Last week, in the pinky, purply Octopus Energy headquarters, I attended a panel discussion hosted by Earth Set about Solar Radiation Modification (SRM), a set of technologies which aim to cool the Earth by reflecting sunlight back into space. The panel was Oliver Morton of The Economist, Colleen Golja of Imperial College London, Alistair Duffey of Reflective, and Mark Symes, who runs ARIA’s climate cooling programme. The seats were uncomfortable and the interior decoration was unconscionable, but the discussion was very interesting. When I asked the panel what China and the USA were doing on it, though, I was left wanting more. Understandably, given the focus of the evening, there was limited discussion on how these behemoths were positioning themselves, or whether they were in alignment. That matters more than anything else about this technology, because whether it is ever used will be decided by those two countries, and nobody else.
I’ve been interested in SRM since my masters in 2019, when I was researching Direct Air Capture (DAC), a related but distinct field of “geoengineering” (I don’t like this term, it is both scary sounding and unspecific) designed to hoover CO₂ out of the atmosphere. One of the leading DAC companies at that time was Carbon Engineering, co-founded by David Keith, then a Harvard Professor, who I learned was also a leading proponent of research into the sci-fi-ish SRM.
I’ll refrain from ham-fistedly explaining the different methods for SRM, or the physics behind how they work (nor could I if I wanted to). If you’re interested, I’ve published a presentation I gave to my research group describing the basic premises of the technology. What I want to talk about here is one method in particular, Stratospheric Aerosol Injection (SAI), and what it means for the sovereignty of nation states. SAI involves flying specially engineered planes into the upper atmosphere and jettisoning tonnes of sulphur dioxide, which transforms into tiny droplets of sulphate and floats around on the rarefied air. As it sweeps around the earth on currents, it reflects the sun’s energy back out to space, cooling the land and sea below. Releasing it anywhere (this is a bit of a simplification) reduces temperatures everywhere.
Research on SAI has two flavours: climate science and governance. The climate science has fairly convincingly made the case (using the same climate models that have accurately predicted global warming due to historic GHG emissions) that this technology would work (GeoMIP; National Academies, 2021), and could cool the earth by 0.5 or 1C pretty cheaply (in the order of 10s of billions a year; Smith, 2020). Most researchers think it could be used for “peak-shaving”, a hilarious term, which, despite sounding like something a man with a receding hairline might ask his barber for, actually means using SRM for a period in the latter half of this century to stave off the worst impacts of climate change, buying us the time we need to decarbonise. Most importantly, it could prevent us from pushing through tipping points (such as ocean current disruption, permafrost and sea-ice melting) which may irreversibly alter our climate, regardless of progress in decarbonisation.
The economics of SAI are thrilling. The corporate tax bill paid to the Irish government by Apple and Microsoft in 2024 was about €11 billion. The science suggests that if we spent that same amount of money on SAI, we could cool the earth by ≈0.5C. If spent on a rail link between London and Birmingham, it might get you half a bat tunnel. We are reasonably sure it would work, and it is cheap as chips. What’s not to like?
A lot of the criticism of SAI has centred around “moral hazard”, the idea that researching this technology would enable those with a vested interest in fossil fuels to delay their inevitable demise. This risk is much exaggerated (the evidence for it is thin). I think we will soon reach, if we have not already reached, a positive tipping point after which the energy transition becomes an inevitability (see EV sales).
There is one huge downside of SAI which I take very seriously, however: regional winners and losers. Changing the amount of energy retained by our climate system creates downstream effects on local weather patterns – the Indian monsoon is often given as the key example (although this has perhaps been overstated; Irvine et al., 2019). So the deployment of SAI becomes a governance problem - who, how, when and why.
The panel’s primary aim was to explain the science to a non-expert audience, which they did very well, but they touched upon governance a few times. Colleen Golja described a research project simulating geopolitical negotiations on the research and deployment of the technology, the results of which have just been published, and in which no agreement was reached. The report found that “issues of strategic competition and state sovereignty drove the simulations, not climate concerns”, with participants more interested in “the capacity to develop and deploy SRM technologies to demonstrate geopolitical supremacy”. Another panellist, asked how we could expect countries to agree on this, gave a qualified defence of the Montreal Protocol, which to this day successfully binds nations together in preventing the release of ozone depleting gases (it helped that the US, and DuPont, wanted it).
We need to be clear-eyed about what the technology involves. SAI is military-adjacent aircraft flying at heights historically reserved for spy-planes, releasing sulphur which will affect conditions in nearly every country in the world. If not agreed to, it infringes upon the sovereignty of nations and peoples over their own countries. We need to be unsentimental in considering how it will be interpreted by geopolitical adversaries who may be sitting around a table negotiating it later in the century. Do we believe that there are any countries in the world that would put the global good before their national interests? Some have argued (e.g. Blackstock et al., 2009) that if the impacts of climate change were so great, countries would be more likely to agree to SAI. I think that is a psychological miscalculation, which ignores the basic human hard-wiring that makes us prefer the risk of inaction to the risk of incorrect action (Ritov & Baron, 1990).
Thought of as an issue of sovereignty, it becomes like any other thing that countries would go to war to prevent (like a rogue state trying to build uranium enrichment facilities). Soon you wash up on the shores of a stark reality: there are two(ish) nations on Earth with the political, economic and military heft to either:
- unilaterally deploy this technology and trust that nobody will pick a fight with them, or
- unilaterally decide to prevent another state from deploying the technology through war.
I’m not the first to wash up here. Joshua Horton, Wake Smith and David Keith (who you’ll remember from earlier) wrote a paper last year making this exact case. They believe that about ten countries have the technical capacity to run a large-scale deployment (I think this is an overestimate, given the demonstrated state capacity of some of the European powers on the list), but only the US and China could do it in the face of strong opposition. They conclude that whether or not those two agree will decide if, when, and how SAI happens. For better or for worse, our fleeting custodianship of humanity is dominated by the USA and China, and I expect that to still be the case in 20 or 30 years, when SAI might become appealing, or even necessary. Human affairs in the 21st century are dictated by these two countries, as they were by the USA and Soviet Union in the second half of the 20th, the two nations who policed the spread of nuclear weapons, and by Britain before that, whose navy controlled the world’s shipping lanes.
We need to do the science on SAI, and we may need to deploy it in the future. But the wider conversation has not caught up with Horton and his colleagues. In terms of governance, a deal between the USA and China is the only show in town. Regardless of how much research we do, or how much we yearn for a more inclusive version of governance, where all the nations around this earth sit at the table and have a say in deployment, there is no world in which this technology is deployed without China and the USA jointly deciding upon how it works. That’s a problem, because the USA and China are geopolitical adversaries, and the race for AI is showing how dangerously that rivalry can be expressed. The USA has cut China off from the most advanced chips, and China has throttled the trade of rare-earth metals in response. Without trust between these two nations, there are only two outcomes. Either SAI remains the subject of speculative research and is never deployed, regardless of how bad climate change gets, or else somebody starts it anyway and we enter into a world where planes are being shot down and countries are using SRM as a weapon and counter weapon.
So, alongside the research, the most urgent thing is to start the practical work of coalition building now, while the stakes are still relatively low. We don’t need a huge treaty, or a global agreement just yet, but we do need US and Chinese scientists working together. We need real discussions between policy makers in these two nations, to iron out the conditions for research and future deployment, and we need a framework for monitoring the stratosphere, to ensure that neither suspects the other of secret research (the uncertainty about each other’s intentions that dogged the ARIA simulations). That is where the rest of the world can play an important role. Middle powers like the UK, which is already spending £57 million through ARIA, could deploy their soft power and resources more effectively by trying to get the US and China in a room, not just by funding science.
At the height of the Cold War, the USA and the Soviet Union, with trust at an all time low, managed to agree to rules for nuclear testing, and installed a hotline so that a misunderstanding didn’t become a war. It can be done, but that kind of trust takes a lot longer to build than a few planes (maybe not bat tunnels), so we should start building it now.
The slides
Download the slides as a PDF or PowerPoint.