H.R. 10152, the Open-Source AI Leadership Act, was introduced last month by Rep. Gabe Evans, R-Colo., and would direct the Department of Commerce to support American-developed open AI models while assessing risks tied to models controlled by foreign adversaries. The House Energy and Commerce Committee's Subcommittee on Commerce, Manufacturing, and Trade is scheduled to mark up the bill today.
The legislation does not regulate data center electricity, water use or emissions directly, but it arrives as AI infrastructure already forces energy decisions well outside traditional IT departments. For companies expanding AI use the technology supply chain may no longer end with the server, cloud provider or data center. Organizations may increasingly need to understand where the model running on that infrastructure came from, who controls it, and what dependencies its use creates.
AI Policy Is Arriving Amid an Already Strained Power Buildout
Lawrence Berkeley National Laboratory's 2025 update to its data center energy usage report estimates that data centers could account for 11.8% of total U.S. electricity use by 2030, with scenarios ranging from 9.5% to 15.3%. That demand is already reshaping how projects get sited, with developers increasingly starting from available megawatts and working backward to determine where a campus can go, rather than treating power as a step that follows land acquisition. A Nature Sustainability study led by Cornell University found U.S. AI servers alone could consume roughly 245 terawatt-hours of electricity and 1.1 billion cubic meters of water annually by 2030, a figure that reinforces just how tightly AI growth and energy-water infrastructure planning are now linked. Against that backdrop, the model running on a given piece of infrastructure becomes another variable in the decision, not a separate one.
Commerce Would Report Findings, Not Ban Any Model
H.R. 10152 would require Commerce to assess risks involving personal and proprietary information, organizational security, model safeguards and national security, and it specifically directs the agency to examine risks to the supply chain of organizations that adopt foreign-adversary models. A business may already evaluate where its processors are manufactured, which data center hosts its workloads, how much electricity capacity is available, and whether its cloud provider can meet sustainability requirements. The bill suggests another question may join that list: who developed the underlying AI model. Model choice can also shape the infrastructure around a workload, since some open-weight models can run on an organization's own servers rather than solely through a third-party API, changing where computing demand lands and who is responsible for the hardware and electricity tied to it, a consideration already shaping how manufacturers and data centers compete for the same constrained grid capacity.
An important note - the bill, as it stands today, would not prohibit companies from using Chinese or other foreign-adversary open models; instead, Commerce would evaluate their adoption, cost, capability and performance against qualified U.S. models, with a first public assessment due within 18 months of enactment and annual reports afterward. It explicitly states it does not give Commerce authority to ban or restrict an open AI model, making H.R. 10152 less a restriction bill than a supply-chain visibility and domestic technology policy measure.
For energy, infrastructure and sustainability leaders, the larger signal is that the definition of AI infrastructure keeps expanding, and Congress is now signaling that the model running inside the physical system may deserve its own scrutiny too.
