Lithium review - 4Q25

Sentiment changes from glut to shortage, does it have any battery left?

An archaic version of the Chinese character for salt 盐 was composed of the ideograms for container, brine, worker, and government official. Salts and brines were a big thing in antiquity, at the core of debates on the role of the central government in economics and production.

Last February, I published a lengthy Lithium Primer, covering the basics of the industry: an explanation of the supply chain and different routes to battery lithium, the economics of the different links in the chain, technical and capital factors affecting the cycle, and the most relevant geographies and companies making up the market.

Today, it seems a lithium review is in order. The reason is simply that the sentiment has quickly shifted, from glut to shortage in a matter of months, between July and September, continuing today, and impacting prices, up 20/30% YTD, and 50/60% since the recent lows in May.

This raises the question: Is a new lithium upcycle coming? Will it squeeze like in 2021?

This article analyzes these questions within a cyclical framework, explaining where the change in sentiment is coming from, what expectations are embedded in certain demand projections, and what we can know about supply with greater certainty, thanks to production and capital deployment restrictions. In addition, it helps build a more complete understanding of the lithium market.

In a following article, we will analyze the impact this has on the cost structure and profitability of different companies in the chain.

Without further ado, we gonna rock down to Electric Avenue!

Note: Unless clarified separetely, all references to tons are lithium carbonate equivalent (LCE).

Disclaimer: The opinions expressed in the Blog are for general informational purposes only and are not intended to provide specific advice or recommendations for any individual or on any specific security or investment product.

Key takeaways

• The narrative in the lithium markets up to June 2025 was a significant excess of supply at least into the late 2020s. This has changed to a potential balance in 2026 and even shortages in 2027.

• The change in narrative came mainly from the demand side, with more optimistic projections for EVs and ESS. Demand is the volatile element of the cycle, whereas supply is reasonably well determined until at least 2027.

• Given (1) the significant spare capacity available, particularly in mining and processing; and (2) the large increases expected for 2026, especially from China, Africa and Argentina, demand should be really bullish to get the market to balance before late 2027. This seems unlikely. Under even moderately conservative demand assumptions, the oversupply remains until 2028, when larger capacity additions are expected.

• Although lithium prices are tilting more bullish than earlier in 2025, they do not yet reflect a full acceptance of the shortage narrative. This is important because lithium market participants have skin in the game. If they fully embraced the shortage narrative, prices should squeeze. I think this skepticism will resolve in early 2026, based on EV sales and ESS deployment. For the bullish view to prevail, sales need to surprise systematically and largely exceed, even double, previous expectations. Under the remaining scenarios, even with good but not explosive demand, lithium will remain oversupplied, and prices would return to more depressed levels.

足久必缺，缺久必足 - Surplus begets shortage, shortage begets surplus

Lithium prices and markets started the year in a relatively gloomy state, with lithium chemicals (lithium carbonate and lithium hydroxide) trading at around ~RMB 75 thousand per ton ($10/11 thousand).

The narrative was that the market was oversupplied by at least 15%, with the excess production going into inventories. In addition, capacity estimates exceeded demand by 40% in mining and 70% in conversion. Adding insult to injury, EV sales were not great in Europe, the US was exiting the energy transition and reducing subsidies on EVs, and China’s economy was not strong enough to warrant excessive bullishness.

The global panic post-tariff announcements in April sent prices in China to RMB 60 thousand per ton (~$8 thousand) by June. This has marked the bottom, and prices have recovered to ~RMB 85 thousand as we speak. The core lithium names (SQM, Albemarle, Tianqi, Ganfeng) are all up 100% in the past 6 months, and many more marginal names are up even more.

Most importantly, and the potential reason behind all this volatility, is an underlying change in sentiment, from an oversupplied market to a quickly undersupplied and potentially short market. Whereas the read just a few months ago was that the lithium market would remain oversupplied at least into the late 2020s, today the read is that next year it could already get tight. Further, the read is that capacity will not be easy to bring to market, leading to large lithium consumers and speculators to panic for supply and further increase pressure on prices.

The role of Chinese demand and supply

Lithium cannot be studied without China: it is becoming the largest miner, it is the largest processor by far, the largest battery manufacturer, the largest end-market for EVs and grid-batteries, and the place where the most liquid exchanges are located.

In this case, China is also the driver of the change in narrative, both on the supply and demand side.

On the supply side, we have the anti-involution (反内卷) and rational/scientific allocation (优化配置) campaigns sweeping across oversupplied industries. These are efforts by the government to curb excessive capacity, mainly via restricting new permits for capacity, and by increasing economic and non-economic thresholds (ex: environmental) so that older/marginal capacity leaves the market. In the case of China, the policies included shifting permission for lithium mining to the national level, counteracting the local incentive to push capacity, with several lepidolite mine shutdowns or strict reviews in the city of Yichun (宜春, Pleasant Spring), province of Jiangxi, the mining core of Chinese lithium ore.

This anti-involution theme is somewhat counteracted by another Chinese campaign, in this case for self-sufficiency, supply chain resilience, and import substitution (自立自强, 产业链韧性, 进口替代). A key comment I found when reviewing papers (example, example) about Chinese lithium production is the concern about the weight of imports on raw materials. This concern may lead the country to expand, not constrain, lithium resource production.

On the demand side, EVs have sold better than originally expected in both China and Europe, both up about 30% YoY, with Chinese EVs already representing ~50% of all sold vehicles. Additionally, China is ramping up Energy Storage Systems (ESS, or grid-level batteries) at an impressive 60/70% rate YoY. ESS already represents close to 20% of global lithium demand, and is becoming a core component of the renewables technology stack. This has increased demand for batteries and lithium for 2025 versus previous estimates.

Finally, the market was sprinkled with recent comments from Ganfeng (赣锋锂业, Forefront of Jiangxi Lithium) and Tianqi (天齐锂业, Heavenly Allignment Lithium), the largest players in global mining/processing, both indicating that the markets will be undersupplied in 2027 and even potentially as early as 2026.

Demand growth scenarios

The goal of this section is to analyze the requirements for end-demand growth (EVs and ESS) to reach certain lithium demand scenarios.

In the original Lithium Primer, I proposed that the determinant of the lithium cycle was supply, not demand, because the former was inflexible in the short term, while the latter was inelastic. Prices would not change demand except marginally in the medium term, and therefore, it was supply that determined prices. Consequently, this article focuses more on the lithium supply.

However, analyzing demand sentiment is important because demand is more volatile (for its own reasons unrelated to lithium pricing), and given supply’s inflexibility in the short term, it can move the market from very oversupplied to less oversupplied, in balance, tight, etc.

Original and revised expectations for 2025

In 2024, lithium demand expanded 20%, and the original expectation for 2025 until about September was 22%, with further deceleration to 16% by 2026 (as reported by CoChilCo or Nomura). This represents a very oversupplied market, as aggregate demand for 2025 was originally expected at ~1.35Mt (figures are from CoChilCo; Nomura figures are closer to 1.15Mt), compared with processing capacities at 2Mt and resources reaching those figures next year.

However, in recent months, the expectations have been revised upwards for 2025, and consequently, as well for 2026. Albermarle 3Q25 call talked of lithium demand expanding 30%. Then we had the comments from GanFeng’s CEO and Tianqi’s Vice Chairman about demand increasing 30% in both 2025 and 2026.

These revised figures would put demand at ~1.45Mt in 2025, and 1.85Mt in 2026, just shy of demand/capacity balance, compared to 1.6Mt for 2026 in the original projection. Add a little fear around inventories and availability into 2027, and this becomes the basis for higher prices.

Where is this bullishness coming from? There are only two sources of growth in lithium: EVs and ESS. The remaining demand (small batteries and industrial uses) represents a stagnant 170Mt, and small vehicle batteries (two wheelers) 70Mt, growing similarly to EVs but from a small base. Among the current bullish commentary, there’s a lot of talk about novel applications like robotics and eVTOLs, but I don't think this can influence the next 2/3 years meaningfully.

On the original numbers from CoChilCo for 2024, EVs consumed 720Mt, and ESS 175Mt. With the recent acceleration to 30% growth in EVs and 60% growth in global ESS, we get to the 1.45Mt for 2025 (1Bt for EVs + two-wheelers, 280Mt for ESS). The question is, can that continue into 2026 and 2027?

EV projections

Starting with EVs, the two relevant markets are China and the EU. 70% of EVs are sold in China, which already has an EV penetration in new cars sold of 50%. China has already announced it is removing subsidies and removing the industry from its strategic list, as it is considered mature. Assuming EVs grow at 30% again and the aggregate growth figure for cars in China is 8%, then EVs will reach 70% of all cars sold in 2027 and ~87% in 2028. Europe, at a penetration of 25%, and with renewed carbon reduction commitments, could sport a faster growth figure for longer, but it is only 20% of the market. I think sustaining 30% in 2026 and even more so forward is not very conservative. The figures are probably closer to 15/20% next year and lower forward.

ESS projections

The largest driver of demand forward is expected to be ESS, already at 20% of demand this year. ESS helps solve the intermittency (solar working during the day but not during night), curtailment (more power produced during the day than the grid can handle leading to some capacity being disconnected), and duck curve (a huge rampup in non-renewable demand when the night comes and solar decreases at the same time as demand increases) problems with renewables, specially solar. The solar + ESS technology pack could really change the economics of renewable energy forward, reducing the burden on grid infrastructure and improving revenue for solar projects, allowing them to sell during the night. Demand is potentially massive. Even today, with 2TW of solar capacity globally, and assuming a 3-hour ESS factor (ESS is mostly needed for 2/4 hours a day between dusk and people going to bed), you get 6TWh of capacity, which is absolutely massive compared to the current 250GWh additions in 2025. BloombergNEF expects capacity to grow at a CAGR of 20%+ for one decade (to reach 7TWh in 2035).

But still, can ESS push aggregate demand to 30% next year and forward? 30% growth in 2026 implies almost 450Mt of net new demand. If EV (plus two-wheeler) growth is 20% next year, it will add ~200Mt (from a base in 2025 after growing 30% of ~1Bt), implying battery demand has to almost double in 2026 from 280Mt to 530Mt to reach 30% growth on aggregate in 2026.

Adding all up

If we start from 1.45Mt and add 30% on EVs (300Mt) and 60% in ESS (170Mt) we get to 1.9Mt in 2026, just shy of balance. Add a little more and 2027 is shortage.

Let’s say instead of that we get 15% in EVs (150Mt) and 60% in ESS (170Mt), then we are talking of approximately ~320Mt in new demand, and getting the market to ~1.77Bt in 2026. If we repeat the growth in 2027 (much less conservative), we get +170Mt from EVs and +270Mt from ESS, totalling +400Mt and a total of ~2.17Bt, reaching balance in 2027, not 2026.

Again, the idea here is not to predict, but rather to work with scenarios to see how optimistic one needs to be for the supply gap to shrink or even flip. Balance in 2026, in my opinion, requires a very bullish view, and balance in 2027 requires a fairly bullish view, about end demand in EVs and ESS.

A final point regards elasticities in the mid-term. I think elasticities in EVs are probably very low, given that battery costs are 10/15% of an EV’s cost, and that lithium itself is 10% of the battery or less (Fastmarkets). That means, even if the lithium price tripled from here, it would only increase the cost of an EV by less than 5%. Meaningful, but not model-breaking. For ESS, I was not able to find good estimates or the data needed to recreate them, but one could intuitively think that the elasticity is a little higher because the batteries play a larger role in an ESS system. If that were the case, then the expectations for ESS deployment could be more price-dependent.

Reviewing supply

Lithium supply is much more inflexible and less volatile than demand in the short term. The key point here is short term, however, because different resources can increasingly change their capacity up or down as timeframes lengthen. Further, contrary to demand, willingness to invest in supply expansion is very price elastic, especially for some resources. This makes the analysis a little more context-dependent. For more details, please refer to the Lithium Primer.

In this section, we will provide a review of the expectations around production from different resources in 2026 and 2027, and how the projections might evolve into 2027, 2028, (2026 being almost defined, given natural lags in capacity), and beyond, to accommodate higher or lower demand.

As an aggregate remainder, the stack of the cost curve in resources goes from brines in the Lithium Triangle (and now China), to Australian spodumene, to African spodumene, and finally, Chinese lepidolite. Separately, we need to analyze ore processing capacity, mainly in China. Brines and processing of ore are the most inflexible portions of the supply chain, and where we might see tightness in the next few years. Let’s go in order.

Brines

Brines today make up 30/35% of the supply of lithium, and I believe they are a long-term anchor of the cycle. One reason for that is that brines sit at the bottom of the cost curve, potentially as much as 40% below spodumene-derived carbonate. Therefore, they are consistently adding capacity even when the remaining portion of the curve is underwater. We will see below that investment in brines has remained strong despite the current downturn in lithium. The second reason is that brines are also one of the slow/inelastic gears of the lithium machine, because production is planned with 18 months in advance, approximately. On a longer time basis, developing a completely new brine operation can take from 5 to 10 years.

Chile

The current brine leader is Chile. Chile produces 300kt/year, basically divided between SQM (240kt) and Albemarle (60/80kt). This is a very cheap supply, profitable even at $8k/t, but both companies are not expecting large increases in capacity into next year, “only” 10/15%.

SQM’s capital expenditures have remained consistent at about $1 billion per year without decreasing despite the cycle downturn, and are expected to continue that way into 2027 at least. The company expects to increase production to reach 260kt of carbonate and 100kt of hydroxide (~150kt of LCE), although hydroxide capacity is not additive to carbonate, only a different route. This implies certain stagnation (still 10% growth) into 1H27 at least, but the possibility of a pickup into late 2027 still exists, if the decisions are taken in early 2026.

In the long term, Chile is attracting other players like Rio Tinto, entering two projects this year in Chile (Maricunga, Altoandinos), which will start to impact production in the 2030s. This shows how Chilean production can continue expanding at the bottom of the curve.

Argentina

Argentina is also extremely competitive and, therefore, expanding production rapidly. Production in 2025 is expected to reach 130kt, with capacity just shy of 200kt. The projection again is for Argentina’s brines to “stagnate” (+10%) into 2027, but I believe this is flexible by late 2027 as well.

Again, this production is expanding because it is profitable, and most large players (including Rio Tinto and Ganfeng again) are investing heavily. These new projects should start hitting production in the late 2020s.

China

Finally, we have China, the big monster, and I think the big disruptor of future supply in brines. Two different deposits are rapidly expanding in China’s Qinghai (青海, Turquoise Sea) and Tibet (西藏, Western Tibet) provinces. Production is expected to reach 230kt LCE in 2026, or basically the size of Chile, growing at 40%+, while still attracting investment. Furthermore, Chinese production is always a wildcard because they can get things done cheaper and faster than anyone else. Again, with the right price, late 2027, early 2028 could very well post higher growth rates than expected.

The developments in Argentina and China are also important testing grounds for DLE technology, which could completely change the lithium cycle by adding flexible capacity from brines (albeit at a higher cash cost). Today, this remains marginal, but it should be evaluated.

Brines on aggregate

Add up all of this, and we can easily see brines in the 700/800kt region next year. The projections slow down into 2027, say 15%+ aggregate between Argentina/Chile at 10% and China at 20%, implying 2027 at about 800/900kt. Based on today’s projections, 2028/29 has a new upcycle in these resources, but I believe that with good pricing, many projects could expand into 2H27.

An important point regarding this flexibility is that, at current prices, brine producers are very profitable. They have all the incentives to expand, as long as they expect prices to remain somewhat elevated into the near future.

Ores and conversion

The remaining market for both carbonate and for the majority of hydroxide comes from ores: spodumene, lepidolite, mainly, plus other more marginal rocks.

Whereas in brines we speak of a single process, because capacity is generally built for extraction/evaporation + conversion into lithium carbonate, for rocks we need to talk about two, generally separate, processes: extraction + concentration, and then conversion. When we talk of spodumene capacity, for example, it refers to the capacity to extract the rock and concentrate it to the commodity level of 5.5%. Then we talk separately of carbonate/hydroxide conversion capacity, which in some cases can be part of a single flex-production plant.

A key difference with brines, also, is short-term marginal flexibility. Mines, concentrators, and converters, do not need to run at full capacity, whereas brine usually has to run at whatever capacity was determined in the past. This has generated a situation where there is significant spare capacity in markets today, particularly in processing, but also in mining.

Over the long term, ore concentration capacity requires 1.5/2 years to expand, and the same or even longer can be said for processing. The current spare capacity, the marginal position in the cost curve, and the length required for large capacity expansions make the mining and processing sectors the drivers of cycle volatility.

Australia

The cost leader in ores is Australia, with almost 500k LCE, mostly in the form of unprocessed spodumene. It is such a cost leader that during the Q2 ‘winter’, most miners remained above the water.

It is true, however, that miners were not investing as much, hurt by their investments in processing, and therefore, capacity is expected to expand by 10% only in 2026 and beyond. This would imply 550kt in 2026 and 600kt in 2027. There is some Australian spare capacity, like Ngungaju (25kt), Bald Hill (20kt), and parts of Mt Cattlin. Fastmarkets reported that ore mines can be returned to production within weeks. This would imply production can get to 600kt and 650kt if prices are supportive, only by restarting capacity.

On a longer time frame, capacity requires at least 1.5 years, potentially more. For example, Pilbara’s Pilgangoora P1000 took 13 months from construction start to first ore, at a cost of AUD 560 million ($370 million) for 1 million tons of concentrate (133kt LCE). Greenbushes’ CPG3 started in late 2023 and will be finished in early 2026, therefore spanning more time (~2.5 years), at a cost of AUD 880 million ($580 million) for 500k tons of concentrate (66kt LCE). This would imply Australian rock capacity would remain relatively inflexible until 2028, beyond the 650kt.

Africa

Higher in the cost curve, we have Africa (mainly Zimbabwe, and now Mali), with projections for up to 280kt LCE in 2026, and potentially 330kt in 2027. Ganfeng’s mine in Mali continues to ramp up, and the continent is expected to reach 400kt in 2030. Finding data on spare capacity or ramp up is more challenging, and whether the continent is faster/slower than Australia is hard to determine, because infrastructure and labor are much worse, but the projects are also led by Chinese companies, which are faster.

China

Finally for ores, we have China, with lepidolite sitting fairly high in the cost curve. Again, I believe the clash between anti-involution and self-reliance will favor the latter over the former, particularly in ore, where the country is still very reliant on imports. For example, for all the talk about Yichun, production is expected to grow 20% next year, to almost 200kt. From the same Chinese source, other regions in Xinjiang (新疆, New Frontier), Inner Mongolia (内蒙古), and Hunan (湖南, South of the Lake) could provide another 140kt, growing by 60%. These estimates were made with lower prices, again allowing for mothballed capacity to join. I think 400kt from China’s ores in 2027 at current prices is not crazy, but this is arguably the most wiggly of all factors, because it is marginal. And again, these are the growth levels during the winter (capacity that was planned in 2023/25 when prices were already low). How much investment could flow to Chinese mining under true fear of scarcity?

Aggregate ores

Adding up, the total ore production could therefore reach 1.2Mt in 2026 and 1.35Mt in 2027 under good price conditions that make marginal spare capacity return to production. Going beyond that before 2027 will be challenging, but as always, the speed of the Chinese is a wildcard factor.

Conversion capacity

Then, moving to conversion, the only relevant region is China. According to Nomura, global processing capacity will stand at 2.2Mt in 2026, but this includes lithium carbonate coming from brines, which we already considered above. Removing those 700/800kt we are left with 1.4/1.5Mt of ore (ex-brine) processing capacity. We are talking here about capacity, not supply, which was lower in 2025 (close to 1Mt excluding brines), because of low utilization in unprofitable plants. Here is where we could see the strongest moves for anti-involution closing plants, although this is unlikely if shortages are expected and prices are rising.

Processing in China is where tightness could become more problematic eventually, because new plants will not come into planning until spare capacity is more utilized, and from then on we should still expect 2 years for construction. Still, based on 1.4/1.5Mt of capacity today, ore would be a bigger problem in the mid-term (2027/28).

Adding up the supply

Supply has a hard ceiling until 2027, with ore mining spare capacity offering some wiggle room, along with ample wiggle room from spare processing capacity.

For brines, we arrive at 700/800kt from brines next year, climbing to 800/900kt in 2027. Above that, it gets complicated before 2H27 at least, potentially early 2028, and that depends on certain bullishness leading brine operators to increase plans in early 2026. I think the big flexible factor in brines is China, because of the bullishness emanating from the country, the concerns with self-sufficiency, the convenient position on the cost curve, and the fact that they are fast at building stuff.

For ores, we get to 1.2Mt in 2026 and 1.35Mt in 2027, assuming that prices are good enough (RMB 75 thousand+) to get mothballed capacity into production. Before 2028 and even into 2029, those numbers become less flexible, in particular because China has high costs in ore, and therefore it will probably not be fast in investing in new capacity. Still, the higher the prices and the bigger the shortage concern, the faster Chinese mines will ramp up.

Finally, we have ore processing at 1.5/1.6Mt, above ore capacity for those same periods, and therefore irrelevant as a limit before 2028. Processing suffers the same problem as ore in China: spare capacity is large, and the margin position is bad, therefore requiring a lengthy period of good prices to induce new capacity, and then additional time to make it a reality.

Overall, we come to 1.9/2Mt in 2026 and 2.15Mt/2.25Mt in 2027. These are the hard limits of the industry in the mid-term.

Final scenarios

Adding all up gives us an idea of where we stand today, summarized in the table below.

That is, current prices below RMB 90k seem, in my opinion, indicative of a market that does not fully buy the mega bullish views of Ganfeng or Tianqi, simply because otherwise there would already be panic, a scramble for inventories, and a squeeze.

Whether we remain in a bullish territory supportive of RMB 80k and above prices (with the potential squeeze) like today or if we return to depressed territory, will greatly depend on how demand for EVs and ESS works next year.

If 2026 starts hot, things can get squeezy very fast, I would believe, because the very bullish scenario already opens the potential for a shortage in 2026, and that fear can cause the squeeze.

If that doesn’t happen, I think the shortage window moves into 27/2028, where we already enter new brine addition territory, both already announced and potential. I think in order to see meaningful capacity additions before 2027, coming mainly from extending brine capacity, we would need high prices in 1H26, enough to convince producers that the shortage is almost certain. That means if 2026 starts without supportive prices, the risk remains for a run because of fears in 2027. This risk would be enough to keep prices where they are today. I think this is the thesis of the aggregate view of the market today. We should remember prices are up a lot on a 6-month basis, but on a yearly basis, they are barely above the cost curve ceiling.

Finally, if 2026 is disappointing, then we would return to depressed prices, barring capacity curtailments, because a 2027 shortage would lose steam, and 28/29 already has new brine capacity entering the market.

Events to follow

To keep an eye on the fundamentals of the market, I would concentrate on three core events:

• Sales of EVs and ESS, both projections and effective

• Announcements of revisions of estimated projections from mines and brines. Announcements of investment in new capacity.

• Announcement of the restart of mothballed capacity in ores and higher utilization in processing