Beyond Copper: The Conductors, Coolants, and Light Links Chasing a 150-Year-Old Metal

America's New Invention Could Replace Copper Forever - YouTube

Materials science & power systems

A wave of 2025–2026 results shows real ways to move charge, heat, and data without copper. None of them retire the metal — and the hype badly outruns the physics.

Feature · Analysis Updated September 2026

BLUF

Copper is not becoming obsolete, but for the first time in a century it faces credible competition in specific, high-value niches. Between late 2025 and mid-2026, peer-reviewed results and shipping products emerged in four areas where copper is the bottleneck: chip interconnects (Cornell's niobium-arsenide nanowires), heat removal (UCLA's record-setting θ-phase tantalum nitride), lightweight bulk conductors (carbon-nanotube fibers from IMDEA in Madrid; copper–graphene composites from PNNL), and data movement (Avicena's microLED optical links that replace copper GPU cabling with light).

The pressure is real: copper set record prices near US$12,000–14,000/tonne in 2025–2026, a 50% U.S. import tariff took effect in August 2025 on national-security grounds, and demand is forecast to climb from ~28 to ~42 million tonnes a year by 2040.

But the popular framing — that "American labs" made copper "obsolete" in 18 months — is wrong on the facts. Several headline results are foreign (Australia, Spain) or years old; the flashiest are lab proofs-of-concept with toxic elements, non-manufacturable processes, or low energy density. The accurate story is substitution at the margins plus continued copper growth, not replacement.

For roughly 150 years, copper has been the default answer to a simple engineering question: how do you move electricity from here to there cheaply? Only silver conducts better among common metals, and silver is far too expensive to string across a continent. So copper carries the current in building wiring, motor windings, transformers, undersea cables, and the interconnects inside every microprocessor.

What changed recently is not copper's chemistry but its context. The electrification of transport and the explosive buildout of AI data centers have collided with a mining industry that cannot expand fast enough. Copper broke successive price records through 2025, trading above roughly US$12,000 per tonne by December and spiking past US$13,000–14,000 in early 2026 — gains of more than 30% on the year.¹ Analysts at S&P Global project global demand rising from about 28 million tonnes in 2025 toward 42 million tonnes by 2040, with a potential shortfall of several million tonnes annually absent new supply.²

Washington now treats this as a security problem, not just a commodity cycle. Following a Section 232 investigation, the United States imposed a 50% tariff on semi-finished copper and copper-intensive derivatives effective 1 August 2025, after Commerce found that import dependence threatened national security.³ The administration's own justification described copper as the Defense Department's second-most-used material, essential to semiconductors, ammunition, data centers, and — in language that will resonate with anyone who has built airborne sensors — radar and missile-defense systems.⁴ The U.S. imports roughly 45% of its copper, and while it exempted refined cathode for now, Commerce owes the President a refined-copper market assessment by 30 June 2026 that could trigger phased duties starting in 2027.³ China, meanwhile, controls roughly half of global copper smelting and refining — the choke point that makes the supply chain a strategic vulnerability.

That combination — scarce metal, strategic exposure, insatiable demand — is what has revived a research question most engineers had filed away as settled: can anything actually beat copper? The federal government put money behind it in 2021 with the Department of Energy's CABLE Prize (Conductivity-enhanced materials for Affordable, Breakthrough Leapfrog Electric and thermal applications), a US$4.5-million, three-stage competition administered by NREL whose explicit goal was to develop conductors that match or beat copper at lower cost or weight.⁵ Several of the results below trace directly to that program or to the national labs it energized.

Interconnects · AI hardware

A wire that conducts better as it shrinks

The most physics-bending result of the past year came from Cornell. Inside a modern processor, billions of transistors are wired together by copper interconnects only nanometers wide, and there copper fails in a way that is fundamental rather than incidental: as the wire narrows, electrons scatter off its surfaces, and resistivity climbs steeply. Interconnect resistance, not transistor speed, is now a leading limiter of chip performance.

A team led by materials scientist Judy Cha, with doctoral student Yeryun Cheon as lead author, reported in Science on 16 July 2026 that single-crystal nanowires of niobium arsenide (NbAs) — a topological "Weyl semimetal" — do the opposite.⁶ In these materials, a population of electrons flows along the wire's surface with little scattering. Shrink the wire and you increase the share of current carried by those well-behaved surface states, so resistivity drops as the diameter falls. Using a technique called thermomechanical nanomolding, the group made single-crystal wires as thin as 40 nm and measured room-temperature resistivity about 70% below that of bulk NbAs.⁷

This is where a careful reading matters, and where the viral version of the story overreaches. As IEEE Spectrum noted in its own coverage, the 40-nm NbAs wires beat same-size cobalt or ruthenium — the industry's leading candidates below copper's practical floor — but they did not outperform state-of-the-art 10-nm copper interconnects. The researchers project that NbAs wires would need to reach roughly 12 nm to cross that line.⁸ And Cha herself is blunt about the caveats: arsenic is toxic, and the nanomolding synthesis is not compatible with the CMOS back-end-of-line processes that fabs actually use. NbAs is a proof of principle that topological semimetals are engineering-relevant, not a copper replacement heading to a fab.

Thermal management · Data centers

The metal that moves heat like a diamond

Copper's other job is to carry heat away, and here it has been the benchmark at about 400 W/m·K for over a century — good enough to line the heat sinks and cold plates in every laptop, server, and inverter. The intrinsic limit on any metal's thermal conductivity comes from how phonons and electrons scatter as they carry energy through the lattice.

A UCLA-led team under Yongjie Hu, working with Argonne National Laboratory and reporting in Science in January 2026, produced single crystals of metallic θ-phase tantalum nitride (θ-TaN) and measured a room-temperature thermal conductivity of about 1,100 W/m·K — nearly three times copper's, and comparable to low-grade diamond.⁹ Synchrotron x-ray scattering revealed why: θ-TaN's crystal structure has an unusually large gap between acoustic and optical phonon branches, which suppresses the phonon–phonon scattering that normally caps heat flow, while its electron–phonon coupling is exceptionally weak.¹⁰ Physics World called it a record that had stood for more than a hundred years, now broken.¹¹

~400Copper thermal conductivity, W/m·K

~1,100θ-TaN thermal conductivity, W/m·K⁹

~2,100Boron arsenide, for context¹⁰

The application is obvious: heat is the wall in AI data centers. If a hot-spot spreader moved three times the heat flux of copper, designers could pack far more compute into the same rack before thermal limits bit. But θ-TaN today is a millimeter-scale single crystal grown under demanding conditions, tantalum is an expensive refractory metal with its own supply concerns, and no one has shown a manufacturable coating or spreader. It reframes the textbook limit; it does not yet cool anything you own.

Bulk conductors · Motors & grid

Doping copper, and doing without it

Not every advance is exotic. At the Pacific Northwest National Laboratory, Keerti Kappagantula's group used a solid-phase process called ShAPE (Shear-Assisted Processing and Extrusion) to blend a trace of graphene — about 18 parts per million — into electrical-grade C11000 copper. The widely circulated claim that this "cuts resistance by 11%" is a misreading. What the team actually measured, published in Materials & Design, was an ~11% reduction in the temperature coefficient of resistance, with electrical conductivity essentially unchanged (up ~1%).¹² That distinction is the whole point: a lower temperature coefficient means the wire loses less conductivity as it heats. Motor windings are today derated to stay in a safe temperature band; a copper composite that holds conductivity when hot could let motors run hotter and harder. PNNL's illustrative figure — that an 11% conductivity gain would translate to roughly 1% motor efficiency — describes the payoff if such a gain were achieved, not a result already in hand.

The bigger structural opportunity is aluminum. It carries about 60% of copper's current per unit area but weighs a third as much and is vastly more abundant, which is why the long-distance grid already runs largely on aluminum. Boosting aluminum toward copper's conductivity — via graphene composites and advanced extrusion — is an active target across the national labs and the CABLE Prize field, precisely because the grid is where tonnage, not nanometers, decides the outcome.

Carbon fibers · Aerospace & EVs

Cables spun from carbon, half the weight

The most credible near-term challenger for weight-sensitive wiring isn't a metal at all. Carbon nanotubes conduct superbly as individuals, but for two decades no one could spin them into macroscopic fibers that kept those properties — impurities and misalignment wrecked the conductivity. In April 2026, a team at the IMDEA Materials Institute in Madrid, led by Juan Vilatela, reported in Science a scalable fix: gas-phase intercalation of tetrachloroaluminate (AlCl₄⁻) ions into the channels between highly aligned double-walled nanotubes, which dopes the fiber without disrupting its structure.¹³

The treated fibers reached room-temperature conductivity of 24.5 MS/m — about 41% of copper's — at roughly one-sixth the weight, and on a per-mass ("specific conductivity") basis they exceed both copper and aluminum.¹⁴ They are also about five times stronger than conventional overhead cable, though moisture degrades them and they need a polymer sheath to hold ~80% of their conductivity. For aircraft, drones, and EVs — where a commercial airliner carries hundreds of kilometers of wiring and every kilogram costs fuel — a conductor at half copper's weight that meets industrial thresholds is a serious proposition. In the United States, Matteo Pasquali's lab at Rice has pursued the same goal from the fiber-spinning side, previously passing 10 MS/m and reaching roughly 80% of copper's conductivity by weight.¹⁵

Energy storage · EVs

A battery with no copper — and no lithium

Every lithium-ion cell contains copper: thin foil current collectors, miles of it in a full EV pack. The Graphene Manufacturing Group (GMG) — an Australian company working with the University of Queensland, Rio Tinto, and the Battery Innovation Center in Indiana — is building a graphene aluminum-ion cell that uses aluminum foil for both electrodes and contains no copper, no lithium, and no cobalt.¹⁶ Its signature figure is charge speed: full charge in about six minutes.

That headline is real, but it comes with an asterisk the video versions omit. As of an April 2026 update, GMG's six-minute cells delivered about 49 Wh/kg, rising to ~101 Wh/kg at a one-hour charge, with targets above 80 and 160 Wh/kg respectively.¹⁷ For comparison, today's automotive lithium-ion cells sit around 250–300 Wh/kg. GMG's chemistry is a fast-charge, abundant-materials play, not a range play, and it remains at battery technology readiness level 4 — laboratory validation. The company targets customer testing in 2026 and small-scale production in 2027. It is a genuine copper- and lithium-free architecture worth watching; it is not, yet, the end of "range anxiety."

Photonics · Telecom & AI

When the answer is to stop using wires

The fastest-moving copper displacement is already shipping, and it works by abandoning electrical conduction entirely. Inside AI clusters, GPUs talk to each other over short, thick copper cables whose reach is limited to about a meter and whose bandwidth and heat are becoming the binding constraint as clusters scale to tens of thousands of accelerators.

Avicena, based in Sunnyvale with a development center in Edinburgh, replaces those copper links with light — but not with lasers. Its LightBundle platform uses dense arrays of microLEDs coupled through multi-core fiber to integrated photodetectors. In August 2026 the company began shipping terabit-class evaluation kits with a 335-channel microLED array running up to 3 Gb/s per channel, at sub-picojoule-per-bit efficiency and shoreline density above 1 Tb/s per millimeter, extending GPU-to-GPU reach past 10 meters.¹⁸ TSMC has partnered with Avicena to build the interconnects — a signal that the "copper wall" in scale-up AI networking is being taken seriously by the industry's most important foundry.¹⁹ (The transcript that prompted this piece calls the maker "Copen Corporation" and the product "neural IO"; neither name checks out. The verified company and technology are Avicena and LightBundle.)

The same logic is retiring copper at continental scale in telecom. In March 2026 the FCC adopted an order (FCC 26-19) streamlining the retirement of legacy copper networks — a 31-day automatic-approval window and federal preemption of slower state rules.²⁰ AT&T, which spends nearly US$6 billion a year maintaining copper that carries fewer than 3% of its customers, has FCC clearance to discontinue legacy service across more than 30% of its ~4,600 copper wire centers, effective in late 2026, on the way to retiring essentially all of it by 2029; the copper is being replaced with fiber and fixed wireless.²¹ Verizon, Lumen, and Fidium are pursuing parallel retirements.²² None of this is a new material beating copper — it is the deliberate removal of copper from an entire access network.

The 2025–2026 field, measured against copper. "Readiness" is the author's assessment, not a formal TRL except where noted.

Material / system	Key metric vs. copper	Lead institution	Readiness
NbAs nanowire	↓70% resistivity vs bulk; not yet < 10-nm Cu	Cornell (Cha/Cheon)	Lab proof
θ-TaN (heat)	~1,100 vs ~400 W/m·K	UCLA (Hu) / Argonne	Single crystal
Cu–graphene wire	↓11% temp. coeff.; ~1% conductivity	PNNL (Kappagantula)	Extruded samples
CNT fiber	24.5 MS/m = 41% Cu; ⅙ weight	IMDEA Madrid (Vilatela)	Scalable process
Graphene Al-ion cell	no Cu/Li; ~49–101 Wh/kg; 6-min charge	GMG / U. Queensland (AUS)	BTRL 4
microLED optical link	replaces Cu; <1 pJ/bit; >10 m reach	Avicena (US/UK) + TSMC	Shipping eKits

What the Video version gets wrong

It's not all "American labs," and not all recent. The standout carbon-nanotube fiber is Spanish (IMDEA, Madrid); the six-minute battery is Australian (GMG / University of Queensland). The CABLE Prize dates to 2021, and the PNNL copper–graphene work was published in 2023–2024 — not a sudden 18-month burst.

The strongest results are proofs of concept. NbAs contains toxic arsenic and can't be made in a CMOS fab; θ-TaN is a lab single crystal of an expensive refractory metal; the GMG cell stores a third to a fifth of lithium-ion's energy. "Better than copper" in a paper rarely means "manufacturable, safe, and cheaper" in a plant.

Copper isn't going obsolete. Even in aggressive electrification scenarios, copper demand is projected to rise by half over the next 15 years. The realistic near-term picture is copper being trimmed at the margins — in weight-critical wiring, in nanoscale interconnects, in AI cabling swapped for light — while the world still needs more of it overall. Two claims in the source transcript could not be verified at all: a figure of "41% better conductivity and 450% more current" for copper–graphene, and a report attributed to an "Advanced Carbons Council." Treat both as unsupported.

The honest synthesis is more interesting than the hype. What 2025–2026 actually delivered is a portfolio: a topological interconnect that rewrites a textbook scaling law, a metal that conducts heat like a gemstone, a carbon fiber that finally meets industrial conductivity thresholds at a fraction of copper's weight, a battery architecture that sidesteps two contested supply chains at once, and a photonic link that is already in customers' labs. Each attacks a different facet of copper's dominance, and each is early. Copper will keep carrying most of the world's current for a long time. But for the first time since the age of the telegraph, the engineering question "what if not copper?" has more than one serious answer.

References

All sources verified against primary press releases, peer-reviewed papers, and official filings. URLs current as of September 2026.

1. Copper 2025–2026: Record-high prices and structural shortage. AuAg Funds, June 2026. auagfunds.com/research-centre/publications/critical-insights-copper
2. Copper surges in metals frenzy; S&P Global demand outlook (28→42 Mt by 2040). Yahoo Finance, 2026. finance.yahoo.com/news/copper-surges-in-unsustainable-rally
3. Section 232 National Security Tariffs on Copper Imports. Congressional Research Service (IN12614), 23 April 2026. congress.gov/crs-product/IN12614
4. President Trump Announces 50% Tariff on Copper. ArentFox Schiff legal alert, 2025. afslaw.com/perspectives/alerts/president-trump-announces-50-tariff-copper
5. DOE Opens Stage 2 of the CABLE Conductor Manufacturing Prize. U.S. Department of Energy, AMMTO. energy.gov/eere/ammto/articles/doe-opens-stage-2-cable-conductor-manufacturing-prize
6. Too thin to fail: an alternative to copper microchip interconnects. Cornell Chronicle, 17 July 2026. news.cornell.edu/stories/2026/07/too-thin-fail-alternative-copper-microchip-interconnects
7. Y. Cheon et al., Surface-dominant transport in Weyl semimetal NbAs nanowires for next-generation interconnects. Science, 16 July 2026. Summary via phys.org: phys.org/news/2026-07-thinner-wires-faster-electrons-quantum.html
8. Topological Materials Could Shrink Chip Interconnects. IEEE Spectrum, July 2026. spectrum.ieee.org/topological-material-nanowire-interconnect
9. UCLA-led team discovers metallic material with record thermal conductivity. UCLA Newsroom, 20 Jan 2026. newsroom.ucla.edu/dept/faculty/tantalum-nitride-record-thermal-conductivity-ucla-research
10. C. Li et al., Metallic θ-phase tantalum nitride has a thermal conductivity triple that of copper. Science 391, 707 (2026). DOI: 10.1126/science.aeb1142. science.org/doi/10.1126/science.aeb1142
11. Metallic material breaks 100-year thermal conductivity record. Physics World, 19 Feb 2026. physicsworld.com/a/metallic-material-breaks-100-year-thermal-conductivity-record
12. B. Gwalani et al., Unprecedented Electrical Performance of Friction-Extruded Copper-Graphene Composites. Materials & Design 237 (2024). PNNL release: pnnl.gov/news-media/electrifying-improvement-copper-conductivity
13. A. I. de Isidro-Gómez et al., Intercalated carbon nanotube fibers with high specific electrical conductivity. Science 392, 395 (2026). DOI: 10.1126/science.aeb0673. science.org/doi/10.1126/science.aeb0673
14. IMDEA Materials researchers develop CNT fibres rivalling copper/aluminium cables. IMDEA Materials Institute, 2026. materials.imdea.org/imdea-materials-researchers-develop-cnt-fibres
15. Rice lab makes case for high-performance carbon nanotube fibers for industry. Ken Kennedy Institute, Rice University. kenkennedy.rice.edu/news/current-news/rice-lab-makes-case-high-performance-carbon-nanotube-fibers
16. GMG Unveils Graphene Aluminium-Ion Battery That Fully Charges in 6 Minutes. Graphene Manufacturing Group, Dec 2025. graphenemg.com/gmg-unveils-graphene-aluminium-ion-battery-that-fully-charges-in-6-minutes
17. GMG's graphene aluminium-ion battery: energy density doubles (49 Wh/kg, 6-min charge), April 2026 update. AlCircle. alcircle.com/news/gmgs-graphene-aluminium-ion-battery-118068
18. Avicena Ships 1 Tbps LightBundle Evaluation Kits for AI Infrastructure. HPCwire, 18 Aug 2026. hpcwire.com/off-the-wire/avicena-ships-1-tbps-lightbundle-evaluation-kits
19. Avicena Launches the World's First microLED Optical Interconnect Evaluation Kit (TSMC collaboration noted). Avicena / BusinessWire, 12 Mar 2026. avicena.tech/avicena-launches-the-worlds-first-microled-optical-interconnect-eval-kit
20. Modernizing Telecommunications Networks (Report & Order FCC 26-19). Federal Communications Commission, 26 Mar 2026. fcc.gov/modernizing-telecommunications-networks
21. AT&T's copper retirement plan plows ahead. Light Reading, 2025–2026. lightreading.com/broadband/at-t-s-copper-retirement-plan-plows-ahead
22. ISPs Want to Continue to Discontinue Copper Service (AT&T, Verizon, Fidium filings). Broadband Breakfast, 21 Aug 2026. broadbandbreakfast.com/isps-want-to-continue-to-discontinue-copper-service

Prepared from primary sources as a fact-checked rewrite of a circulating video transcript. Baseline properties used for comparison: copper ≈ 58–60 MS/m and ≈ 400 W/m·K; aluminum ≈ 37 MS/m at ~30% of copper's density. Readiness assessments are the author's and should not be read as vendor commitments.