As AI infrastructure demands push rack power density into the hundreds of kilowatts, data center engineers face a clear bottleneck: conductor capacity. In response, the industry has charted two dominant paths — raising direct current voltages to 800V (led by Nvidia) and disaggregating rack components (led by Microsoft and Meta).

Both approaches are bold and valuable, reducing copper use and adding modularity. But they share a core assumption: there aren’t any better conductor options.

This isn’t true anymore.

A recently matured class of conductor, high-temperature superconductors (HTS), offers a third pathway. Deployed at scale around the world, HTS cables have near-zero electrical resistance, enabling radical reductions in voltage levels, copper usage, cable footprint, and energy losses. It is time for data center designers to bring this technology into the conversation.

The hidden cost of copper

According to Nvidia, delivering 1MW per rack at 54 VDC would require 200kg of copper per rack. Copper is not only becoming supply-constrained and unsustainable, but the weight, space, and thermal losses are untenable. The industry approaches of voltage escalation and disaggregation will help, but they introduce new challenges - 800VDC is harder to interrupt and requires high-voltage qualified technicians, and both approaches add complexity and compatibility issues with low-voltage IT hardware. The current workstreams are engineering trade-offs, not clean solutions.

Why now? What’s changed for HTS

Superconductors aren’t new, but commercial deployment has rapidly accelerated this decade.

• Nexans recently powered a station in the Paris metro system and is building upon more than a decade of cable deployment experience.
• South Korea has utility-connected HTS cable systems operating today, designed by cable giant LS Cable.
• Vision Electric Superconductors recently demonstrated a 200 kA HTS busbar for an aluminum smelter, highlighting the industrial readiness of high-current use cases.
• And the industry is broadly scaling: companies like VEIR (which recently raised $75m with National Grid and Microsoft backing), TE Magnetics, Supernode, NKT, and others are bringing fresh manufacturing capacity online.

The capabilities and potential of superconductors are now reaching data centers. The 2025 Superconductivity for Data Centers Workshop revealed that nearly a dozen superconductor companies are building data center-targeted solutions: from Nexans and VEIR’s MV cabling designs to LS Cable’s first public utility-connected data center project.

Advantages for data center designers

HTS solves the power density problem at the root. It delivers: a more than 80 percent reduction in cable footprints, in addition to zero-resistive losses and low voltage operating levels. Increased efficiency and reduced PUE without cooling excess cable heat is another benefit of the technology, along with the ability to centralize high current distribution in low-voltage, dense buses, without needing to redesign IT hardware for high-voltage.

In short: denser, safer, and more efficient.

Implementation is underway

Data center operators are already designing with 3000+ amp HTS cables—not for distant use cases, but to simplify substation and campus-level MV designs today.

That same momentum can carry HTS all the way to the rack. Most deployed HTS cables today terminate in a single cooling module. However, data center racks will require dozens of tap-offs for each individual rack. That adaptation may take two pathways:

1. New termination systems, built into modular cryostats. Multiple HTS providers, including Vision Electric Superconductors, are exploring development for this challenge.
2. Cryogenic-native compute: Fully immersing compute in LN₂, eliminating the need to bring power back to room temperature. It’s radical, but not unprecedented — the Cray-2 computers in the 1980s and quantum computers today already operate in cryogenic environments.

Either approach can be designed to leverage the same cooling infrastructure already being implemented for larger-cable HTS adoption, and to potentially help other systems like cooling and fire suppression.

Where to go from here

This isn’t a call to abandon high-voltage or disaggregated systems. They are vital workstreams and will unlock valuable lessons and designs. But if the industry invests in them alone, it risks optimizing around copper’s limits instead of moving past them. It is time to move beyond legacy technologies to meet the new challenges presented by today’s power-hungry AI data centers. Let’s design with a conductor built for the AI era.