Lithium has two chokepoints, and they sit in different countries

Most writing on lithium supply risk treats lithium as one commodity with one chokepoint, and the chokepoint named is usually China. China does refine about two thirds of the world’s lithium chemicals (see the lithium supply chain page). But most of what China refines is used in China. Look only at what crosses borders and lithium splits into two products, with a different country at the centre of each.

In 2023 Chile shipped 79 per cent of the world’s traded lithium carbonate. China shipped 74 per cent of the traded lithium hydroxide. Which of those a battery maker is exposed to depends on what it makes.

One element, two chemicals

Cathode makers don’t buy lithium metal. They buy a lithium chemical, and two of them matter.

Lithium carbonate (Li₂CO₃) is the older form. Chile makes it from brine: water pumped from beneath the Atacama salt flat is evaporated in ponds, and the lithium is precipitated as carbonate. It goes into lithium iron phosphate (LFP) cathodes and lower-nickel NMC.

Lithium hydroxide (LiOH, usually sold as the monohydrate) is what high-nickel cathodes use. They have to be fired at lower temperatures to keep nickel out of the lithium layers, and hydroxide, which melts at 462 °C against carbonate’s 723 °C, reacts more completely there. Most of the world’s hydroxide has been made by converting spodumene, a hard-rock mineral mined mainly in Australia, in Chinese refineries.

They start from different raw material and end up with different buyers, so there’s no reason to expect them to share a chokepoint. They don’t.

The numbers

These come from UN Comtrade bilateral trade for 2023, taken from both reporting directions, on the six-digit codes HS 2836.91 (lithium carbonates) and HS 2825.20 (lithium oxide and hydroxide):

	lithium carbonate	lithium hydroxide
traded value	$11.5bn	$9.1bn
traded mass	288,000 t	185,000 t (about 163,000 t LCE)
largest exporter	Chile, 79%	China, 74%
second	Argentina, 8%	Chile, 15%
supplier concentration (HHI)	0.64	0.57
largest buyers	China 57%, Korea 22%, Japan 10%	Korea 70%, Japan 19%

LCE is lithium carbonate equivalent, the usual way to compare lithium chemicals by their lithium content. On that basis, more lithium crossed borders as carbonate than as hydroxide.

Competition authorities treat a Herfindahl-Hirschman index above 0.25 as highly concentrated. Both products are more than double that.

Who depends on whom

Each row gives the share of a buyer’s imports that came from its single largest supplier:

buyer	carbonate	hydroxide
China	88% from Chile	net exporter
South Korea	82% from Chile	78% from China
Japan	60% from Chile	89% from China
EU (12 reporters, extra-EU imports)	86% from Chile	50% from China

Korea is the case to stop on. It relies on Chile for one lithium chemical and on China for the other. A Korean cathode maker that moves from high-nickel to LFP doesn’t shed its exposure. It moves it from China to Chile.

This is the trade-data version of a result Cheng, Fuchs, Karplus and Michalek reported in Nature Communications in 2024: cathode chemistry changes which disruptions a battery supply chain is exposed to.

The EU row cuts against the usual story. For carbonate, 86 per cent of what these twelve countries bought from outside the EU came from Chile, above the 65 per cent single-country benchmark the EU’s Critical Raw Materials Act sets for 2030. For hydroxide, China’s share was 50 per cent. In this data, Europe’s sharpest single-country lithium dependency is on Chile, not China. The volumes are small, though: $244m of carbonate and $251m of hydroxide, a few per cent of world trade, and France is not among the reporting countries.

The four-digit code hides this

Trade statistics are coded at several levels of detail. The four-digit code that many analyses use for lithium carbonate, HS 2836, simply means carbonates. By weight it is mostly soda ash: 23.6 million tonnes traded in 2023, against 288,000 tonnes of lithium carbonate. The biggest shippers by weight are the United States and Türkiye.

On the four-digit code, Chile’s share of trade value is 45 per cent and the HHI is 0.24, just under the “highly concentrated” line. On the six-digit code the figures are 79 per cent and 0.64. The same trade, measured through a coarser code, looks like a moderately competitive market.

My earlier piece on trade data used the four-digit codes. Its lithium figures hold up at six digits: China took 87.6 per cent of its carbonate imports from Chile on the finer code, because nearly everything Chile ships to China under HS 2836 is lithium carbonate. What the coarse code distorted was the world picture, not the bilateral one.

When the network breaks

Concentration tells you who matters but not what happens next. For that I ran a linear-threshold cascade. Remove one country. Any country that loses more than a set share of its trade, the threshold, fails too, which removes its trade from its partners, and so on until nothing else fails.

The table gives the share of countries in each network that fail:

threshold	Chile removed, carbonate	China removed, hydroxide
50% or lower	100%	100%
60%	97%	100%
70%	26%	54%
90%	3%	8%

Both networks collapse almost completely unless countries can absorb losing more than 60 per cent of their trade. Ouyang and colleagues reported this “robust-yet-fragile” pattern for the lithium network in Environmental Science & Technology in 2024: resilient to random failures, fragile to losing the right node. On the four-digit carbonate code, which dilutes lithium with soda ash, removing Chile stops causing a full collapse once the threshold reaches 40 per cent, so the coarse code would also have made the carbonate network look much harder to break than it is.

Nobody knows the true threshold for lithium, so it’s swept rather than fixed. The mfa-engine dashboard runs the same sweep for every material it covers.

A caution about the numbers themselves

Chile and its customers disagree about what the carbonate was worth. They don’t disagree about how much there was: Chile recorded 129,000 tonnes shipped to China, and China recorded 139,000 tonnes received. The prices are another matter. Chile’s records value those shipments at $19,700 a tonne and China’s at $41,700. The same gap shows up for Korea ($28,800 against $48,600) and Japan ($35,300 against $57,500), but not for the United States ($20,000 against $19,600).

Lithium prices collapsed during 2023, so timing and contract pricing could explain part of it. The data alone can’t say which side is closer to the true price. It does mean the dollar shares above carry real uncertainty. Shares by weight tell the same story: Chile shipped 75 per cent of traded carbonate by mass, China 71 per cent of the hydroxide.

What this does not do

• It covers one year, 2023, and not a normal year for lithium prices.
• It measures trade, not production. China’s domestic refining, most of the world’s total, doesn’t appear unless it’s exported. Neither does the step before hydroxide: Australian spodumene concentrate travels under a different code.
• Thirty countries report in this dataset. Countries that don’t report appear only through their trading partners.
• HS 2825.20 includes lithium oxide as well as hydroxide, and the LCE figure assumes it is all hydroxide monohydrate.
• The HHI here uses every country’s export share. Computed from importers’ own records only, as the mfa-engine dashboard does, it comes to 0.73 for carbonate and 0.62 for hydroxide. The conclusion doesn’t change.

The code and cached data are at github.com/ibtisamkhan96/crm-trade-network. The material flow model these trade numbers feed into, including what recycling can and can’t recover, is mfa-engine. Trade data is from UN Comtrade.