Bringing artificial intelligence to the next generation of fusion energy

We are working with Commonwealth Fusion Systems (CFS) to bring the reality of clean, safe and unlimited fusion energy closer to the world.

Fusion, the process that powers the sun, provides clean, abundant energy without long-lived radioactive waste. To make it work here on Earth, it means maintaining the stability of the ionized gas, called plasma, at temperatures exceeding 100 million degrees Celsius – all within the limits of a fusion energy machine. This is a very complex physics problem that we are working on using artificial intelligence (AI).

Today we announce the establishment of research cooperation with the company Commonwealth Fusion Systems (CFS), a world leader in fusion energy. CFS is pioneering a faster path to clean, safe and effectively unlimited fusion energy with its compact, powerful tokamak machine called SPARC.

SPARC uses powerful, high-temperature superconducting magnets and aims to be the first-ever magnetic fusion machine to generate net fusion energy – more energy from nuclear fusion than is needed to sustain it. This breakthrough is known as break-even and is a key milestone on the path to profitable fusion energy.

This partnership builds on our groundbreaking work using artificial intelligence to effectively control plasma. With academic partners in Swiss Plasma Center at EPFL (École Polytechnique Fédérale de Lausanne)we have shown that deep reinforcement learning can control tokamak magnets to stabilize complex plasma shapes. To cover a broader scope of physics, we have developed CHESTa fast and differentiable plasma simulator written in JAX.

We are now turning this work over to CFS to accelerate the timeline for delivering fusion power to the grid. So far, we have cooperated in three key areas:

• Create fast, accurate and differentiable fusion plasma simulations.
• Finding the most efficient and reliable path to maximizing fusion energy.
• Using reinforcement learning to discover novel control strategies in real time.

The combination of our artificial intelligence expertise with CFS's cutting-edge equipment makes this an ideal partnership to advance fundamental discoveries in fusion energy for the benefit of the global research community and, ultimately, the world.

Fusion plasma simulation

To optimize the operation of the tokamak, we must simulate the flow of heat, electricity and matter through the plasma core and the interaction with the systems surrounding it. Last year we released TORAX, an open-source plasma simulator built for optimization and control, expanding the range of physics issues we could address beyond magnetic simulation. TORAX is built in JAX so it can easily run on both CPUs and GPUs and can seamlessly integrate AI-powered models, including oursto achieve even better performance.

TORAX will help CFS teams test and refine operational plans by running millions of virtual experiments before SPARC is even enabled. It also gives them the flexibility to quickly adjust their plans once they receive the first data.

This software has become the backbone of CFS's daily work, helping it understand how plasma will behave under different conditions, saving valuable time and resources.

Finding the fastest way to maximum energy

Operating a tokamak involves countless adjustments to the various “knobs” available, such as magnetic coil current, fuel injection, and heat output. Manually finding the optimal tokamak settings to produce the most power while maintaining operating limits can be very inefficient.

By using TORAX in combination with reinforcement learning or evolutionary search methods such as AlphaEvolve, our AI agents can explore a vast number of potential operating scenarios in a simulation, quickly identifying the most efficient and reliable paths to net energy generation. This can help CFS focus on the most promising strategies, increasing the likelihood of success from day one, even before SPARC is fully commissioned and operating at full capacity.

We are building infrastructure to explore different SPARC scenarios. We can look at maximizing the fusion power produced under various constraints or optimizing for endurance as we learn more about the machine.

Here we illustrate examples of a standard SPARC pulse simulated in TORAX. Our AI system can evaluate many possible triggers to find the settings we think will work best.

Through our growing network of collaborations within the fusion research community, we will be able to validate and calibrate TORAX against previous tokamak data and high-fidelity simulations. This information will provide confidence in the accuracy of the simulations and help us adapt quickly once SPARC becomes operational.

Developing an AI remote control for real-time control

In our previous work, we showed that reinforcement learning can control the magnetic configuration of a tokamak. We are now increasing the complexity by adding simultaneous optimization of more aspects of the tokamak's performance, such as maximizing fusion power or managing the SPARC thermal load, so that it can operate at high efficiency with greater margin over the machine's limitations.

Operating at full power, SPARC will release enormous heat concentrated in a small area that must be carefully managed to protect the solid materials closest to the plasma. One strategy SPARC can use is to magnetically move exhaust energy along the wall, as shown below.

In the initial phase of our collaboration, we are exploring how reinforcement learning means can learn to dynamically control the plasma to efficiently distribute this heat. In the future, AI may be able to learn adaptive strategies more complex than anything an engineer could create, especially when balancing multiple constraints and goals. We could also use reinforcement learning to quickly tune traditional control algorithms to a specific impulse. The combination of pulse optimization and optimal control can push SPARC further and faster to achieve its historic goals.

Combining AI and Fusion to build a cleaner future

In addition to our research, Google invested in CFSsupporting their work on promising scientific and engineering breakthroughs and aiming to commercialize their technologies.

Looking to the future, our vision goes beyond optimizing SPARC operations. We are laying the foundations for artificial intelligence to become the intelligent, adaptive system at the heart of future fusion power plants. This is just the beginning of our journey together and we hope to share more details about our collaboration as we reach further milestones.

By combining the revolutionary potential of artificial intelligence and nuclear fusion, we are building a cleaner and more sustainable energy future.