Baseline: The five critical raw material systems

The big picture

Before examining each material individually, it is useful to consider the five materials together. Looking across the system reveals patterns that are not apparent in the individual material profiles and provides context for the detailed, material-level analysis that follows. The following figure provides an overview of the material systems, comparing each material across the major stages of the value chain—from ore refining and processing through manufacturing, in-use stocks, and end-of-life management. It immediately illustrates that, although these materials share a designation as critical raw materials, they differ fundamentally in both scale and system behaviour.

Production volumes span more than two orders of magnitude: at the refinery stage, copper is produced in tens of millions of tonnes annually, while lithium is produced in tens of thousands.⁴¹ The difference is even greater upstream at the ore stage. These disparities matter because they shape the challenges that each material faces and the role that circularity can realistically play. Strategies that are effective for a high-volume material such as copper may be irrelevant for a specialty material like lithium, and vice versa.

These relationships are not static. Copper dominates each stage of the system at the beginning of the projection period, but under the baseline assumptions, its refined and manufactured output declines toward 2050 while silicon continues to increase, meaning that silicon’s refined output could eventually grow to overtake copper. The relative importance of these materials therefore changes over time, reinforcing the need to consider each material on its own terms.

Supply-demand balance and resilience

Four of the five materials experience prolonged periods of supply stress. Copper, silicon, and lithium each spend 14 to 15 years below the threshold required to meet demand, cobalt around 11 years at different points in time, while nickel experiences only two years of supply stress. This points to structural supply concerns for the materials with longer supply gaps, whereas shorter episodes of stress are more likely to reflect localised events. The two are different kinds of problems: short, sharp pulses—geopolitical events, export bans, trade disputes—are fluctuations and liquidity problems, where material still exists globally but becomes temporarily hard to access, driving potential price spikes, logistics disruptions, and speculative buying. This is more typical of materials with patterns of significant oscillations, such as nickel. Longer, structural gaps, such as declining ore grades and geological scarcity, are capacity and planning problems, involving demand shifts and potentially compounding geopolitical concentration risk. This produces a steadier trajectory: a gradual, sustained decline in the service level, mirrored by a gradual, sustained widening of the supply gap, as seen with copper. However, the number of stressed years alone provides only a partial picture of system resilience.

The character of supply stress differs markedly between materials. Copper and lithium experience almost entirely continuous periods below the threshold of necessary supply, signalling sustained supply gaps and a structural imbalance between supply and demand that, once established, does not recover within the modelled period. Cobalt follows the opposite pattern, with shorter but repeated episodes of supply stress, reflecting a supply system that adjusts unevenly rather than gradually. Silicon combines one prolonged period of stress with additional shorter episodes earlier in the projection.

The resilience indicators reveal important differences between the materials. Lithium accumulates by far the largest cumulative resilience triangle area, reflecting shortages that are both prolonged and deep. Copper and silicon experience similar cumulative stress despite differing patterns, while cobalt accumulates markedly less because its shortages are generally shallower. Nickel remains comparatively resilient across all metrics considered. These findings demonstrate that frequency, duration, and depth of supply stress are distinct characteristics, pointing to different vulnerabilities that call for different responses and not a single intervention.

Strategic reserves provide further context for interpreting these results. In the baseline, reserves are never amassed for copper and silicon, available but never utilised for nickel, and actively drawn on for lithium and cobalt. For lithium in particular, reserve releases provide substantial additional supply through the 2030s and 2040s, partially masking the severity of the underlying supply imbalance. A similar, though less pronounced, effect is observed for cobalt. This highlights the important role that strategic reserves can play in managing sustained gaps between supply and demand.

Upstream: Extraction, processing, and manufacturing

Upstream efficiency, and the ore grades that influence it, vary considerably between materials. Cobalt is a special case: because it is recovered almost entirely as a by-product of copper and nickel mining, rather than mined in its own right, its supply is structurally constrained by the ore grades and production decisions of those host metals rather than by demand for cobalt itself. For the other materials, the model reflects the best available evidence. Copper follows a declining trajectory, consistent with well-documented ore grade decline.⁴² Nickel and cobalt remain broadly stable, in line with the historical record, while lithium and silicon are assumed to remain flat because there is limited evidence to support projecting future changes in ore grades.

These differences have important implications for the physical footprint of the system. Copper accounts for the majority of ore moved across all five materials, making the upstream system largely a copper story in terms of volume. Cobalt, by contrast, stands out not because of the amount of material moved, but because its supply depends on copper and nickel production, limiting how independently it can respond to changes in demand.

Upstream loss profiles differ between materials. For copper, nickel, and cobalt, the largest upstream losses occur during manufacturing. Silicon, by contrast, experiences most losses during semi-fabrication, reflecting the purification and wafer-slicing processes that define its value chain. Lithium has historically followed a similar pattern with more losses in semi-fabrication due to dispersive applications, although this shifts over time as battery applications become increasingly dominant.

These contrasting loss profiles demonstrate that opportunities to improve material efficiency are highly material-specific. Rather than applying generic circular manufacturing strategies, interventions should target the stages where losses are greatest within each individual material system.

Midstream: Accumulation and in-use stocks

Each material accumulates in-use stock throughout the projection period, but the pace and concentration of that growth vary considerably. As shown in the figure below, copper, silicon, and nickel continue to build large stocks in long-lived, established applications such as wiring and cable, ferroalloys, and stainless steel. The absolute volumes vary depending on the material: copper accumulates the most at approximately 650 million tonnes, silicon accrues 380 million tonnes, and nickel builds up around 140 million tonnes by 2050. Lithium and cobalt, by contrast, grow much more rapidly from far smaller initial bases, driven almost entirely by battery deployment. Lithium's stock rises by around 2,700%, from approximately 150 thousand tonnes in 2020 to over four million tonnes by 2050, while cobalt's stock grows by around 350%, from approximately 460 thousand tonnes to over two million tonnes.

The results also highlight a structural timing constraint that recurs throughout this analysis. The applications driving stock growth are either those with the longest service lives or the most recently deployed, meaning the material entering the in-use stock today will not become available for recovery until well after the projected supply constraints begin to emerge. This is an inherent feature of stock-based systems rather than a limitation of collection or recycling infrastructure. As a result, the contribution that end-of-life circularity can make within the modelled period is fundamentally constrained, regardless of how well the recovery system performs.

This timing constraint reflects a structural feature of the energy transition itself. Renewable energy systems are fundamentally stock-based: generation, storage, and transmission infrastructure must first be built before delivering energy over decades of service. As a result, the energy transition concentrates material demand in the early years, while its benefits accrue over a much longer period. The speed and depth of the transition therefore determine both how much material is needed and how quickly. A faster transition compresses this demand into a shorter timeframe, increasing near-term pressure on supply even as it brings forward the point at which these growing in-use stocks begin to reach end of life and become available for reuse.

Downstream: End-of-life and circularity performance

Viewed together, the following figures show that the five materials fall into three broad groups in terms of circularity performance.

Nickel and copper form an established group. Both begin the projection with high collection rates—56% for copper and 78% for nickel in 2020—which improve only marginally over time. Their End-of-life Recycling Rates are also high but remain broadly stable, reaching 60% for nickel and 57% for copper by 2050. Secondary supply is driven primarily by end-of-life material rather than new scrap. Together, these patterns suggest systems operating close to their practical ceiling under current conditions, where further gains depend less on collecting more material than on improving what happens after collection.

Cobalt and lithium form a developing group. Both make substantial gains in collection and circularity over the projection period, starting from relatively low levels and reaching stronger performance only later in the timeframe. Cobalt's collection rate rises from approximately 40% in 2020 to 58% by 2050, while its End-of-Life Recycling Rate increases from 20% to 35% and its Recycling Input Rate from around 10% to 34%. Lithium starts from a lower base but grows faster in relative terms: its collection rate rises from 20% to 56%, its End-of-Life Recycling Rate from under 1% to almost 27%, and its Recycling Input Rate from 0% to 22%.

Silicon is an outlier. Unlike the other materials, its collection rate remains largely flat throughout the projection period, increasing only marginally from around 15% in 2020 to 17% in 2050. Its recycling indicators remain extremely low, with an End-of-life Recycling Rate at 1.4% and a Recycling Input Rate of just 0.6% by the end of the period, up from 0.8% and 0.2%, respectively. Downcycled volumes exceed recycled volumes tenfold. In other words, silicon is recovered, but primarily as a lower-value material rather than being returned to high-purity silicon value chains.

This final observation highlights a broader limitation of interpreting recycling indicators. The headline indicators used here capture functional recycling only, excluding both downcycling and remanufacturing. For materials where downcycling is the dominant fate of collected material, these indicators correctly identify weak circularity performance. However, they bring two very different outcomes together: material that is never captured for recycling at all, and material that is captured but recycled into a lower-value use, losing its original function in the process. These represent fundamentally different challenges requiring different interventions. The former is a collection problem. The latter is a problem of maintaining material quality and value after collection—one that can be shaped as much by decisions made before collection as by the recycling process itself.