They outline the key challenges of the sector and produce economic models to compare development approaches. First, they explain that, to make an impact, fusion power needs to scale up rapidly once proven. However, the energy sector’s innovation cycle moves slowly, especially for large infrastructure projects. Unlike solar panels or wind turbines, which can be incrementally improved through frequent iterations, fusion reactors take many years to build which is a hindrance to effective learning.

This creates a risk of getting locked into a suboptimal technology. If only one concept is pursued and massive investment goes into scaling it up before sufficient learning has occurred, better approaches might be overlooked.

So, how do we ensure this doesn’t happen to nuclear fusion? The researchers developed economic models and used probability curves to compare sequential development, where one concept is tested at a time, with parallel development, where multiple concepts are tested simultaneously.

They calculate the optimal strategy by balancing the extra costs of parallel development against the increased probability of success. Key in the analysis is a method to estimate future revenues, factoring in the fact that these are worth less the further in the future they are. Put simply, they estimate how much we can spend to get fusion energy earlier, and with higher probability.

Their key finding was that the mathematically optimal strategy is to develop 5 to 10 different fusion concepts in parallel.

This matches the current industry structure, where over 40 startup companies are pursuing about a dozen fusion concepts, collectively attracting over $7 billion in private investment. While this might seem wasteful, Lopes Cardozo and Ward demonstrate it is economically rational. Their model shows that, when potential future revenues are large but uncertain and the learning cycle for each concept is long, parallel development maximises expected returns by finding the best approach faster.

Fusion power has the promise of great societal value, but it is presently unclear which technical approach will succeed. Whereas it may seem smart to try to ‘pick the winner’ at an early stage, the team’s paper shows why testing multiple concepts in parallel will potentially bring fusion power to market decades sooner.