I’d like to point out just how easy a non-carbon form of capitalism would be—and not just to imagine, but to accomplish. Here it is:
Over the next ten years, build about 5,000 standardized 5 TWh nuclear reactors worldwide and retire all fossil-fuel plants. Mandate a 20-year switch to electric vehicles. This would cost around $3 trillion per year, which isn’t much, and would cut carbon emissions by about 80 percent. Done.
Note that I’m not recommending this, nor saying that it would be mere child’s play. I’m just saying it’s far more feasible than reconstructing the entire global economy over the next decade …
Where to begin? This post will be completely unoriginal, but if it brings well-worn facts to the attention of a few new readers, it will be worthwhile. (Shade of MK: don’t be afraid to repeat the truth.)
One: no nuclear power reactor can be built anywhere without a cast-iron government guarantee replacing third-party liability insurance and government assumption of responsibility for waste, in practice a non-market price too (Hinkley C). A mass reactor buildout is the most socialist project you can imagine.
Two: As of mid-2018, 50 power reactors were under construction worldwide. The industry struggles and fails to build them on time, let alone to budget. The 50 have on average been under construction for 6.5 years. “The average construction time of the latest 53 units in nine countries that started up since 2008 was 10.1 years with a very large range from 4.1 to 43.5 years.” (World Nuclear Industry Status report, 2018, pdf). This is the best today’s battle-hardened reactor builders, the survivors in a steadily declining industry, can do. Building 5,000 at a time would have to be done by completely inexperienced workers, engineers and project managers. What could possibly go wrong?
Three: Drum thinks that worries about safety and costs can be addressed by using new designs: Generation IV, thorium, small modular reactors. No full-size commercial prototype exists for any of them, or even a fully licensed design. The normal procedure would be to build a handful of prototypes (at best 5 years construction, plus a crash two years for design and approval). Then you mass-produce the most successful prototype, betting there is one (another 6 years, allowing just one for design). So you don’t make any impact on emissions for at best 13 years, beyond the deadline for action set by the IPCC. Historical experience suggests 20 years. Plan B is to forget about prototyping and go straight to building 5,000 full-sized power reactors using completely untested designs and builders with no experience. This is not sane. (H/t to John Quiggin for emphasizing the timing issues.)
Four: Lazards’ 12th survey of US generating costs gives the comparative unsubsidised LCOEs per Mwh as:
- Utility PV solar $40 – $46 (midpoint $43)
- Onshore wind $29 – $56 (midpoint $42.5)
$57 – $148 (midpoint $102.5)$112 to $189 (midpoint $151)
The wind and solar numbers are hard data from hundreds of developers. The nuclear ones are optimistic guesses, as no new reactor site has been started in the USA for just short of a decade. This shows what real capitalists think of the technology.
I wonder what snake-oil salesman Lazards got the $57 low number from: it’s entirely out of line with recent experience in the USA and Europe with reactor construction. [Update: Sorry about that] The unsubsidised tag is incidentally a joke – there is no market price for the state guarantees for nuclear. So even in the best-case scenario, new nuclear in the USA will cost twice three times what you can get wind and solar for. There is no reason to think the global pattern will be significantly different. [Update: Jacobson gives the cost ratio as 2.3 to 7.4 , consistent with my rough estimate]
BUT, say nuclear advocates, nuclear reactors offer reliable! baseload! supply (90% availability) while wind and solar are intermittent (hiss hiss) and have to be firmed with gas or storage. This is a half-truth. The true part is that wind and solar do need firming with despatchable storage and transmission; I’ll go with Andrew Blakers’ calculated markup of 50% for Australia (pdf). But nuclear plants are lumpy and also need backup in case they have unplanned outages, an infrequent but regular event. (Simultaneous mechanical failure of equivalent volumes of wind turbines or solar panels is is no more likely than a giant asteroid strike.) Most of the world’s pumped hydro storage was built as backup for nuclear plants. Let’s say we need a 25% safety margin, either in the form of excess nuclear plants or the same mixture as for renewables. That pushes up your costs.
In addition, the claimed “baseload” feature is a bug to grid managers. Electricity demand cycles over the day and the year. Let’s just look at the daily cycle. A random day in January in California:
The minimum load, about 21 GW, is in the small hours. The peak, at 6 pm, comes in at about 32 GW, 50% more. So California would, in Drum’s scenario, need 23 GW of nuclear (at 90% availability) to cover the baseload. But going all-nuclear, it needs another 12 GW to cover the peak – running around half the time. That hits the LCOE, which is based on running round the clock. The average capacity factor drops to 26.5/35 or 76%. It’s lower again if you superimpose the annual cycle. By itself, the lower CF pushes up the LCOE by at least 15%. At first sight that’s not too bad, but power reactors are designed to run all the time. Ramping them up and down imposes stresses and shortens life, though the French are forced to do it with their huge nuclear fleet.
Taking these factors together, the various operational constraints just about balance the availability advantage of nuclear and maintain the proposition that 100% nuclear would cost at least
twice three times the 100% renewables solution.
Drum costs his thought experiment at $3 trn a year. Using the same back-of-a-napkin methods, say $1 trn for EVs, leaving $2 trn for electricity, of which $0.5 trn for renewables, so $1.5 trn for nuclear reactors or $15 trn over ten years. I have shown that
half two-thirds of this, $7.5 $10 trn, is an opportunity cost over wind/solar/storage. In what universe does it make sense to waste trillions of dollars on an unreliable and complex technology with problems we know all too much about?
I could go on: the waste headache, proliferation risks (Iran), the negative learning curve (in contrast to the well-behaved ones for wind, solar, and batteries, and known flat costs for pumped hydro and transmission), and fading public acceptance after Chernobyl and Fukushima. But the issues I’ve covered are enough to make it crystal clear that nuclear power is an obsolete, expensive and multiply risky technology we no longer need. Time to draw a line, break the bad news tactfully, and present the gold watch for long and moderately faithful service.
What annoys me most about Drum’s little jeu d’esprit is the subtext that since a massive rollout of nuclear power is evidently impossible, so is the whole GND or fast transition. This is false. As I’ve noted here before, cost estimates of the energy component of the GND from supporters (Jacobson) and foes (Holtz-Eakin), come in quite similar, under $1 trn a year, before netting out the current investment in oil and gas. This is large but practicable in the $20 trn a year US economy.
Zooming in, FERC thinks there are 188 GW of renewable generating plants being planned in the USA for installation in the next three years, say 63 GW continuous equivalent at an average 33% CF. The remaining coal capacity a year ago was 268 GW or 144 GW continuous equivalent. If all the planned renewables are built, they will make 44% of the coal fleet redundant in three years. Not all of them will see the light of day, but most will because they are cheaper and, with storage and gas backup, do the job just fine. A complete phaseout of American thermal coal in the next decade is not only feasible: it’s what current trends predict.
(Shade of Mark Kleiman again: Is it about something that matters? Check. Have you proofread for typos? Check. Source links and quotations verified? Check. [Update: not carefully enough, I’m afraid.] As short as needed to make the point? Guilty, but you know me.)