This is a linkpost for https://futureoflife.org/podcast/johannes-ackva-on-managing-climate-change/ and https://x.com/FLIxrisk/status/1704960893417586992?s=20
In this podcast, Gus Docker and I discuss a broad set of climate-related issues from a broadly effective altruist lens. I also say “kind of” far too often.
Here’s the chapter list
00:00 Johannes's journey as an environmentalist 13:21 The drivers of climate change 23:00 Oil, coal, and gas 38:05 Solar, wind, and hydro 49:34 Nuclear energy 57:03 Geothermal energy 1:00:41 Most promising technologies 1:05:40 Government subsidies 1:13:28 Carbon taxation 1:17:10 Planting trees 1:21:53 Influencing government policy 1:26:39 Different climate scenarios 1:34:49 Economic growth and emissions 1:37:23 Social stability.
Stuff that I haven’t covered before
As I’ve done a bunch of podcasting over the last 15 months (here, here, and here), I was originally worried about overlap, but – to Gus’s credit – I think the overlap is not more than 20-30% and the FLI podcast also helps clarify some common misconceptions on what my views are (whereas the 80k podcast has most detail on philanthropy, our methodology, and prioritization, and the Volts podcast is the shortest overall introduction to our key ideas).
Here are some personal highlights of materials covered that I have not covered before (or in far less detail):
- (1) Significantly more on specific energy technologies and the big picture of energy transitions, in particular on energy density, energy transitions, and lock-in
- (2) How the current moment is quite different from the past in terms of US climate policy and how money is not the binding constraint on energy innovation anymore, the role of permitting reform and other regulatory interventions
- (3) Why it seems reasonable to think that most variance in energy technology outcomes is the result of social choices (rather than innate technological characteristics)
- (4) Why there is no alternative to a strong role of government in energy innovation
- (5) Which type of philanthropic interventions look most promising
- (6) Why carbon pricing is unlikely to be the main story
- (7) Whether we only recommend hedging against the worst worlds (no)
- (8) How significantly the risk of extreme scenarios has reduced.
- (9) A quantitative sense of the difference made by technological change compared to carbon pricing in moving things towards low-carbon competitiveness.
Materials discussed
- All of our work can be found at founderspledge.com/climate.
- Emissions by sector: https://ourworldindata.org/emissions-by-sector
- Energy density of different energy sources estimated from existing plants and extrapolated into the future: Noland et al 2022
- Emissions forecasts: The source I discussed in our conversation was this working paper from LSE, which estimates a distribution (with about 0.2-0.6% for 5-6C, compared to Wagner and Weitzman from 2016 suggesting a 5% probability of 6C or more), similar work was published in Science earlier this summer.
- “Structure of climate risk and its implications for high-impact philanthropy” is the talk I reference on risk management.
- Study on carbon pricing and the implicit carbon price one would have needed for German solar to be equivalent to what was possible with tech-specific subsidies (700 USD/t):
- Why not simply plant trees? Conclusion is to focus on avoiding deforestation, which should be easier (but see next bullet)
- Review study in Science showing REDD+, the primary attempt to avoid deforestation, not working, with only 6% of claimed additional deforestation reductions actually being additional (p.3-4).
- Decoupling, 24 high-income countries were able to achieve this over the last decade (Global Carbon Budget Summary Highlights)
- Premature deaths from air pollution, about 7m/year, from fossil fuels but also other sources (e.g. biomass).
Hi Johannes, I really enjoyed the structure of the interview and your detailed and careful answers. This made it easier to pinpoint a part of the interview where I think you might have too much confidence.
This part is around (3) on variance around outcomes. If I understand correctly, the argument put forward in the podcast is more or less that if we had given equal social choice to nuclear as we have done with wind and solar, nuclear would be highly likely to have followed similar cost reductions as wind and solar. I think we disagree but to clarify it might be helpful for you to put some numbers on it? Perhaps something like X% chance of reducing the LCoE of SMRs by more than 50% from the only built SMR where I could find some cost data, where I very roughly calculated the LCoE of as $127/MWh (might be low-balling, might be hidden costs as Russia is not known for transparancy). However, I take it from your statements you think there might be a ~70% chance of SMRs having become competitive with wind and solar if we have decided to support that technology similarly. Wind and solar have middle-of-the-range LCoEs at around $50/MWh and $60/MWh respectively, eyeballing Lazard's charts.
I think this is overly optimistic but not impossible. So my disagreement is more about the strength of your claim, not that it is impossible in all possible alternative worlds. I would very initially put something like a 30% chance that SMRs with a similar deployment in MW to wind or solar would end up below $80/MWh and maybe a 5% chance of getting closer to the $50-$60 range of solar and wind.
One main reason for this is that nuclear is not modular. Moore's law, solar cost decreases and also the Carlson curve are all dependent on massive scale, factory manufacturing, etc. Professor Bent Flyvbjerg at Oxford covers this quite well (especially solar, but also touching on nuclear and wind) and bases it on extensive data his team has collected. As an example that I think I have used on the forum before, the HTR-PM's first reactor took ~10 years to build. I doubt the first solar panels or wind turbines took that long.
I will stop writing as this comment is already long but would be happy to have a conversation about this. Perhaps there is something I am missing. I am coming from a project management and engineering background so I could be biased towards being somewhat dismissive of social and political influences.
Hi Ulrik,
thanks for your comment and for engaging!
I think there is a mix of (1) looking at things at (a) different time-scales, (b) and geographical levels (~ differences in perspective), (2) misunderstandings of what my view is here, so let me try to clarify:
(1) (a)
When I speak about social choice as the primary driver of techno-economic outcomes, I am taking a multi-decadal view on the level of the energy system at large, which is quite different from the perspective of a project manager and engineer in the short-term. It is certainly true that right now, as I discuss in the pod, it is easy and fast to build renewables and it is slow and difficult to build nuclear.
All I am saying is that the fact that this is so is the result of long processes of differential societal commitment, in the case of solar and wind it is the result of policy support since the 1970s to getting renewables cheap and, in the case of nuclear, similarly long efforts by large constituencies to make nuclear expensive and hard to build (+ other factors, as we discussed here).
(1)(b)
It is also important to not conflate project delivery times with energy system transformation at the system level. At the same time as renewables are cheap and easy to add to the grid, France was much more faster and complete in decarbonizing the grid in the 1970s and 1980s with nuclear than has been achieved with renewables to date (as I also discuss in the pod, the circumstances of France in the 1970s are not replicable).
In this context, it is also important to point out that the value proposition of SMRs or any other clean firm power option does not lie in meeting the LCOE of solar but rather in providing an energy system function that is distinct from the one that solar is providing. If one wants to choose an adequate cost analogue it would be solar LCOE + cost of seasonal storage + transmission in most contexts, i.e. marginal solar can be cheap, with more expensive clean firm power options in terms of LCOE still being quite valuable.
For (1)(a)(b) you can observe all of the things you mention and what I say can still be true, i.e. that these data points are the result of long-term societal choices and what matters is the ease and speed of energy system transformation, not individual projects. What I would say is that, philanthropically speaking, shaping the long-run system level picture is causally more relevant.
(2)
I think your comment sometimes mixes arguments about SMRs and traditional nuclear and I think about them as quite distinct:
Large-scale nuclear: As I discuss in the pod we see a 5x difference in building time for reactors when we compare France in 1970s to Western countries today. All of this difference is the result of social choice, it is not that large-scale nuclear is a new technology. It is not all affectable social choice (i.e. we could not induce the conditions that triggered 1970s France), but the variance is clearly not inherently technical. If we had had a pro-nuclear environmental movement that got concerned about climate change in the 1990s, this could look very different.
SMRs: As I said on various occasions, I am not confident we will get cheap SMRs, I think about it as a bet worth making. But for SMRs we should principally expect similar learning curves than for other modular technologies, that is the whole idea behind SMRs. They don't need to reach the LCOE of solar to be a valuable addition to the energy system as long as transmission + seasonal storage remain relevant barriers to a 100% intermittent renewable grid. And, here, at the same time as we have spent hundreds of billions on making renewables cheap, even in the US we still have a Nuclear Regulatory Commission that makes nuclear innovation harder than it should be. Given that the primary reason that renewables are cheap today are the efforts of jurisdictions actively anti-nuclear (Germany, California, (Denmark)) and that no country has made a bet on SMRs that is parallel to the bets that these jurisdictions have made on solar and wind, it seems quite plausible that the learning curves and cost reductions could have been induced for SMRs as well had they had similar support as renewables had.
Thanks for the clarifications Johannes, I think I agree with you on most points then. And you have changed my mind especially in one area and that is whether either a nuclear or alternatively a renewables strategy is best to pursue for a country/region before they embark on an energy transition. I think this question is still super relevant as most future emissions reductions/avoidance will have to come from Asia and Africa where renewables are still not widely deployed (this makes me question Western "development" organizations pushing hard for renewables without being clear about this as a somewhat risky bet). These countries actually today have a choice between liberalising and letting solar and wind rip, or, on the other hand, being a bit more heavy handed and using state control to push for a nuclear strategy. Bangladesh is just building its first nuclear plant and I used to think that was a bad strategy, but perhaps not (Although I fear that decision was made less based on a genuine and well thought through long term plan and more based on political pressures, I am initially sceptical of Rusatom that runs the project). I guess I had taken for granted a liberal electricity market and hence I felt that the nuclear path was really suboptimal as LCoE would matter. But if one can assume an ability to exercise strong state control, this changes my mind.
I think your point on system level costs is super relevant. In fact, I have been frustrated by the focus on LCoE in wind energy, as if you are building close to other wind farms, LCoE does not really matter that much as the market price (not taken into account in LCoE) will be low whenever your wind farm is producing. What matters is the revenue you capture, which can be high by producing small amounts at high prices (high LCoE). I think I even saw a study that cost optimized the Swedish grid for either a ~40% nuclear or ~90% renewable grid and the 2 systems cost came out pretty much the same. But we know for a fact we can get to a grid that is ~100% nuclear. But we have never before made a grid with close to 100% renewables (some people point to Denmark, but it is far from a good example being super connected to hydro in the north and lots of other generation/demand in the south). So they might have the same cost, but the renewable option is more risky. This is something you have changed my mind on. I hope this to some degree captures (perhaps with a lot of engineering simplification!) your argument about long time scales and total system cost.
I think perhaps our remaining disagreement is around the modularity and potential learning curve of SMRs. But I think our disagreement here is a matter of degree, not kind. I think we both think there is a chance that SMRs could come significantly down in cost with more deployment. I think perhaps you feel more optimistic about this than me but maybe not by much. And in any case I am not sure it matters that much as I think we both agree that the bet (or hedge as I like to think of it) is worth making, especially given that a lot of the world is taking the more risky path of widespread renewables instead of pursuing the nuclear option and we must be humble about the challenges we could face when renewables start dominating the grid. Perhaps I am also a little less worried (but still worried enough to support SMRs!) about the challenges of renewables. Especially industrial electrification makes me optimistic. An example of my optimism is that here in Sweden they are thinking about making steel from electricity and they are planning on large thermal reservoirs with hours (if not days) of thermal storage in order to quickly ramp electricty consumption up and down (I think GW scale!) based on electricity prices and requirements for frequency response.
Hi Ulrik,
Thanks for the good exchange -- that all makes sense.
I am unsure whether we disagree on learning rates for SMRs, we are just in the process of building a comparative tool to clarify our expectations of the returns of different innovation advocacy bets and, IIRC, SMRs sit in the middle range there based on stuff like Mahotra and Schmidt (2020?, from memory) on design complexity and customization and how this shapes expectable learning rates.
We'll publish this later in the fall and then we'll see whether we disagree:).
Your detailed work is likely to make me update so let's see. And you probably do this already but it seems intuitively worth looking at learning rates per unit produced, and not only MW. Solar panels might be 2000 units per MW, wind turbines 0.2 and SMRs 0.003 units per MW. Just feels like solar panels have a massive advantage here in terms of learning.