Editorial Article
Ninety-Six Thousand Six Hundred Tonnes Is a Policy Number
The number that matters is written down, not dug up
Cobalt is usually discussed as a scarcity problem. It is not. It is an administrative one. About 70 per cent of mined cobalt comes from the Democratic Republic of Congo, and the quantity permitted to leave is set by ARECOMS against an export allowance of 96,600 tonnes. In July 2026, roughly 20,000 tonnes — about USD 1.1 billion of material — sat at risk against that cap.
Twenty thousand against ninety-six thousand six hundred is 20.70 per cent of the annual allowance hanging on one administrative decision. That is the shape of the risk, and it is worth being precise about why it is a better risk to hold than most of the others in this chain: a quota has a number and a date on it. A cathode chemistry’s cost curve has neither.
The decade-long engineering programme a quota can undo
The battery industry spent ten years engineering cobalt out of the cell, and it worked. LFP — no cobalt, no nickel — went from a niche chemistry to the volume default. The numbers inside the Pheonix forecast show how completely that succeeded as a share argument even while the cobalt-bearing segment kept growing in absolute terms.
EV battery value runs from USD 91.7 billion in 2025 to USD 369.5 billion in 2033 — a 4.03× market, an 19.03 per cent CAGR. LFP cathode value over the same window runs from USD 8.30 billion to USD 15.99 billion at 8.54 per cent. So cathode’s share of the container it ships in falls from 9.05 per cent to 4.33 per cent.
| Measure | 2025 | 2033 | Change |
|---|---|---|---|
| EV battery market value | USD 91.7bn | USD 369.5bn | 4.03× · 19.03% CAGR |
| LFP cathode value | USD 8.30bn | USD 15.99bn | 1.93× · 8.54% CAGR |
| LFP cathode share of EV battery value | 9.05% | 4.33% | −4.72 points |
That is a market doubling inside a container quadrupling. And the consequence for a cobalt quota is the opposite of the comforting reading: the cobalt-exposed segment is becoming smaller, denser and more valuable per unit. A quota does not need to hit the volume market to matter. It has to hit the part of the market where energy density is the product, and that is precisely the part the industry could not engineer cobalt out of.
Four chokepoints, four different instruments
Cobalt is one line in a table that is often read as a single risk. It is not one risk; it is four instruments, and they fail in different ways.
| Chokepoint | Concentration | Instrument |
|---|---|---|
| Cobalt — DRC | ~70% of mined supply | Export quota (ARECOMS, 96,600 t) |
| Spherical and synthetic anode — China | >90%, under 1% made outside | Industrial policy and graphitisation capacity |
| Lithium refining — China | 60–70% | Processing concentration |
| Nickel — Indonesia | 31.5% rising to ~60% | Resource nationalism and ore-export rules |
The instruments matter more than the percentages. A quota can be extended, tightened or lifted by decree and its effect is immediate and price-visible. Industrial policy is slow, capital-intensive and largely irreversible on a five-year view — a graphitisation plant that exists cannot be willed away, and one that does not exist cannot be qualified in a quarter. Processing concentration is a margin question before it is an availability question. Resource nationalism is a contract question.
A supplier who plans against the cobalt cap and not against the anode concentration has planned for the interruption that announces itself and ignored the one that does not.
Compliance arrives after the forecast has stopped looking
There is a timing problem in how this chain is modelled. The forecast’s acceleration to a peak annual growth of 23.67 per cent in 2029 is grounded explicitly in announced capacity from Samsung SDI, VW, LG-Honda, Umicore and Albemarle, starting between 2024 and 2027. It is, in other words, a factory list converted into a growth rate — the same conflation that produces the wilder gigawatt-hour numbers elsewhere in the storage complex.
Against that, the compliance calendar runs later. The EU battery passport lands in February 2027, the year the acceleration begins. Recycled-content targets land in 2031 — two years after the model’s peak year and after the window has effectively stopped looking. So the two things that could most change the composition of demand are both outside the part of the curve where the forecast is confident.
This is not an argument that the forecast is wrong. It is an argument about what it can be used for. A growth rate built from announced factories is a statement about capacity intentions; a battery passport and a recycled-content mandate are statements about which material is allowed in which cell. The second can invalidate a plant that the first counts.
What would change this reading
One: the cap being lifted or extended without interruption. If ARECOMS settles the allowance upward and the 20,000 tonnes clear, the July 2026 event becomes a price spike rather than a structural constraint — and the argument that quotas are the forecastable risk weakens with it.
Two: the processing route moving rather than the mine. Congolese material diverted to Indonesia or China for intermediate processing changes the exposure from an export allowance to a logistics and ownership question. That would not remove the chokepoint; it would relocate it to a place with more participants and less transparency.
Three: recycled content becoming a supply source rather than a compliance cost. The 2031 targets are the only mechanism on this list that changes the concentration itself rather than the price of living with it. If secondary cobalt scales ahead of schedule, the quota loses its grip for a reason no quota negotiation can undo.
The wider frame — why a market that quadruples while its unit price falls implies roughly 5.59 times the volume, and why the cathode inside it grows at less than half the container’s rate — is set out in Four Times the Market, Half the Material Share. The point of both pieces is the same: in this chain the volume forecast is the least interesting number, and the instrument behind the material is the one that decides what gets built.
Frequently Asked Questions
How concentrated is cobalt supply really?
About 70 per cent of mined cobalt comes from the Democratic Republic of Congo, and the export of it is administered by ARECOMS against a 96,600-tonne cap. In July 2026 roughly 20,000 tonnes — about USD 1.1 billion by value — sat at risk against that cap, which is 20.70 per cent of the annual allowance. A quota is a different kind of risk from a geological one because it has a number and a date.
Why does a cobalt quota matter if the industry is moving to LFP?
Because the two markets are growing at different speeds. EV battery value goes from USD 91.7 billion to USD 369.5 billion at 19.03 per cent, while LFP cathode value goes from USD 8.30 billion to USD 15.99 billion at 8.54 per cent. Cathode's share of the container falls from 9.05 to 4.33 per cent — the market is growing toward the chemistry that needs no cobalt, which leaves the cobalt-exposed segment smaller, denser and higher-value, and that is exactly where a quota bites hardest.
Is the cobalt cap the only chokepoint in the battery supply chain?
No, and the others are governed by different instruments. China holds above 90 per cent of spherical and synthetic anode production with under 1 per cent made outside it, and 60–70 per cent of lithium refining, while Indonesia has moved from 31.5 per cent to roughly 60 per cent of nickel. Those are industrial policy, processing concentration and resource nationalism respectively. A supplier planning against one instrument and not the others is planning for one kind of interruption.
When do the compliance rules hit relative to the forecast's growth peak?
After it. The model's peak annual growth of 23.67 per cent arrives in 2029 and is grounded in announced capacity starting between 2024 and 2027. The EU battery passport lands in February 2027 and recycled-content targets in 2031 — the second of those after the forecast window stops looking. Compliance risk is therefore not in the forecast at all, which is the reason to hold it separately rather than read it as a cost line.