As 2025 drew to a close, a growing sense of dread began to rise amid increased reports of an apparent shortage of memory hardware. But days into 2026, a new supply chain gap is set to emerge: copper.

A study by S&P Global warns of a “substantial shortfall” in copper supply to such a degree that the problem poses a “systemic risk for global industries, technological advancement, and economic growth.”

A projected supply deficit could reach 10 million metric tons by 2040, with demand set to surge by 50 percent compared to current levels.

“Copper is the great enabler of electrification, but the accelerating pace of electrification is an increasing challenge for copper,” said Daniel Yergin, vice chair of S&P Global, who co-chaired the study. “Economic demand, grid expansion, renewable generation, AI computation, digital industries, electric vehicles, and defense are scaling all at once, and supply is not on track to keep pace.”

A shortfall, super-sized by servers

The study suggests that a shortfall in the metal is being widened further by growth in the digital infrastructure space, suggesting that AI and data center demand, alongside defense spending, could represent a combined four million metric tons of additional demand.

Beyond being employed in power delivery and cooling systems, copper is an integral part of the data center networking component, being found in everything from Ethernet cables to transceiver modules.

Copper is used alongside tungsten in chip-level circuitry above the 10-nanometer level, with the metal offering low resistance to electromigration compared to aluminum. And for interconnects, copper-based cables, including DACs and ACCs are increasingly being used to support short-reach, high-bandwidth connections between servers and switches within a single rack.

S&P’s report did state, however, that the shift away from copper to fiber optics in the interconnect space could result in a decline of overall copper intensity in data centers by “four to five metric tons per megawatt, which is of limited impact compared to the overall copper intensity of 30-40 metric tons per megawatt for non-crypto data centers.”

Beyond fiber, there’s a concerted effort into developing photonics technologies as an alternative to both copper and fiber for interconnects, though Nvidia CEO Jensen Huang said last August that copper would dominate for some time, adding: “We should stay with copper for as long as we can, and then after that, if we must, we’ll use silicon photonics.”

“Copper demand for data centers is forecast to increase from 1.1 million metric tons in 2025 to 2.5 million metric tons by 2040,” the S&P report reads. “AI training data center-related copper demand will account for 58 percent of total copper demand in data centers by 2030.

“Annual copper demand [as pertained to data centers] in 2040 could be anywhere between 1.7 and 2.7 million metric tons – a span that underscores both the uncertainty and the scale of the challenge ahead.”

Aurian De La Noue, executive director for critical minerals and energy transition consulting at S&P Global Energy, said: “Multilateral cooperation and regional diversification will be crucial to ensure a more resilient global copper system – one commensurate with copper's role as the linchpin of electrification, digitalization, and security in the age of AI.”

Copper adds to memory and laser woes

Copper joins a growing number of components and materials that could well place significant strain on digital infrastructure buildouts.

A shortage of memory chips has already caused panic buying amid price hikes. The drought stems from OpenAI signing massive deals with manufacturers Samsung and SK Hynix to secure up to 900,000 DRAM wafers per month, the equivalent of 40 percent of the total global DRAM output for its Stargate project.

The entire hyperscale data center buildout is also to blame, with AI workloads requiring vast amounts of memory, which has already left the consumer PC market to the wayside to such an extent that long-time manufacturer Micron opted to get out and focus on enterprise.

As a result, Hewlett Packard (HP) warned late last year that supply chain issues are a genuine roadblock set to impact its 2026 profitability. Dell and Lenovo have warned of cost increases, and SK Hynix has drawn up plans for low-power wafer factories to meet demand.

Meanwhile, Nvidia’s apparently aggressive efforts at shoring up its capacity of electro-absorption modulated lasers (EML) – used to send support transmissions of large amounts of data over long distances – could lead to a potential worldwide shortage.

A TrendForce report published in December claimed that Nvidia’s EML purchases, which are likely to be used in the production of co-packaged optic (CPO) iterations of its Spectrum-X Ethernet and Quantum-X InfiniBand networking fabric solutions, may force wait times for related components beyond 2027.