Exponential View - Nuclear Fusion w. First Light Fusion
Since the last Notes covering nuclear fusion (Oct 1st), there have been a number of big announcements demonstrating huge momentum in the space. Helion announced a $500mm round, with a further $1.7bn committed pending certain milestones to pursue their magnetic-inertial hybrid approach. And Commonwealth Fusion Systems (CFS), the start-up spun out of MIT, announced a $1.8bn round to build their first net gain (energy out > energy in) plant. Climate Tech VC have an interview with CFS following that announcement here. So that’s $2.3bn (or $4bn counting milestone capital) on the table for fusion. PLUS, Lowercarbon Capital, investors in CFS as well as two other fusion companies, are raising a dedicated Nuclear Fusion fund. Visualised inflows from Azeem’s newsletter:
It should be said in the context of the climate counterstrike that fusion doesn’t have an imminent role to play in decarbonisation (or anything else for that matter). For that we need to deploy existing tech as fast as possible. However, we are going to need a mind-boggling amount zero-carbon energy into the distant future, not only to stabilise the climate, but to contribute to all of the other things we want to do as a species. And that makes fusion terrifically exciting and worthwhile pursuing.
First Light Fusion is taking an inertial confinement approach to creating fusion. I first came across them when I visited Oxford Science and Innovation with Future Planet Capital a few years back. They stuck in the mind for having the greatest ever origin story, taking inspiration from the amazing pistol shrimp, which is the size of a thumb, but can create noise equivalent to a blue whale and temperatures of almost 5000 degrees centigrade. I found this to be an especially rich conversation, so could recommend listening to the full thing.
To achieve positive energy balance for fusion, plasma needs to hit a level of a triple product of temperature, density, and time maintained. Magnetic confinement approach, used by CFS, Tokamak Energy, etc, has a lower density and a longer time. The inertial confinement approach used by FLF has a much higher density (“tens or hundreds of times that of metals”) but only lasts for a tiny fraction of a second. [The energy balance ratio is referred to as Q, so Q>1 is more energy out than energy in or “net gain”.]
Other inertial confinement approaches, such as in the National Ignition Facility (NIF) at Lawrence Livermore National Lab, have been using high-powered lasers. FLF’s approach is to use a projectile fired with a rail gun to smash a capsule containing the fuel.
This approach is much more energy efficient than the laser approach, putting about 10% of the energy into creating the reaction, rather than 0.01%. Also it is much more straightforward from an engineering perspective. The laser approach requires a very complicated and fragile geometry of 192 mirrors precisely aligned, whereas the projectile system is more straightforward. This allows them to run more accurate tests in simulation and make faster progress.
FLF have already seen 10,000x improvement in the triple product output. They are now pushing to do another x100 performance output.
In general there is real meaningful progress being made in fusion. The NIF hit a Q of 0.7 this year, a joint record, and expect to hit net gain next year.
Progress over time has come with a series of ‘s-curves’ as new approaches have been tried, taken some time to understand, and then hit inflection points of delivering real progress. [It is worth checking out the animation at the end of Azeem’s newsletter to see the progress over the last decades, still below:]
The rate of progress in technical challenges as described by Elon Musk is a function of 1. How much is learned in each iteration and 2. How long each iteration takes. Nick suggests that in addition you want to try to build some parallelism into the process by having secondary learning goals for iteration that inform some other part of the process and don’t jeopardise the primary objective of that particular sprint.
FLF have yet to create fusion, so their milestones are first to create fusion, then to demonstrate net gain, then to build a plant that can create electricity.
It’s important to focus on keeping as many parts as standard as possible to reduce complexity and shorten the time to deployment. One element is the size, FLF will initially focus on 150MW plants. That will allow them to buy off-the-shelf components, turbines etc. When you get to GW scale then there is more idiosyncratic engineering. [Echoes Bill Brown first principle for developing climate solutions - use off-the-shelf, readily available components - Jan 15th Notes.] Additionally, they will aim to keep the costs of the plants at <$1bn, which will make them easier to finance.
Once the first working fusion power plant is built there will be a lot of things that can still be optimised on it. Nick Hawker suggests that we will need to resist the temptation and focus on deploying more of the same to get the scale quickly. [I find it somewhat ironic to hear a ‘deploy, deploy, deploy’ sentiment from someone developing fusion technology, and even within that a fairly novel approach.]
Timing for fusion:
2030 - likely to have a fusion reactor connected to the grid somewhere.
2050 - likely to be a part of the energy mix, but not a dominant part
2070 - likely to be a major part of the energy mix