A ProPublica investigation published on November 6, 2026, revealed that US Secretary of the Interior Doug Burgum, who is promoting the construction of AI data centers on federal lands, personally holds between $18.1 and $64.4 million in investments in AI companies, while his subordinates in land matters own shares in Nvidia, Meta, Microsoft, and stakes in xAI. Concurrently, the Bureau of Land Management (BLM) is reviewing permits for approximately 15 data centers and 17 infrastructure projects in six western states. These decisions will determine where and under what conditions new computing power for the American AI industry will appear.

image
image

What happened

On November 6, 2026, ProPublica published an investigation by journalist Mark Olalde regarding the personal finances of officials issuing permits for AI infrastructure on US federal lands. Interior Secretary Doug Burgum, co-founder of the venture firm Arthur Ventures, holds between $18.1 and $64.4 million in investments linked to AI companies, and earned $2.5–15.1 million in profit from the sale of AI stocks in 2025 alone. Bureau of Land Management (BLM) Director Steve Pearce owns shares in Nvidia, Meta, Microsoft, Apple, and Alphabet, and has invested in data center developer Titan Development. Deputy Secretary of Agriculture (USDA) Michael Boren holds $1.5–6.1 million in xAI and up to $500,000 in Valar Atomics, a company that filed an application for the Project Beehive complex of more than 450 small nuclear reactors on BLM lands in Utah. At the time of publication, the BLM was reviewing permits for approximately 15 data centers and 17 infrastructure projects, including nuclear plants, in Arizona, Nevada, Oregon, Utah, Wyoming, and Idaho. According to journalists' calculations, 380 administration appointees have financial ties to Nvidia, more than 320 to Alphabet, and more than 240 to Meta.

Context

Understanding the situation requires recognizing the role of the agencies involved. The Bureau of Land Management and the Forest Service distribute access to vast federal territories in the American West, where the AI industry is seeking sites for energy-intensive facilities: the industry's bottleneck is increasingly becoming land, energy, and water rather than model architectures. Burgum sets the public frame by comparing AI development to the Manhattan Project and calling the current situation an "AI arms race"—in this logic, the speed of permit issuance becomes a competitive advantage. Among the applications under review are not only data center sites but also energy projects, so agency decisions will also determine the industry's future energy framework. Against this backdrop, CREW and ethics specialist Richard Painter assess the configuration in the ProPublica material as a conflict of interest in terms of "visibility": the mere overlap of personal portfolios with permit-issuing authorities diverges from transparency standards.

Why this matters for the industry

For the AI industry, the signal is twofold. Federal lands are becoming a key platform for hosting data centers and energy infrastructure, so permit issuance rules effectively become rules for accessing future computing power. However, applications are being reviewed by people with personal positions in Nvidia, Alphabet, Meta, xAI, and infrastructure developers, so competition for sites may come down not to technology but to access to the permitting channel: an advantage that cannot be bought with technology and can easily be taken away with a change in personnel or administration policy. Until BLM applications from the queue are approved, the pipeline remains a list of announcements rather than confirmed capacity, and teams planning capacity should factor in a political risk premium for each project on these lands. A separate effect is the demand for transparency: the investigation provides a structured body of data on permits and asset owners, based on which monitoring systems for application statuses can already be built.

Why this matters for users

For readers, this is a specific mechanism through which changes can reach daily life and wallets. Land, water, and electricity in the southwestern states where construction is planned are limited resources, and permit issuance by the BLM and Forest Service can influence their prices, and thus regional tariffs and rents. Those evaluating AI service providers or planning capacity in the US should understand that nothing changes instantly; all projects are still under review, so current computing prices and availability remain unchanged. It makes sense to track two indicators: which companies will receive land plots and electricity contracts, and under what disclosure conditions. If within a horizon of approximately two years some facilities actually go into construction, the supply of computing power in the US will increase, and its geography will shift to the southwest—this will eventually affect the cost and availability of AI services.

What is currently unknown / limitations

The weak points of the evidentiary base are known. Asset ownership ranges are taken from financial disclosures and denote ranges rather than exact amounts. The investigation records asset ownership and the progress of application reviews but does not show a single specific decision made under a personal conflict of interest. The material relies on one journalistic work by ProPublica: there is no independent second confirmation, official responses from agencies to the publication have not yet followed, and the Hacker News article had 2 points and 0 comments at the time of checking. Valar Atomics' Project Beehive remains an application rather than a verified technical concept: timelines, engineering risks, and the actual energy density of the small nuclear reactor complex are not disclosed, so it is premature to include it in industry plans.

Sources

Author

Look at AI, editorial team