Artificial intelligence has begun to transform the way many energy companies do business — from transmission planning to document retrieval at nuclear facilities. And the technology is also helping to accelerate and advance scientific research, such as the discovery and refining of battery materials.
Now, a new report suggests it can also take us a step closer to what many consider energy’s holy grail: nuclear fusion. Released last week, the report — which is a survey of existing research — details how “advancements in artificial intelligence and high performance computing are accelerating the development of fusion energy technologies.”
Amazon Web Services — which also announced last week that its high performance cloud platform is in use by Commonwealth Fusion Systems, the fusion giant backed by Bill Gates and Jeff Bezos — supported the research.
The potential of both AI and high performance computing, or HPC, to accelerate scientific progress, paired with their urgent need for a source of clean, baseload power, put them in a “symbiotic relationship with fusion,” according to Sehila González de Vicente, global director of fusion energy at CATF. AI brings fusion closer to fruition in two broad ways, according to the report: it changes the process of scientific discovery, and it enhances enabling technologies.
“We see a unique opportunity for these technologies to progress hand-in-hand, with AI helping accelerate fusion’s development while fusion provides the clean energy necessary to sustain AI energy demand,” she said in the press release announcing the report.
AI has the potential to significantly shorten the “years-to-decades-long” manual process of “building machines and testing ideas, generating datasets, and then choosing the next design based on empirical scaling,” the report found.
The technology, for example, can enable a scientist to narrow down which fusion designs could be viable and cost-effective. Once they have the conceptual design, they can use AI and HPC to run both short-term and long-term simulations and figure out which adjustments are needed, rather than having to do the testing on physical devices.
Jaydeep Deshpande, principal engineer in the open innovation team at Commonwealth Fusion Systems, told the report’s authors that thanks to HPC, models that would have taken weeks to run now take just days.
AI and HPC could also help fusion developers overcome technical challenges. Much like the technology helped Microsoft find a new battery material in just two weeks, AI can be used to predict material behaviors in the harsh conditions required by fusion, and potentially improve a material’s molecular structure to make it more resistant.
Clean Air Task Force is developing a machine learning-enabled database “that consolidates all fusion-related materials data,” with the goal of reducing the time and money spent on developing new materials.
Other applications of AI in the commercialization of fusion include the optimization of the manufacturing of high-temperature superconductors, and advanced monitoring capabilities and diagnostics.
Commercializing nuclear fusion still remains far off — but its pursuit has attracted interest and money from major investors like Sam Altman and Reid Hoffman, as well as the Biden administration. The number of fusion companies has quadrupled since 2017, and they have raised at least $8 billion in equity, according to the CATF report; of that total, $7 billion has been secured in the past four years.
That’s a lot of money for a technology that, as the report pointed out, some skeptics joke “is thirty years away, and will always be thirty years away.”
Recent announcements include nuclear fusion startup Pacific Fusion Corp, which raised $900 million Series A last month, and Denver-based Xcimer, which raised a $100 million Series A backed by investors such as Breakthrough Energy Ventures and Emerson Collective in June. Commonwealth Fusion Systems, which is considered the richest fusion startup in the world, has raised over $2 billion to date.
If any of these companies achieve the scientific breakthrough of nuclear fusion, the report suggests fusion’s levelized cost of electricity, which includes all costs over a power plant’s lifetime, could be comparable to that of other energy sources, at around 4.7 to 7.9 cents per kilowatt-hour.
