The real reason Google DeepMind partners with fusion energy startups

Energy startup Commonwealth Fusion Systems announced Thursday that it is working with Google’s DeepMind division to use AI to fine-tune and even improve the operation of future Sparc reactors.

The companies’ plan is to use specialized DeepMind software, known as Torax, to simulate the plasma burning inside the CFS reactor. They also plan to combine Torax with AI models to help CFS find the best way to achieve fusion power generation.

Fusion power promises to deliver large amounts of electricity with zero emissions from a nearly limitless fuel source: water. AI companies are bullish on fusion startups as a power source to power energy-hungry data centers. Google seems to be eyeing them as potential customers as well.

This isn’t Google’s first foray into nuclear fusion. The technology company collaborated with another fusion startup, TAE Technologies, to use AI to study how plasma behaves inside TAE’s fusion machines.

There’s a reason Google keeps coming back to this issue. Because AI may be uniquely suited to enable fusion power generation.

One of the biggest challenges facing fusion startups is keeping the plasma inside the reactor hot enough for long enough. Unlike nuclear fission reactions, which are self-sustaining, fusion reactions are difficult to maintain outside a star like the Sun. Without such mass and gravity, the plasma is always at risk of spreading out and disappearing.

In CFS reactors, powerful magnets replace gravity to keep the plasma contained, but it’s not perfect. Reactor operators must develop control software that allows the equipment to continuously respond to changing plasma conditions.

The problem is that there are almost too many knobs to turn, certainly beyond human ability. That’s the kind of problem that AI is good at. Experts cite AI as one of the key technologies that have enabled the industry’s impressive advances over the past few years.

CFS is currently building a demonstration reactor, Sparc, in a suburban area outside Boston. The device is about two-thirds complete and, when completed in late 2026, the company predicts it will be the first fusion device to generate more electricity than is needed to run the plant itself.

Google said Torax can be used in conjunction with reinforcement learning or evolutionary search models to find “the most efficient and robust paths to producing net energy.” The two companies are also exploring whether AI can be used to control the operation of nuclear reactors.

In August, Google participated in CFS’ $863 million Series B2 round along with NVIDIA. Google also announced earlier this year that it would purchase 200 megawatts of power from CFS’ first commercial power plant, Ark, to be built outside Richmond, Virginia. The technology company is also an investor in TAE Technologies, a competitor of CFS.