A nail in a lead coffin

A damning chart by Arnulf Grubler of IIASA in Austria, via Joe Romm:

Figure 13: Average and min/max reactor construction costs per year of completion date for US and France versus cumulative capacity completed

Remember that the French nuclear programme had the most favourable institutional and political environment imaginable - a centralised polity, a stable political consensus administered by a technically-trained élite, a single capable purchaser insisting on maximum standardisation - and costs still went up.

Why the negative learning curve? Grubler and Romm have good ideas. They think that as you gain experience with building reactors, you discover more ways things can go wrong, so you add a layer of complexity, which later on leads to more problems, and so on. I’d add that as you make reactors more complex, you increase the amount of highly skilled engineering, management, regulatory, and political labour required.


While the nuclear industry is often quick to point at public opposition and regulatory uncertainty as reasons for real cost escalation, it may be more productive to start asking whether these trends are not intrinsic to the very nature of the technology itself: large-scale, lumpy, and requiring a formidable ability to manage complexity in both construction and operation. These intrinsic characteristics of the technology limit essentially all classical mechanisms of cost improvements - standardization, large series, and a large number of quasi-identical experiences that can lead to technological learning and ultimate cost reductions - except one: increases in unit size, i.e., economies of scale.

Anyway you don’t need an explanation to know that the chart dooms nuclear energy. Nobody can afford a technology of increasing costs. The free market understood this long ago, and nuclear power is still only kept afloat by generous subsidy and public guarantees of long-tail liabilities like waste disposal.

The latest corporation still active in in the sector to head for the door is the behemoth Siemens, which built all 17 reactors of Germany’s nuclear park. They pitch the decision as a response to German public opinion and Merkel’s decision to denuclearize German power, but it’s surely a heart a commercial one: their renewables business is the company’s fastest growing sector, and you put your effort where the money is.

I’m sure that some commenters on this blog will let engineering romanticism override their libertarian principles and say that nuclear power has been killed by over-regulation rather than its own weaknesses. Basically, a fusspot public that will not weigh the risks correctly. Naval marine nuclear reactors don’t seem to have the same problems, because navies accept the higher risks. That may be so, but the fusspot public is the sovereign people, and it really, really does not like very nasty radiation accidents from installations that have repeatedly been described as safe, safe, safe. If the industry had been honest and said from the outset “these is no such thing as completely safe, every machine or drug that works is dangerous, life is all trade-offs”, things might have been different. But it did not; and will die by its lies.

I do not see this as good news. Denuclearization requires a huge investment in renewables just to tread water on carbon emissions. Without nuclear, it’s harder to guarantee carbon-free electricity supplies to deal with the variability of wind and solar. On the other hand, closing the chapter releases scarce technical and managerial resources to concentrate on stuff that works: a learning curve requires people to learn. Not to mention risk capital, though not much has being going nuclear’s way recently.

Question: is there any reason to think that Grubler’s problem won’t apply to fusion, when ITER finally gets it to work?

A lot more effort should be going into geothermal. Here is some rock music recorded deep under Khazad-dûm the hot-dry-rock pilot plant at Soultz in Alsace, as supercritical hot water under high pressure cracks the granite.


  1. prasad says

    “is there any reason to think that Grubler’s problem won’t apply to fusion, when ITER finally gets it to work?”

    What’s wrong with the obvious one? The fundamental trouble with fission is keeping things from going critical. Lose control and you have a horrible safety issue. Fusion isn’t like that - unless you do an awful lot of clever stuff to overcome internuclear repulsion, you don’t get anything happening at all. Fission wants to blow up and become a bomb. Fusion wants to stop.

  2. Ebenezer Scrooge says

    I’m a nucleophile from way back. One of the reasons why I always tuned out the nucleophobes is that their arguments kept shifting with time-typically a sign of an intellectually weak position. However, Grubler’s chart (and Wimberley’s accompanying argument) nicely refutes this objection. The sound you hear is of somebody reconsidering their position.

    (“when ITER finally gets it to work”. Fusion has always been like Brazil-always the country of the future. However, the future finally might be coming for Brazil-maybe I shouldn’t be skeptical about fusion.)

  3. says

    I can’t tell if the dollar/franc values are inflation-adjusted. If not, then some of the upward curve is inflation, since the graph spans a few decades.

    It is surely significant in any case that nuclear construction is now more expensive per watt than well-sited photovoltaics. And I doubt nuclear would win much on maintenance costs either (though I’d love to see numbers).

  4. Tom Womack says

    The dollar/franc values are inflation-adjusted (maybe the axes have been relabelled since you put your post).

    prasad: I think fission usually wants to melt down and become a puddle, rather than blowing up and becoming a bomb; but neither circumstance is very good for the property value of your reactor, and applying that amount of thermal energy to anything containing chemistry causes it to get enthusiastically bigger, catch fire if it has any chance to catch fire, and start spreading dust (which is thoroughly radioactive dust since it’s spent time next to an unconstrained fission reactor) in many directions.

  5. Dennis says

    @ Prasad

    The reaction ITER uses is tritium + deuterium = helium4 + 14.1 MeV n

    The helium is fine, but the neutron is problematic. The plan is to use the neutrons to breed tritium from Li6, but it’s very unlikely that they can avoid neutron activation of other things in the vicinity of the reactor, like the reactor itself.

    Note that the goal of ITER is to have a relatively clean burning nuclear process. It’s not and can’t be completely clean. The reactor itself might be relatively fail-safe, although I’d hesitate to classify anything that produces 500MW of energy over 15 minutes or so as incapable of exploding. That is a lot of power to be shedding.

  6. James says

    Sorry if I did not read carefully, but did they incorporate reduced externalized costs? I know there were no major accidents in France, but if the risk did go down with the increasing costs, then they essentially internalized the costs of those risks and the chart would not be entirely correct.

  7. says

    James: The provisional version of Grubler’s paper (only covering France) is here. Romm links to the final version behind a paywall. Grubler discusses the measurement issues in some detail. His data are corrected for general inflation and levelized. The costs include decommissioning. I couldn’t see any discussion of dramatic changes in methodology over the life of the programme: this isn’t the pluralistic American environment, where different utilities might calculate things differently, the French data originally come from stable centralised bureaucracies (EDF, CEA).
    Short of a future expert refutation of Grubler’s final paper by someone with credentials equivalent to his, I suggest we accept his finding.

  8. Things that are useless says

    Anywhere you go on this mudball, you drill down 100,000 feet and you will have all of the heat you will ever need to generate electricity permanently.

    RE: Navy accepting risks. Why hasn’t the Thresher power plant ever been recovered? Why hasn’t the hydrogen bomb that was lost off of Thule Air Base in Greenland ever been recovered? The technology is there, the will is blocked by a bunch of two bit politicians who have been promoted way too highly in the military. Totally incapable of foreseeing a disaster in the making.

  9. Don says

    It’s been a long time since there was new U.S. commercial nuclear construction, but back then a huge proportion of the work at a nuclear construction site (memory says 40%) was re-work. That argues for the hypothesis that the technology is expensive because it’s inherently complicated.

  10. SamChevre says

    As the article is behind a paywall, I have a very basic question. How does this cost escalation compare to that of building roads, or coal-fired power plants, over the same timeframe? (It’s my impression that per-mile costs for new roads, and per-megawatt costs for coal-burning plants, has increased dramatically faster than inflation.)

  11. BM says

    (The high end of that graph is *still* cheaper than solar photovoltaic, isn’t it?)

    From the engineering end, I think what a nuclear engineer would tell you would be: yes, the standard PWR and BWR designs are “inherently complex”. The core, as a thing in itself, has all sorts of obvious failure modes. Each of these can be kept in check by a backup system. But the backup system needs a backup system, and a failover system, and a backup-failover-system, and failover-backup-system-diagnostics, etc., and so on to the edges of human ability to manage.

    However, I’d like to emphasize that this fact is known to nuclear engineers, and they sincerely think they can move in a different direction with modern plant designs.

    If you look at the (entirely theoretical AFAIK) “generation IV” reactor designs, they’re all intended to go in exactly the cheapo direction you describe—they’re supposed to be inherently safe. That’s supposed to mean that you can cut the pumps, turn off the computers, and walk away, and the abandoned hulk won’t release radiation. That doesn’t mean that the plant itself—the core—isn’t complicated, but it does mean that the complications can peter out without reaching 99.999999% reliability. You DO need six billion dollars’ worth of failsafes if your worst-case scenario is “we forced the abandonment of a city of 50,000″. You do not need six billion dollars’ worth of failsafes if your worst-case scenario is “we had to pull out one of the modules and buy a new one.” There’s also a lot of design progress towards small-scale, modular reactors—100MW is typical—that could have the mass-production economics you mention.

    This doesn’t make me an optimist, especially post-Fukushima, but the nuclear engineers I know are aware of the economic barriers to building future $6B super-power-plants, and they’re aware of the human-factors and complexity barriers to operating them safely. Which gives me some hope.

  12. says

    Shorter BM: Nuclear engineers to taxpayers - yes, we screwed up first time, sorry, so please let us have a second go with a different untested design philosophy!
    SamChevre: I can’t find data quickly and I have no comparative advantage over you in Googling. In conventional fossil power, gas turbines are much cheaper per kw than coal so the substitution must have driven down average fossil costs. Wind and solar have conventional learning-by-doing curves. The cost of solar PV modules fell by 7% in the USA last month, on trend not a blip.
    These curves again result from inherent characteristics of the technologies: small unit size, allowing unproblematic scaling; a large number of producers, and sharp competition between them in a global market; a standardized range of products; a short time to installation, allowing quick feedback and rapid product cycles. All this applies more to solar than wind, so the wind learning curve is flatter, but it still slopes in the right direction. Solar thermal is just starting, but the same factors will apply, especially to the steerable mirrors and control gear. Geothermal does have one inherently bespoke characteristic - the geology of each site is unique, as with an oil well. But once you’ve decided where to drill the holes (there is a risk of small earthquakes if you get it wrong), the rest can be standardised and we can confidently expect economies of scale.

  13. Barry says

    prasad says:

    “What’s wrong with the obvious one? The fundamental trouble with fission is keeping things from going critical. Lose control and you have a horrible safety issue. Fusion isn’t like that – unless you do an awful lot of clever stuff to overcome internuclear repulsion, you don’t get anything happening at all. Fission wants to blow up and become a bomb. Fusion wants to stop.”

    First, incorrect - fission reactors don’t want to blow up.
    As has been pointed out, both technologies release large amounts of heat, which must be continuously dumped. Losing cooling to a large fusion reactor would be catastrophic, and would well lead to an explosion.

    “Also, the waste products are much less of a problem.”

    Fusion reactors will have a great deal of material which has been exposed to an extremely strong neutron flux for a long period of time. What *do* neutrons do to many elements?

  14. Barry says

    Another comment on cost - I’d bet my money that nuclear power buries a far higher percentage of cost through government subsidies, ranging from small (the Navy runs a large training program for nuclear engineers and technicians) to large (long-term disposal of waste, and providing long-tail insurance).

  15. prasad says

    “First, incorrect – fission reactors don’t want to blow up.”

    Jeez, I was speaking colloquially :) Anyway, I don’t see how much the distinction matters for the point I was making, which is that with fission it is at least potentially hard to immediately stop the reaction from happening when things go really wrong, and that leads to all sorts of potential horrible problems. Whereas if something breaks with an ITER type plant and say their magnets or heat transfer or whatever breaks, roughly point is you don’t need to DO anything (not being in the solar core) profound to stop fusion from happening. A huge ITER accident is like (say) a large portion of the LHC or Fermilab blowing up. Bad, sure, but fission-bad.

    Re waste, if you actually think the waste problem is as bad as with fission, shrug.

  16. BM says

    Barry: “Losing cooling to a large fusion reactor would be catastrophic, and would well lead to an explosion.” That’s not really true. A fusion reactor’s heat load goes to zero, immediately, when you stop deliberately injecting and confining new fuel. This is in contrast with fission reactors, for which “turning them off” really means “turning the energy source down by 95%, then waiting a month”. Likewise: “Fusion reactors will have a great deal of material which has been exposed to an extremely strong neutron flux for a long period of time.” Yes, and the idea is to choose materials (blankets) whose activation products, if any, are short-lived and easy to dispose of. Fission reactors have no such options; fission is *fundamentally* about turning U and Pu and Th into highly-radioactive fission products, and nature doesn’t give you many handles on what those products are.

    (In the absence of a fusion technology that’s anywhere near being relevant for the warming/energy/peak-oil issues of, say, the next two decades, this is all of strictly academic interest.)

    But, James, I disagree with your assessment. This process—try something, learn in what ways it’s too expensive, try again—is really how engineering works. Let’s talk about this in terms of the curve you posted, rather than in terms of the accidents (about which more below). This curve shows that engineers have tried to do something, and it turns out to be expensive to do. That’s normal. You don’t find out what parts of engineering are difficult until you try, and you can map out a similar trajectory for many complex technologies. Take the V22 Osprey—another example of trying out an idea, finding that it’s unsafe, spending billions to fix it, and deciding it was the wrong approach. You don’t look at that and declare a *general* failure of aerospace engineering. Take the Space Shuttle—we spent $100B on this idea for a reuseable space plane that could launch once a week; it turned out to be so complicated that it needed $500M worth of maintenance on every launch. But you don’t take this as a generic indictment of rocket science; you take this as a lesson in value engineering, and you tell your rocket scientists to *keep it simple* in their next design. Another analogue may be drug development. The cost of modern cancer medication has skyrocketed; you have to commit $100M (or whatever) to put in your best-effort drug engineering, and only at the end do you find out whether the product has $100M worth of value. Again, that’s a lesson learned, and your drug-engineers can be asked to take it into account on their next round of engineering.

    It’d be nice to be talking about this in engineering/value terms; I’m aware that we also need to talk about it in collateral-damage terms, since the 50,000 people evacuated from the Fukushima area were (unlike the 14 dead Shuttle astronauts and however-many-dozen Osprey crash victims) never told that they were participating in a “lesson in value engineering”. I don’t have an answer to this; it’s ugly and I struggle to make sense of it. The problem is that 10 million coastal Bangladeshis have *also* not, to my knowledge, given informed consent to the Important Global Task of keeping US gas prices low—nor to the task of keeping French electricity investments below $5000/kW—-nor, more to the point, did they agree that a high-probability (global warming) risk to Dhaka is a fair trade for reducing the long-tail risk to Cadarache or Santa Cruz or Chalk River.

    Everything we do—build more PWRs, decommission existing PWRs, build Hyperion modules—is an engineering/economic choice with upsides and downsides. It looks like the (small, long-tail) downsides of Gen III reactor modules are preferable to the (large, immediate) risk of burning all the coal that we’re on track to burn. I believed that at $1000/kW and I still believe it at $6000/kW.

  17. Tangurena says

    >back then a huge proportion of the work at a nuclear construction site (memory says 40%) was re-work…

    Back then, reactor vessels were usually made in multiple pieces and welded together on site. The welds needed to be inspected by xray as that was the only way to be sure of anything under the surface.

    Today, reactor vessels are made in one piece, by a single company in Japan. They only make 4 per year and have a 15 year lead time. They also have a master smith making samurai swords, but my understanding is that the waiting period for those is a lot longer than for a reactor vessel.

    >I’d bet my money that nuclear power buries a far higher percentage of cost through government subsidies…

    According to the book “The Binding Curve of Energy”, the feds paid nuclear plant operators $500,000/kg of produced plutonium. A 1000 megawatt reactor will make about 290kg of plutonium per year. It was this subsidy that lead the head of the AEC to proclaim that “[in the future nuclear generated electricity will be] too cheap to meter” - because the production of plu is the main product and the electricity generation is a by-product. In the early 1970s, the US stopped purchasing plutonium with killed the economics of operating reactors. And after all, if you lose $100,000,000 of revenue per year, you end up hurting. But of course it is far more politically correct to blame tree hugging lawyers, Jane Fonda and “The China Syndrome” than to blame cold hard economics.

    The US military has about 100 tons of plu, about half of it in bombs. Commercially produced plutonium stockpiles are around 1700 tons, almost all of it sitting in spent fuel casks. If it weren’t for the Nuclear Nonproliferation Act of 1978 banning reprocessing fuel, I think the majority of those private plutonium stocks would have been turned into MOX fuel and used to power more reactors.