Companion piece to How Safe and Clean Is Nuclear Power, Really?, which deals with emissions and deaths. This one is about whether we need nuclear at all.

The question

I live in Austria, a country that voted against nuclear power before it ever produced a kilowatt-hour. In 1978 a referendum blocked the finished Zwentendorf plant by 50.5% to 49.5%, and since 1999 a constitutional law has kept the country nuclear-free. Austria gets away with it largely because of mountains: hydropower supplies most of its electricity.

Most countries don’t have the Alps. So the question is worth asking straight: now that wind, solar and batteries are cheap, can they decarbonise electricity without nuclear?

My answer has two parts, because the question hides two different ones:

  • Without nuclear? Yes. A zero-carbon grid can be built without it.
  • Without any firm clean power at all, meaning only wind, solar, batteries and transmission? Probably, but in the grid models discussed below, usually at a noticeably higher cost than a system that also has some clean power that can run through long shortfalls. Nuclear is one way to provide that; hydro, geothermal and others are alternatives.

By “clean” I mean close to zero emissions. Some firm options, like gas with carbon capture, cut emissions sharply but not to zero; I note where that matters.

The rest of this piece explains why the second question is the one that matters.

Where we are

In 2024 low-carbon sources produced 40.9% of the world’s electricity. The largest contributors were hydro (14.3%), nuclear (9.0%), wind (8.1%) and solar (6.9%), with bioenergy and other renewables making up the rest1. Wind and solar together overtook hydro for the first time. Nuclear’s share fell to a 45-year low, not because reactors closed en masse, but because everything else grew faster. Solar in particular has become the engine of growth: its generation doubled in three years.

Cost per kilowatt-hour: renewables win clearly

The most quoted cost comparison is the levelised cost of electricity: everything a plant costs over its life divided by everything it produces. Lazard’s 2025 estimates for new plants in the US, without subsidies2:

Axes: each bar spans Lazard’s low and high cost estimate for a new US plant of that type, in dollars per megawatt-hour (1 MWh = 1,000 kWh), without tax credits. Look for: the gap between the nuclear bar and everything else. Takeaway: in Lazard’s US numbers, new nuclear costs roughly three times as much per unit of electricity as new wind or solar, comparing the middle of each range. Lazard bases its nuclear estimate on Vogtle, the only recent US reactor project built from scratch, so other countries’ costs can differ.

Storage has followed the same curve. BloombergNEF’s 2025 survey put battery pack prices for stationary storage at about $70 per kWh of storage capacity, down 45% in a single year3. That is the purchase price of the battery packs alone, not of a finished installation, and not comparable with the per-MWh generation costs above. Ten years ago grid batteries were a niche; today they are routinely bundled with solar farms.

Why cost per kWh is not the whole story

LCOE treats every kilowatt-hour as the same product. Electricity isn’t. When every solar panel is producing at midday, power prices sometimes go negative; on a still winter evening, the same kilowatt-hour can be worth many times more.

Covering those hours happens on three time scales, and they get harder in that order:

  • Hours. Solar drops at sunset just as demand peaks at dinner. Tools: batteries, and demand shifting (moving flexible use, like charging cars or heating water, to other times). Commercially established, and getting cheaper.
  • Days to weeks. A cold, cloudy, windless week, the Dunkelflaute. Tools: interconnectors, hydro, building more wind and solar than average demand needs, firm power. Hard: covering a whole week from batteries takes many times more storage than covering an evening.
  • Seasons. In central and northern Europe, solar-rich summers and demand-heavy winters. Tools: seasonal storage (hydrogen, very large pumped hydro) and firm power. Hardest: the storage options are expensive, inefficient or limited by geography.

Batteries are superb at the first time scale. Where there is enough midday solar to charge it, a four-hour battery can shift that energy into the evening peak day after day. But adding one doesn’t make solar firm. Lazard’s California example shows the gap. A new solar plant costs about $51 per MWh, or $77 per MWh with a four-hour battery. The grid also needs enough capacity to cover its highest-demand hours, and solar alone counts for little there. Adding Lazard’s estimate of the backup capacity still needed brings solar alone to about $137 per MWh and solar with a battery to about $142: the battery costs more up front but reduces how much extra backup is needed, so the totals end up close2. Lazard stresses that even these totals cover peak-hour reliability, not round-the-clock supply through long lulls.

The second and third time scales are where costs climb, for a specific reason: storage sized for a two-week lull is fully needed only a few times a year, so its cost is spread over very little use. Hydrogen adds another penalty, losing well over half its energy on the round trip from electricity back to electricity. Germany saw what the second looks like in December 2024, when wind output fell about 85% below normal during a cold spell and short-term wholesale prices briefly neared €1,000 per MWh, more than ten times the year’s average4.

What the grid models say

How much firm power is worth can be modelled. Sepulveda, Jenkins and colleagues ran 912 scenarios across several emissions limits, technology assumptions and two US regions5. In the zero-emission cases, having firm low-carbon resources available (nuclear, natural gas with carbon capture, bioenergy and similar) lowered the cost of the grid by 10–62% compared with relying on wind, solar, batteries, demand flexibility and transmission alone.

The range is wide because the answer depends on region and on cost assumptions. And the study did not model very cheap storage that lasts for days or longer; if that arrives, the value of firm power could shrink considerably.

Other modellers, notably Jacobson and colleagues, model a 2050 US energy system running entirely on wind, water and solar, with storage, that keeps the lights on and has lower annual costs across all sectors (electricity, transport, buildings, industry) than their projected business-as-usual system6. Those systems include hydro and geothermal, so they show that a grid without nuclear can work; they don’t directly test Sepulveda’s question of how much cheaper it gets when more firm options are allowed. The real disagreement is about the cost of the final stretch, and how much transmission and storage can realistically be built.

So the modelling answer is: firm clean power is valuable, by a margin that depends on the region. Whether nuclear should provide it is a cost-and-speed question.

Nuclear’s real problem: time and money

Nuclear’s case is not hurt by its safety record or its emissions; as the companion piece shows, both hold up even under pessimistic assumptions. Its case is hurt by how recent Western reactors were built:

  • Olkiluoto 3 (Finland): construction began in 2005, completion was planned for 2009; regular production started in April 20237.
  • Flamanville 3 (France): construction began in 2007, completion was planned for 2012; it was connected to the grid in December 2024. France’s Court of Audit puts the construction cost at €13.2 billion against an initial estimate of about €3.3 billion (both in 2015 euros); including financing and other costs, the total reaches about €23.7 billion in 2023 euros8.
  • Vogtle 3 and 4 (USA): construction began in 2013, after licensing in 2012, with operation expected from 2016; the units entered service in 2023 and 2024, and the cost rose from an original estimate of about $14 billion to more than $30 billion9.

A reactor that takes 15 years instead of 5 is not only expensive. Every year of delay is a year of fossil generation that might have been replaced sooner, which is the core of Jacobson’s opportunity-cost argument. Other countries (South Korea, China, the UAE’s Barakah plant built by Korean firms) have built faster, which suggests the problem is partly one of lost industrial practice rather than physics. That is encouraging for the long term and unhelpful for the next ten years.

The easy decision: keep what already exists

There is one part of this debate where the numbers are lopsided. An existing, safely operating reactor’s construction cost is already spent. What matters for keeping it running is its operating costs plus any safety upgrades or refurbishment for a longer life, and the IEA’s 2022 analysis projects that most lifetime extensions will produce power at well under $40 per MWh in 2030 (in 2020 dollars), competitive even with new solar and wind in most regions10.

Germany provides the cautionary example. A study of the reactors it shut after Fukushima found that, over 2012–2019, their lost output was mostly replaced by coal and imports, at an estimated social cost of €3–8 billion a year (in 2019 euros), most of it from an estimated 800 extra deaths a year from air pollution11. The study ends before Germany closed its last three reactors in April 2023.

So, can renewables and storage do it without nuclear?

  • Without nuclear: yes. Wind, solar, batteries, hydro, transmission and other firm sources can decarbonise a grid.
  • Without any firm clean power: possibly, but usually at a higher cost, especially in regions with dark winters.
  • Most of the job is cheapest with wind, solar and batteries, built as fast as grids and permitting allow.
  • Existing reactors: keep them running while they are safe. Closing them early has tended to mean burning more fossil fuel.
  • New nuclear is one candidate for firm power. I think it should get that role where it can be built at a reasonable cost and speed, and not elsewhere.

The question to bring to the next argument isn’t “renewables or nuclear?” It is “what is the cheapest way to keep the lights on through the worst winter weeks, in this region, at close to zero emissions?” The answer differs between Spain and Finland, and it may change as storage costs fall.

  1. Ember (2025). Global Electricity Review 2025. ember-energy.org ↩

  2. Lazard (2025). Levelized Cost of Energy+ (LCOE+), June 2025, p. 8 (LCOE ranges) and pp. 28–30 (cost of firming intermittency). PDF at lazard.com ↩ ↩2

  3. BloombergNEF (2025). Lithium-ion battery pack prices fall to $108 per kilowatt-hour despite rising metal prices. Press release, December 2025. about.bnef.com ↩

  4. Clean Energy Wire (2024). Short-term power prices spike amid new ‘Dunkelflaute’ in Germany, most customers unaffected. 13 December 2024. cleanenergywire.org ↩

  5. Sepulveda, N. A., Jenkins, J. D., de Sisternes, F. J. & Lester, R. K. (2018). The role of firm low-carbon electricity resources in deep decarbonization of power generation. Joule, 2(11), 2403–2420. doi:10.1016/j.joule.2018.08.006 ↩

  6. Jacobson, M. Z. et al. (2022). Zero air pollution and zero carbon from all energy at low cost and without blackouts in variable weather throughout the U.S. with 100% wind-water-solar and storage. Renewable Energy, 184, 430–442. PDF at stanford.edu ↩

  7. World Nuclear News (2023). Finnish EPR enters regular electricity production. world-nuclear-news.org ↩

  8. Cour des comptes (2025). La filière EPR : une dynamique nouvelle, des risques persistants. January 2025, pp. 26–28. PDF via vie-publique.fr ↩

  9. American Nuclear Society (2024). Vogtle-4 enters commercial operation. ans.org ↩

  10. International Energy Agency (2022). Nuclear Power and Secure Energy Transitions. iea.org ↩

  11. Jarvis, S., Deschenes, O. & Jha, A. (2022). The private and external costs of Germany’s nuclear phase-out. Journal of the European Economic Association, 20(3), 1311–1346. doi:10.1093/jeea/jvac007 ↩