The asset outlives the argument for it
Nuclear plants are now routinely operated well beyond their original design life. Sixty-year operation is the working assumption for modern designs, eighty-year licence renewals are established practice in the United States, and serious work is under way on what a hundred-year operating life would require. Set that against the industries that justify new capacity. The demand case for a reactor approved today will be superseded, probably more than once, inside the asset’s life.
This is not an argument against building for AI demand. It is an argument against designing the whole proposition around it. The compute industry is a strong anchor customer and a poor planning horizon. A reactor sited, sized, connected and financed on the assumption that one sector’s growth curve continues for sixty years is carrying a risk that the sector itself would never accept on its own balance sheet.
Consumers are transient; generation is not
Across its life, the UK fleet has powered heavy industry that has since closed, manufacturing that has moved, changing domestic heating patterns, electric vehicle charging that nobody modelled, and now data centres. The plants did not need to be redesigned for any of it, because electricity is fungible and the grid handles the matching.
That is the strongest argument for keeping new nuclear grid-connected and system-integrated rather than tying it tightly to a single private consumer. A grid-connected plant can serve a wider market over its operating life, although demand, prices and connection terms still carry risk. A plant built behind the meter for one customer has a market for as long as that customer’s business case holds, and then it has a problem that is expensive to solve retrospectively.
Where a private wire arrangement genuinely makes sense, and sometimes it does, the design should still preserve the grid route. The cost of building in export capability at the outset should be assessed for the site. Retrofitting it to a stranded asset can be costly.
What long-life thinking changes
Taking the asset life seriously changes several decisions that are usually made on shorter horizons.
Siting. A site chosen for proximity to today’s demand cluster may be poorly placed for tomorrow’s. Sites chosen for durable physical characteristics — grid access, cooling, maritime logistics, land for later units, established community familiarity — hold their value regardless of which industry is paying the bills in 2060.
Offtake. Power purchase agreements run for ten to twenty-five years. Reactors run for sixty to eighty. The structure needs to state plainly who carries the merchant exposure in the decades after the anchor contract ends, and the answer should be arrived at before financial close rather than inherited by whoever owns the asset at the time.
Capability. A sixty-year asset needs sixty years of people who can operate, maintain, inspect and regulate it. That is three to four working generations. The skills pipeline is not a construction-phase problem that closes when the plant connects; it is a permanent operating requirement, and the supply chain that supports it needs an order book stable enough to justify investing in it.
Modification and obsolescence. Nothing inside the plant lasts sixty years without intervention. Control and instrumentation systems will be replaced several times, each replacement carrying a safety case and regulatory burden. Designing for modification, and keeping the safety case in a state that supports change rather than resisting it, is a material contributor to long-run cost.
The useful reframing
The AI demand surge is a genuine opportunity for nuclear, and it has done more to restart serious investment conversations than a decade of energy policy. It should be taken for what it is: a credible anchor that improves the financing case for assets whose value extends far beyond it.
The projects that will look well-judged in 2060 are the ones that used today’s demand to get built while designing for a world in which that demand has been replaced by something nobody has thought of yet. That is the ordinary discipline of investing in infrastructure that lasts longer than the industries that pay for it.
NCL advises clients on long-horizon nuclear investment decisions, from siting and feasibility through to offtake structure and operating model. Get in touch to discuss a specific case.