AI Data Centers Demand Revolutionary Approach To Power: The Case For Hybrid SuperCapacitors

For the first time in more than a decade, U.S. electricity consumption is on the rise. Between the mid-2000s and early 2020s, electricity demand remained mostly flat. Now, it is increasing and expected to grow at an average rate of 1.7% per year, driven in no small part by the expanding data center industry.

While legacy data centers were measured in processing power or storage capacity, today’s facilities are measured in megawatts. It’s no surprise that U.S. data centers consumed 183 terawatt-hours of electricity in 2024, roughly 4% of the country’s total electricity consumption. By 2030, that number is predicted to increase by 133%, largely due to the rise of artificial intelligence.

“AI data centers consume power fundamentally differently than traditional data centers,” said Alan Forster, senior director of sales for Musashi Energy Solutions.

Most energy systems have been designed to protect the load — be it computers, medical equipment and so on — from the power source, Forster said. With AI, that remains true, but designers must now also consider protecting the source — meaning the grid and power generation components — from the load itself, which presents increasingly complex management challenges.

AI workloads operate differently from traditional data center applications, Forster said. While traditional data centers have components that operate independently, AI systems work in tandem with the graphics processing and server activities, rising and falling synchronously based on workload demands. This means that even minor issues can rapidly escalate into significant problems due to dynamic shifts in usage.

“Large power peaks and valleys are generated multiple times per second,” Forster said. “Power is moving around so much, so fast, and due to the synchronous behavior of AI power, it’s exceeding traditional limits and technology designs currently available, causing tremendous strain.”

According to Forster, effective power solutions for AI data centers must meet three critical requirements: high power delivery, high resiliency and high safety standards 

Batteries function adequately, but while they provide superior energy capacity, they lack the necessary response speed due to lengthy charge and recharge cycles and have limited operational lifespans.

Lithium batteries offer improvements over lead-acid batteries. While lead-acid battery life is measured in hundreds of charge/recharge cycles, lithium batteries achieve thousands of cycles. However, this still falls short for AI applications, which require tens of millions of cycles or more. Lithium batteries also pose safety risks, including thermal runaway, which can lead to fires or explosions.

In response, the data center and power industries are exploring innovative approaches to meet these demanding energy requirements. One approach, Forster said, involves scaling up existing solutions. However, this strategy amplifies the inherent limitations of current systems: reduces sustainability, increases safety risks, adds complex maintenance requirements, and higher CapEx and OpEx costs.

“Rather than simply oversizing the infrastructure, sustainable solutions are needed that actively work to level power in real time,” Forster said. “That’s where Hybrid SuperCapacitors come in.”

Hybrid SuperCapacitors offer exceptional resilience and high power, react in microseconds, and have significantly extended cycle life, thousands of times greater than lithium batteries, all without the risk of thermal runaway. 

Musashi’s patented UL-listed technology has led this development for more than a decade. Major industry leaders are currently deploying Musashi’s Hybrid SuperCapacitors in their data center racks, and the technology continues to evolve in response to advancing industry requirements.

“Google has a reference design for a one-megawatt rack that will require a whole new power infrastructure,” Forster said. “Hybrid SuperCapacitors will play an important part in managing the power in the most demanding loads today and in the future.” 

“The objective is to enable AI data centers to operate safely and reliably to support the requirements of AI-driven infrastructure,” he said. “Musashi will play an essential role in the infrastructure of the future.”

This article was produced in collaboration between Musashi Energy Solutions and Studio B. Bisnow news staff was not involved in the production of this content.

Studio B is Bisnow’s in-house content and design studio. To learn more about how Studio B can help your team, reach out to studio@bisnow.com