Physics / mechanism
Advanced nuclear is used here as an umbrella term for fission reactor concepts pursued as alternatives to the conventional operating fleet. The available sources treat it as a funding and deployment category rather than describing a single reactor physics package: the US Department of Energy’s ARPA-E ran a dedicated programme that awarded $24 million across 10 advanced nuclear reactor projects. The spread of that award across ten separate projects indicates a portfolio approach to early-stage reactor concepts rather than convergence on one design.
The commercial case advanced in the sources rests on the load profile rather than on physics novelty. Fission is presented as a baseload generator, with the argument that steady output is well matched to datacentre demand, while levelised cost of electricity is identified as the weak point of the case ref. The same analysis is sceptical on both timeline and cost, concluding that nuclear is a partial rather than a complete answer to datacentre power demand ref.
Key parameters that decide the category, on the evidence supplied, are therefore delivered cost per MWh, time from contract to power on the grid, and regulatory clearance for specific offtake structures. The sources do not give reactor-level parameters such as coolant, fuel enrichment, outlet temperature or power rating for any of the funded concepts.
Competitive landscape
The sources do not support a design-by-design comparison of advanced reactor types. They position the category against the demand side instead: hyperscale datacentre operators are treated as the marginal buyers, with Google and Microsoft described as actively pursuing nuclear supply, Amazon reported as having had an arrangement blocked, and Meta also named as pursuing nuclear ref. The competitive question posed is whether nuclear can beat alternatives on LCOE, with the source’s own answer being that it likely cannot fully displace them and functions as one component of a datacentre power mix ref.
Evidence base
- ARPA-E launched a programme awarding $24 million to 10 advanced nuclear reactor projects, announced 22 July 2020.
- A November 2024 investor analysis concluded that nuclear is “probably not the answer” for datacentre power, but a partial answer ref.
- The same analysis summarised its position as “LCOE bad, baseload good”, identifying cost of energy as the constraint and steady output as the attraction ref.
- As of November 2024, Google and Microsoft were reported as pursuing nuclear supply for datacentres, an Amazon arrangement had been blocked, and Meta was also engaged ref.
- Scepticism in the November 2024 source is directed specifically at timeline and cost rather than at technical feasibility ref.
Frontier (open questions)
- Which reactor concepts received the 10 ARPA-E awards, and what were their coolant, fuel and power-rating specifications?
- What delivered LCOE, in $/MWh, do advanced reactor developers contract at for datacentre offtake, and how does that compare with firmed renewables plus storage on the same grid?
- What was the regulatory basis for the blocked Amazon arrangement, and has that precedent been resolved for behind-the-meter nuclear supply?
- Did any of the ARPA-E-funded projects reach a construction permit or first criticality, and on what schedule relative to their original claims?
Synthesised 2026-08-31 from 2 KB sources by the resynth pipeline; citations are KB source slugs.