Nuclear encyclopedia

Chapters / The way forward

The way forward

What nuclear power is for in a decarbonising grid, what could plausibly be built by 2040, and what has to go right.

Status: the 2026 pivot documented; pipelines and SMR reality check to come. Release v0.8.1, 20 November 2026. Bracketed [source-id] links point to the source register.

The case for nuclear power in 2026 rests on firm, carbon-free electricity at a moment when grids are absorbing variable renewables faster than storage is arriving, and when data-centre demand has made twenty-four-hour clean power suddenly valuable. The case against rests on cost, construction time, waste and the stubborn coupling of civilian and military nuclear histories.

The plausible paths are enumerable: life extension of the existing fleet, a Western build-out that fixes its cost disease, continued Asian expansion, SMRs finding a factory model, and fusion arriving too late to matter for 2050 targets but early enough for 2070. This chapter keeps each path honest: what is announced, what is financed, what is actually under construction.

The pivot is real

The IEA's January 2025 assessment, The Path to a New Era for Nuclear Energy, found interest at its highest since the 1970s oil crises: more than 40 countries supporting expansion, 63 reactors (over 70 GW) under construction, lifetime extensions decided for more than 60 reactors (about 15 per cent of the fleet) in five years, and annual investment up by half since 2020 to over $60 billion. By November 2026 the construction pipeline had grown to 88 GW, the most in forty years. Nuclear supplies just under 10 per cent of world electricity, the second-largest low-emissions source after hydropower. [iea-new-era] [wna-cop31]

The tripling pledge

At COP28 in 2023, 22 countries launched the Declaration to Triple Nuclear Energy by 2050; by COP31 in November 2026 the count had reached 39, including eight of the ten largest economies. The World Nuclear Association's 2026 outlook finds declared government targets, if delivered, reach about 1,457 GWe by 2050, comfortably past the roughly 1,200 GWe tripling mark, with existing nuclear countries supplying 1,303 GWe of it. The gap is not ambition but delivery, which is what the rest of this chapter is about. [wna-cop31] [wna-outlook-2026]

A new buyer: the data centre

The demand shock that changed the politics is firm clean power for AI data centres, and the emblematic deal is a restart: Constellation's Crane Clean Energy Center, the undamaged Three Mile Island unit 1, closed in 2019, now being returned to service under a 20-year power purchase agreement with Microsoft, with a $1 billion federal loan and a target of the second half of 2027, ahead of its original 2028 date. Restarts and uprates are the fastest megawatts available: the site, the reactor and the grid connection already exist. [constellation-crane] [constellation-crane-2025]

The bottlenecks

The IEA names the imbalance plainly: of 52 reactors that started construction worldwide since 2017, 25 are of Chinese design and 23 of Russian design, and enrichment capacity is similarly concentrated, with Russia's Rosatom holding over 40 per cent of world capacity. A Western and Indian build-out therefore has to rebuild heavy-forging supply chains, fuel services (including HALEU for many advanced designs) and a construction workforce, not just sign targets. This is a decade-scale industrial project, and it is the honest reading of why announced pipelines exceed what will actually be poured by 2035. [iea-new-era] [wna-enrichment]

Waste: the Onkalo answer

On 22 October 2026 Finland granted Posiva the world's first operating licence for final disposal of spent fuel: the Onkalo repository at Olkiluoto, licensed to 2070 for 6,500 tonnes of uranium. The first deposition hole intended for actual disposal was drilled in August 2026, more than 400 metres down; disposal itself begins only after final commissioning inspections. The forty-year Finnish process, siting by consent, stepwise licensing, no schedule pressure, is the template Sweden, France, Switzerland and Canada are now following. The technical question is settled; what the world has lacked is the institutional patience. [posiva-licence]

What has to go right

The evidence so far supports a short list. Repeat designs at fleet scale rather than bespoke megaprojects: Barakah delivered four units near contract where first-of-a-kind EPRs and AP1000s broke budgets (see the economics chapter). Restart and extend before building new. Treat SMRs as an unproven industrial bet until a factory exists, not as a plan. Standardise licensing across borders so a design approved once is not re-litigated per country. And keep the weapons side of the ledger in view: a larger nuclear world only works if the non-proliferation regime in the treaties chapter holds.

Still to come in this chapter

  1. Country pipelines separated into committed, financed and aspirational, once the fleet dataset schema is approved
  2. SMR reality check: first-of-a-kind projects, dates and slippage, tracked release by release
  3. Proliferation and security risks of a larger nuclear world, and the governance on offer
  4. India 2047: what the three-stage programme could still deliver (with the India chapter)

Claims in the final text will carry citations to the source register; anything unverifiable will land on the gaps page instead.