Superconductors: The Supply-Side Story

TL;DR

Five separate industries are converging on the same materials supply chain in the same decade: fusion magnets, grid-scale energy storage, medical imaging, AI compute interconnects, and quantum computing. All of them depend on high-temperature superconducting tape — primarily REBCO, with niobium-titanium and Nb₃Sn for legacy low-temperature applications. Global REBCO production is currently measured in kilometers per year across roughly half a dozen suppliers worldwide.¹ A single fusion power plant absorbs tens to hundreds of kilometers of tape per machine. ITER alone consumed roughly 100,000 km of Nb₃Sn strand for its central solenoid.²

The physics has been solved. The materials science is mature. The factory does not exist at the scale these industries need.

This is the gap an applied materials company sits inside.

§1 — The tape that broke the dam

In 2012, Dennis Whyte’s MIT graduate students ran the numbers on a fusion reactor built around REBCO (rare-earth barium copper oxide) tape and discovered that the device could be roughly 1/40 the volume of ITER for the same fusion power output.³ What had been a generation-defining megaproject became a problem you could fit in a warehouse in Devens, Massachusetts. Commonwealth Fusion Systems was spun out of that homework assignment. SPARC’s first toroidal field magnet was complete by January 2026.

The tape itself is the entire story. REBCO carries current densities orders of magnitude higher than copper at superconducting temperatures, and it does it at 20–77 K — temperatures reachable with liquid nitrogen rather than liquid helium. Strong field, smaller machine, cheaper coolant. The same logic that collapsed ITER into SPARC applies to every other application that needs a strong magnetic field in a small volume.

§2 — The five industries pulling at once

Fusion. CFS, Tokamak Energy, Type One Energy, Proxima, Thea Energetics, and a long tail of stellarator and FRC startups all need REBCO tape (or in TAE’s case, fewer magnets, but still high-field). The cumulative magnet bill for a commercial fusion plant fleet is megameters of tape per year. Global capacity today is not in that league.

Grid storage and protection. Superconducting Magnetic Energy Storage (SMES) and Superconducting Fault Current Limiters (SFCLs) are entering grid deployments. SFCLs, in particular, are getting active utility interest because the alternative — pyrotechnic and resistive limiters — is reaching the limits of what NFPA 855 (battery storage) and updated grid codes can tolerate as renewable penetration rises. Grid-scale superconducting devices are not science fiction; they are commercial-stage hardware that has been commercialized in pockets and is now being asked to scale.

Medical imaging. Every clinical MRI on Earth uses a superconducting magnet. The installed base is roughly 50,000 machines globally, with an annual replacement and expansion volume in the low thousands.⁴ All of those have historically used low-temperature niobium-titanium magnets cooled in liquid helium. The helium supply chain has been brittle for two decades and is getting worse. High-temperature superconducting MRI magnets cooled by closed-cycle cryocoolers (no liquid helium) are an obvious replacement — and the OEMs know it. The transition is slow because the magnet supply isn’t there.

AI compute interconnects and storage. Covered in The Datacenter Water Floor. At sufficient scale, superconducting interconnects collapse the cooling-water budget by orders of magnitude. The required tape volumes look a lot like the fusion volumes.

Quantum computing. Superconducting qubits (Google, IBM, IQM, Rigetti) and the cryogenic plumbing around them aren’t direct REBCO consumers in the way fusion magnets are, but they pull on the same low-temperature infrastructure: dilution refrigerators, niobium thin films, microwave cabling, magnetic shielding. The supply chain serving 10–100 commercial-grade dilution refrigerators per year is a different beast from one serving 10,000.

§3 — Where the floor is

REBCO tape today is produced primarily by SuperPower (US), Faraday Factory (Japan, with manufacturing in Russia historically — supply disrupted post-2022), SuNAM (South Korea), Shanghai Superconductor Technology (China), and a handful of smaller producers. Public capacity figures are imprecise; the rough industry consensus is that global annual REBCO output is in the range of a few thousand kilometers per year.¹

Compare that to copper magnet wire: a single mid-size copper-wire mill ships tens of thousands of tons annually, easily megameters of conductor. The superconducting tape industry is roughly four orders of magnitude smaller than the copper wire industry by length. Closing even one of those orders of magnitude is a multi-billion-dollar capital build.

The bottleneck is not the materials chemistry. REBCO synthesis is well understood, with multiple deposition routes (MOD, PLD, MOCVD) at production scale. The bottleneck is:

1. Substrate availability. REBCO is grown on textured nickel-tungsten or stainless-steel substrates with buffer layers. The substrate supply is a niche industry serving the tape industry — a niche serving a niche.
2. Yield and throughput. REBCO tape is sold in 4-mm wide strips, hundreds of meters long, with strict critical-current uniformity tolerances. A single bad meter in a 500-meter spool can make the spool unusable for a magnet wind. Yield improvement, not capacity, is often the binding constraint.
3. Capital intensity. A new REBCO line is hundreds of millions of dollars and takes 3–5 years to qualify.

This is a supply chain that resembles the early semiconductor industry of the 1970s more than it resembles the mature commodities markets the magnet customers come from.

§4 — Where a materials company fits

Armadillo Labs is not a tape manufacturer. We are not building a REBCO line. The question we ask is one layer deeper: what if the next-generation superconductor isn’t REBCO?

REBCO is the current commercial workhorse. It is not the end state. Higher-Tc, higher-field, easier-to-grow superconductors would change the magnet economics of every industry above. Hydrogen-rich superhydrides under pressure achieve room-temperature superconductivity but at hundreds of gigapascals — useless for engineering today, but a forcing function for the search.⁵ Cuprates and nickelates have been the source of every Tc record above ~30 K for forty years. There are still families that haven’t been screened.

Our screening tools — phonon-mediated transport models, electron-phonon coupling estimators, and Materials Project integrations — are built to find and triage candidate superconductors at speed. Most candidates fail. That’s the point. The ones that survive a multi-stage screen are the ones worth the DFPT and the experimental synthesis. The economics of the magnet industry — and the cooling-water economics of the datacenter industry, and the helium economics of MRI — make this search worth doing on its own merits.

The supply-chain story for superconductors is real. It is also the story of a single material class. The deeper bet is that the material class itself is going to broaden in the next decade. Whoever finds the next workhorse owns the next twenty years of the magnet industry.

---

Footnotes

1. Aggregate from public disclosures by SuperPower (Furukawa subsidiary), SuNAM, Shanghai Superconductor Technology, Faraday Factory Japan, and U.S. DOE / ARPA-E REBCO supply chain workshop reports, 2022–2024. ↩ ↩²

2. ITER Organization, “Magnets” technical overview, https://www.iter.org/mach/magnets — central solenoid Nb₃Sn strand cumulative procurement. ↩

3. Sorbom, B. N. et al., “ARC: A compact, high-field, fusion nuclear science facility and demonstration power plant with demountable magnets” Fusion Engineering and Design 100, 378–405 (2015). Earlier MIT design exercises (2012–2014) established the volume scaling. ↩

4. IMV Medical Information Division, “MR Market Outlook Report” (annual series); industry-standard installed-base figure of ~50,000 clinical MRI systems globally. ↩

5. Drozdov, A. P. et al., “Conventional superconductivity at 203 kelvin at high pressures in the sulfur hydride system” Nature 525, 73–76 (2015). Subsequent LaH₁₀ and YH₉ results, 2019–2021. ↩