friction point in data center development. Environmental review and grid interconnection queues routinely add years to large computing projects, and OpenAI wants the federal government to treat AI infrastructure as a category deserving expedited, standardized treatment rather than case-by-case review. The mention of "AI-powered permitting tools" is notable: OpenAI is suggesting that the permitting bottleneck itself could be attacked with the same class of technology the permits would enable.

The company calls streamlined permitting "essential" — its strongest single word in the filing — to enable "substantial, timely capital investments" in and construction of AI supercomputer hubs. The placement of permitting ahead of financial considerations in the submission signals where OpenAI sees the greatest delay risk.

Money and risk

On financial and contractual terms, OpenAI advocates for lease agreements that are "predictable" and "easy to execute," along with what it calls innovative approaches to derisk private investment and accelerate completion of AI supercomputer hubs both on and off federal lands.

The company's requested federal mechanisms range from competitive electricity tariffs to targeted tax incentives. OpenAI argues these would do two things: significantly mitigate risk for private investors, and accelerate deployment of critical AI infrastructure — with a claimed multiplier effect of "magnifying overall economic impact for the nation."

The ask amounts to a structured subsidy argument. OpenAI is not requesting a direct handout in this filing; it is requesting that the federal government lower the cost of capital and energy for AI infrastructure buildouts through market-shaped instruments. Whether the DOE adopts any of these mechanisms in its eventual solicitation will indicate how far Washington is willing to go to keep frontier AI compute construction inside the United States.

An open door, with conditions

OpenAI does not commit to building on federal land. The submission states plainly that "further review of any ultimate public solicitation will be required" before the company can determine whether co-locating its infrastructure on public lands would "indeed expedite and facilitate" its continued leadership in AI research and development.

That conditional stance makes the RFI response a negotiating position rather than a pledge. OpenAI is telling the DOE what would need to be true — fast permits, predictable leases, cheap power, de-risked capital — before it would bring its data centers onto government property.

The company closes its public summary on an optimistic note, saying it is "hopeful" that these considerations will be "front of mind" as the federal government works to expedite construction of the supercomputer hubs and energy generation infrastructure OpenAI says the country needs "for continued American AI leadership."

The pairing of compute hubs with energy generation in that final sentence is deliberate. The scale of infrastructure OpenAI envisions requires dedicated power, and the DOE — which manages both federal land portfolios and major energy programs — is one of the few agencies positioned to bundle the two. The DOE's forthcoming solicitation, informed by responses like this one, will show whether the federal government moves from soliciting opinions to offering terms that companies such as OpenAI will accept.