Unleashing innovation to advance American Energy.

What do the power grid, artificial intelligence (AI), fusion energy, and Secretary of Energy Chris Wright all have in common? If you are a user of electricity from the public power grid in the U.S., there is actually quite a lot! According to Secretary Wright’s first secretarial order, the Energy Department’s primary mission is to “promote energy abundance, demonstrate leadership in scientific and technological innovation, steward and strengthen our weapons stockpiles, and meet Cold War legacy waste clean-up commitments.”

One of the ways that Secretary Wright has identified to do this is to “unleash American energy innovation”. For the Trump Energy Department, achieving this revolves around three initiatives:

1. Improving our supply chain security and manufacturing competitiveness.
2. Using existing dispatchable energy sources (fossil fuels, nuclear, geothermal, and hydropower) to increase energy availability while decreasing costs.
3. Investing DoE research and development funds in areas promising true technological breakthroughs, such as high-performance computing, quantum computing, AI, and fusion energy. It is the last of these three initiatives that is most interesting from our perspective.

The Connection Between AI, HPC, and Fusion Energy

Fusion energy, AI, and high-performance computing (HPC) have a uniquely synergistic relationship that mutually benefit each other. Modeling the magnetohydrodynamics of high-temperature, high-pressure plasmas is one of the most difficult computational challenges in physics, and it is critical to the design and optimization and commercialization of fusion energy. This is particularly true for identifying the conditions that can potentially cause instabilities in high-energy fusion fluids, such as turbulence and heat flow within a plasma. HPC and AI also helps to provide ways of avoiding, or “healing” of plasma flows. Finally, MHD modeling has also been found to be critical to predict the physics of fusion “support systems” such as thermal blankets, where the optimal capture of fusion energy into thermal energy (“heat”) is critical, as is avoiding the loss of that thermal energy within the thermal blanket structure.

At the same time, AI computation (even running on HPC clusters) is extremely demanding from an energy perspective. Utilizing AI and HPC to design better fusion systems can (ironically enough) provide the power that AI and HPC require to become more widespread and take on even greater computational problems. An early indicator of the scop of this problem can be seen in Virginia, where more than 25% of the state’s electricity is consumed by datacenters today. Furthermore, generative AI systems are expected to make this much worse, increasing 15%-20% per year through 2030, roughly doubling the power consumed by datacenters. It is not only the additional power draw of AI which is a problem; it is also the destabilizing effect that this draw (which can be very peaky) will have on a grid which is not made to handle such demand. Fusion could provide the extra power that datacenters require to fulfill AI’s demand for power.

AI Can Also Help Stabilize the Grid

As they say “every cloud has a silver lining”, and this is also true of AI and the electrical grid. There are several ways in which AI can “help” the grid:

• Improving management of the power grid: Today the power grid at the state and regional level is managed by a number of independent system operators (ISOs), who “watch” grid demand every few seconds, and make corrections to supply. However, AI can do this much faster, and across multiple points on the grid. This can not only help improve the response to peaky demands, but also will help with the integration of renewable energy resources.
• More effective maintenance of grid infrastructure: In addition to better managing grid supply vs demand, AI can also implement predictive maintenance strategies. Knowing that an infrastructure component is about to fail can help avoid larger grid failures.
• Improved Cybersecurity for the Grid: For some time there have been significant concerns around the cybersecurity of the US power grid, amplified by the number of cyberattacks on various grid elements. Like other cybersecurity issues, AI has the potential to effectively tackle these issued.
• Smart “Behind the Meter” Components for Enterprises: AI can also be used to improve the “private grids” used by large enterprises and manufacturing concerns, such as load balancers, reducing their cost and negative impact on the grid.

Conclusion: Can Chris Wright Make All of This Happen?

As Thomas Edison said, “Genius is one percent inspiration and 99% perspiration”, and this applies as much to executing energy strategies as it does to technological advancements. Making all of this happen across several technologies, while not foolproof, can certainly be helped along by the right investments with Department of Energy R&D funds.