The relentless growth of artificial intelligence is placing unprecedented strain on global power grids, and the companies building the infrastructure to support it are scrambling for solutions. Traditional power transmission and distribution networks are proving insufficient to meet the soaring energy demands of AI data centers, leading tech giants like Microsoft to explore innovative technologies like high-temperature superconductors (HTS) to dramatically improve efficiency and capacity. The challenge isn’t simply generating more power, but delivering it reliably and sustainably to these energy-intensive facilities.
According to the U.S. Energy Information Administration (EIA), annual transmission and distribution losses in the U.S. Average around 5 percent, a figure that climbs significantly in other parts of the world. This inherent inefficiency, coupled with the exponential increase in demand from AI, is driving a search for alternatives. Microsoft, along with Amazon Web Services and Google Cloud, is investing heavily in research and development to overcome these hurdles. The core of this effort centers on HTS, a technology that promises to revolutionize how power is transmitted and utilized within and around data centers.
Superconductors: A Potential Game Changer for Data Center Efficiency
High-temperature superconductors offer a compelling solution to the limitations of conventional copper wiring. Unlike copper, which encounters resistance as electricity flows through it—resulting in energy loss as heat—superconducting materials exhibit almost no electrical resistance when cooled to cryogenic temperatures. This means significantly less energy is wasted during transmission, and a much higher current can be carried in a smaller space. “As superconductors take up less space to move large amounts of power, they could help us build cleaner, more compact systems,” explained Alastair Speirs, general manager of global infrastructure at Microsoft, in a blog post.
The benefits extend beyond efficiency. HTS cables are smaller and lighter than their copper counterparts, reducing the physical footprint of power infrastructure. They also minimize voltage drops, ensuring a more stable and reliable power supply. Microsoft estimates that next-generation superconducting transmission lines could deliver capacity an order of magnitude higher than conventional lines at the same voltage level. This increased capacity is crucial for accommodating the ever-growing power needs of AI data centers.
Veir and the Advancement of HTS Technology
To accelerate the development and deployment of HTS technology, Microsoft has invested $75 million in Veir, a company specializing in superconducting power technology. Veir’s conductors utilize HTS tape, primarily composed of rare-earth barium copper oxide (REBCO), a ceramic superconducting material deposited as a thin film on a metal substrate. “The key distinction from copper or aluminum is that, at operating temperature, the superconducting layer carries current with almost no electrical resistance, enabling very high current density in a much more compact form factor,” said Tim Heidel, Veir’s CEO and co-founder.
The Challenge of Cryogenic Cooling
While “high-temperature” superconductors operate at warmer temperatures than their traditional counterparts, they still require cryogenic cooling to function. Veir addresses this challenge with a closed-loop liquid nitrogen system, circulating the coolant through the length of the cable, re-cooling it, and recirculating it. Heidel emphasized that liquid nitrogen is a readily available, cost-effective, and safe material widely used in various industrial applications. Veir prefers to manage the cooling system externally, feeding liquid nitrogen lines into the data center to minimize the internal footprint and operational complexity.
The economics of HTS are currently a limiting factor. The rare earth materials and complex cooling systems add significant costs. But, Heidel argues that the technology is most compelling in situations where power delivery is constrained by space, weight, voltage drop, or heat. “In those cases, the value shows up at the system level: smaller footprints, reduced resistive losses, and more flexibility in how you route power,” he explained. He anticipates that costs will decrease as manufacturing scales up and standardization improves.
AI Data Centers as a Proving Ground
AI data centers are uniquely positioned to drive the adoption of HTS technology. Hyperscalers are willing to invest in higher-efficiency systems to offset the substantial energy costs associated with AI workloads. Husam Alissa, Microsoft’s director of systems technology, noted that HTS manufacturing has matured, improving both cost and supply availability. “Our focus currently is on validating and derisking this technology with our partners with focus on systems design and integration.”
The increasing demand for computing power, as highlighted by the EIA’s projections that computing will account for 20% of commercial sector electricity consumption by 2050, underscores the urgency of finding more efficient power delivery solutions. Data center computing is already significantly more energy intensive than general computing, and this gap is expected to widen. The adoption of HTS, alongside other energy-saving measures, will be critical to mitigating the environmental impact of the AI boom and ensuring a sustainable future for the technology.
Microsoft and its partners are continuing to refine and test HTS technology, with a focus on system design and integration. The next steps involve validating the long-term reliability and scalability of these systems in real-world data center environments. As the technology matures, it could pave the way for a more efficient and sustainable power infrastructure for the rapidly expanding world of artificial intelligence.
What are your thoughts on the potential of high-temperature superconductors to address the energy challenges of AI? Share your comments below and let us realize how you think this technology will impact the future of data centers.
