The United States Is Rewriting the AI–Energy Equation
Artificial intelligence now drives everything from cloud computing to national competitiveness — but its growth depends entirely on energy. In 2025, the U.S. Department of Energy (DOE) and National Nuclear Security Administration (NNSA) took a historic step by inviting private companies to build AI data centers powered by clean, reliable energy directly on federal land.
Announced in late September 2025, the initiative builds on DOE’s July site-selection process that identified Savannah River Site in South Carolina and the Oak Ridge Reservation in Tennessee as immediate focal points for AI and energy infrastructure projects. The Requests for Proposals (RFPs) issued on September 30 invite private developers to co-locate digital and power-generation assets directly on DOE land — an unprecedented move in U.S. energy policy that merges innovation, industry, and national security interests.
DOE’s Vision: Co-Located Data Centers With Clean Power
The DOE selected four sites — Idaho National Laboratory, Savannah River Site, Paducah Gaseous Diffusion Plant, and Oak Ridge Reservation — as the first hubs for AI and energy co-development. These locations were chosen for their access to existing energy infrastructure, transmission capacity, and readiness for rapid permitting.
The DOE’s goal is to establish self-sufficient campuses that generate or integrate on-site energy systems rather than rely solely on external grid power. The energy mix may include advanced nuclear reactors, hydrogen systems, renewables, or natural gas with carbon capture — whichever best ensures reliability, affordability, and low emissions.
AI’s growth is driving unprecedented electricity demand. U.S. data centers already consume roughly 4–5% of national electricity and could exceed 9% by 2035 as compute workloads surge. To meet that demand sustainably, the DOE is enabling collaboration between technology developers, utilities, and clean-energy innovators to provide 24/7 power with minimal carbon intensity.
By combining energy generation and computation within the same physical footprint, this model reduces grid congestion, transmission losses, and project timelines. Rather than making AI dependent on the grid, DOE’s vision positions AI infrastructure as a stabilizing part of the grid itself.
Why Hydrogen Fits the AI Energy Model
As AI data centers expand, hydrogen is emerging as a crucial clean-energy carrier. Hydrogen can store surplus energy, provide backup power, and stabilize grid operations during peak loads. DOE’s framework encourages a wide range of low-carbon hydrogen technologies, both green and blue, to be explored as part of site proposals.
One such innovation is FARST (Fluidised Autothermal Reforming Syngas Technology), a next-generation process for producing hydrogen from natural gas with integrated pre-combustion decarbonization. While independent from DOE programs, FARST exemplifies how industry-led solutions can support co-located data center models by supplying clean, dispatchable energy on demand.
FARST: A New Pathway for Clean Hydrogen Production
FARST combines partial oxidation and steam reforming within a fluidised-bed autothermal reactor, producing a hydrogen-rich syngas stream while capturing CO₂ before combustion. This “pre-combustion” capture yields a low-carbon hydrogen product with lifecycle emissions below those of conventional steam methane reforming.
Because FARST operates under pressure, CO₂ capture is more efficient and less energy-intensive. Comparable autothermal reforming (ATR) studies indicate lifecycle emissions as low as 3.9 kg CO₂ per kg H₂, demonstrating the potential for cost-competitive, low-carbon hydrogen.
When integrated alongside AI facilities, hydrogen from technologies like FARST could feed fuel cells or turbines to deliver firm, low-carbon baseload power, while heat recovery and carbon capture help improve overall efficiency.
The Power of Co-Location
Placing AI data centers and clean-energy systems on the same site offers multiple synergies. Shared utilities, combined waste-heat recovery, and flexible energy balancing transform the economics and environmental performance of both assets.
Hydrogen turbines or fuel cells can supply stable baseload power; stored hydrogen can cover peaks or backup; and excess energy can support local industry. This integration creates a multi-revenue, low-carbon energy ecosystem — one where digital infrastructure actively contributes to energy security and grid stability.
Savannah River and Oak Ridge: First Test Sites for the Model
At Savannah River, a 310-square-mile complex once central to the Manhattan Project, NNSA has opened 3,100 acres for private AI-energy proposals due December 5, 2025. At Oak Ridge, home to Oak Ridge National Laboratory (ORNL), proposals close December 1.
Both locations are envisioned as AI-energy innovation zones where industry partners can deploy nuclear, hydrogen, renewable, or hybrid systems to power next-generation compute infrastructure. Savannah River’s expertise in nuclear materials management and Oak Ridge’s research heritage make them natural proving grounds for co-located clean-energy hubs.
Addressing the Challenges
Integrating advanced energy systems with AI compute at this scale is ambitious. Developers face regulatory coordination, capital-intensive construction, and complex interconnection processes.
The DOE’s Speed to Power Initiative, launched in September 2025, addresses these hurdles by streamlining permitting and accelerating grid connections for large-scale energy-intensive projects, including AI facilities. Leveraging federal land and transmission assets helps reduce risk and timeline uncertainty for private developers while maintaining strict environmental and safety standards.
The Blueprint for a New Infrastructure Era
If successful, Savannah River and Oak Ridge will establish a blueprint for integrating AI and clean energy — a replicable model that could combine nuclear, hydrogen, renewables, or carbon-managed natural gas into a unified, low-carbon digital infrastructure.
AI and energy are no longer separate domains; they are converging into one strategic ecosystem. Embedding compute capacity within clean-power hubs allows the U.S. to scale digital capability and energy resilience simultaneously.
Technologies like FARST blue hydrogen show how industry-driven innovation can align with this shift, offering scalable, low-carbon solutions for powering AI’s next frontier.
The future of digital infrastructure is under construction — not just in data halls but across America’s evolving energy landscape. The companies that merge AI intelligence with energy innovation will lead the era ahead, where every teraflop of computation is powered by cleaner, smarter energy.