AI data centre energy planning is moving from grid requests to dedicated power campuses. In July 2026, the US National Nuclear Security Administration selected Amentum to negotiate a lease for a proposed 1-gigawatt artificial-intelligence data centre at the Savannah River Site in South Carolina, paired with about 2 gigawatts of on-site generation.
The scale is striking, but the wording matters: this is a selection for negotiations, not a final lease or a completed project. The proposal still faces commercial negotiations, permitting, safety and security reviews, and other federal approvals.
Savannah River selection opens lease negotiations
The NNSA announcement dated 20 July 2026 describes a phased public-private project. Its headline components are a 1-gigawatt data centre and approximately 2 gigawatts of dedicated generation, with natural gas acting as a bridge toward nuclear energy.
Amentum was chosen to enter negotiations after a competitive request for proposals. The agency explicitly states that the selection does not constitute a final lease award. Any agreement depends on successful negotiations and applicable regulatory, permitting, safety and security processes.
The Savannah River Site is a large federal nuclear complex near Aiken, South Carolina. An earlier NNSA request for proposals identified ten tracts covering 3,103 acres for potential AI and power infrastructure.
Why pair 1 gigawatt of computing with 2 gigawatts of generation?
A data centre’s headline capacity is not the same as its average electricity use, and a power plant’s nameplate capacity is not the same as continuous output. A larger generation portfolio can provide operating margin, cover maintenance, support variable workloads and potentially export electricity when the computing campus is drawing less than its maximum.
The proposed ratio also signals that electricity is becoming part of data-centre architecture rather than a service added after a site is chosen. Developers need to coordinate servers, substations, cooling equipment, fuel supply, transmission and backup systems as one integrated design.
One gigawatt is comparable to the output of a large conventional power station. A campus approaching that scale would be an industrial energy project as much as a digital one.
Natural gas first, advanced nuclear later
The government description says natural gas would bridge to nuclear energy. Gas turbines can generally be deployed faster than a new nuclear plant, and they can provide controllable output around the clock. That makes gas attractive when computing demand is expected before new reactors could realistically be licensed and built.
The bridge creates a climate and infrastructure dilemma. A gas-powered campus would produce substantial carbon emissions unless its operation, efficiency and emissions controls were carefully managed. A future nuclear phase could supply firm low-carbon electricity, but advanced reactor schedules, costs, licensing and fuel arrangements remain uncertain.
The announcement does not identify a reactor design, a firm commissioning date or a final generation mix. Treating the nuclear element as guaranteed would therefore go beyond the public evidence.
AI is changing electricity forecasts
The International Energy Agency’s analysis of energy and AI says data centres consumed about 415 terawatt-hours of electricity globally in 2024 and projects demand could more than double to around 945 terawatt-hours by 2030. AI is not the only workload, but accelerated computing is an important driver of new capacity.
In the United States, the Energy Information Administration has linked rising commercial electricity consumption to large computing facilities. Concentrated demand can create local bottlenecks even when national generation appears adequate, because new transmission lines, substations and power plants take years to plan and build.
What dedicated power could solve—and what it cannot
On-site generation may reduce the amount of new supply that utilities must procure for a single customer. It can also shorten the physical connection between generators and computing equipment, support resilience and give the developer more control over expansion.
It does not make the campus independent of the wider energy system. Gas still needs pipelines and fuel markets. Nuclear projects rely on regulators, specialised supply chains and long-term waste management. Data centres may still require grid connections for backup, balancing or electricity exports. Water use, land impacts, noise and local emergency planning also remain public concerns.
Lease, financing and power permits: milestones to watch
The next meaningful update would be a completed lease with a defined project company, financing structure and development schedule. After that, the key evidence will include environmental reviews, permits, interconnection agreements, gas-supply plans and specific generation technologies.
For the nuclear phase, watch for a named reactor design, a licensing pathway and a credible construction sequence. For the data centre, the useful details will be phased computing capacity, cooling design, water requirements and actual power-utilisation targets.
A blueprint under negotiation
The Savannah River proposal illustrates where the AI infrastructure race may be heading: enormous computing campuses planned alongside their own power systems. It also shows why announcements need careful reading. The project is significant because of its proposed scale and federal setting, but its final form may change considerably during negotiations and regulatory review.
If it proceeds, the campus could become a test of whether rapid AI construction, reliable electricity and a transition from fossil fuel to advanced nuclear can be coordinated in practice. For now, it is best understood as a high-profile blueprint—not an operating 1-gigawatt data centre.
The proposal is part of a wider shift explored in our report on why AI data centres are planning their own power plants.
Sources and further reading
- US Department of Energy/NNSA: Amentum selected for Savannah River negotiations
- US Department of Energy/NNSA: Savannah River request for proposals
- International Energy Agency: energy demand from AI
- US Energy Information Administration: electricity demand and large computing facilities
Reporting note: the Savannah River development remains proposed and subject to negotiation and approvals. FutureTechDose covers AI infrastructure and energy-sector progress for a general audience. This article is informational and does not provide investment advice.
Leave a Reply