Tearing apart the Transnet manganese export corridor collapse

Manganese.

The financial press is covering South Africa right now as a currency story. The rand is weak, the macro is messy, load-shedding has a new name, Eskom has a new CEO. None of that is the story. The story is on a 861-kilometre stretch of iron-ore and manganese export rail that is falling apart at a rate that the ore producers have stopped politely mentioning and started quietly screaming about.

The Sishen–Saldanha iron ore line and the manganese corridor running through the Northern Cape to Port Elizabeth — two of the most strategically concentrated export arteries on the continent — are being strangled by locomotive attrition, signal failures, and a state logistics company that has been running on deferred maintenance since 2019. Transnet Freight Rail, which controls both lines, is operating with roughly 40% of its available locomotive fleet. The number has been sliding for three years and it just crossed a threshold.

The consensus framing is that this is a South Africa sovereign risk problem — and therefore someone else's problem. The equity guys model it as a discount on South African producers, clip it into a political risk factor, and move on. That framing misses the actual mechanism entirely. This isn't about South African producer margins. It's about where the world's manganese ore goes and whether the steel and battery industries in China, Japan, and Europe can source a substitute. They cannot.

South Africa produces roughly 38% of global mined manganese output. It holds an estimated 80% of the world's known economic manganese reserves. When the rail stops moving the ore, the ore stops being ore. It becomes rock sitting in a pit 900 kilometres from a port. The market has not priced this correctly. It has priced a logistics inconvenience. What's building is a physical cut-off.

Strip the narrative. Here's the engineering.

The Transnet manganese rail corridor runs from the Northern Cape through to Port Elizabeth — 970 kilometres. It was designed to move around 16 million tonnes per year. Actual throughput has been declining steadily since 2020 as locomotive availability collapsed. The current run rate is estimated at below 10 million tonnes annualised, which means exporters are sitting on ore they cannot ship. They are stockpiling at mine-side. The stockpiles have nowhere to overflow into. The mines are starting to cut production not because demand fell, but because they physically cannot store more unshipped ore.

On the demand side, nothing has softened. Manganese is not a discretionary input. It goes into every tonne of steel — roughly 6 to 9 kilograms per tonne as an alloying agent — and into the cathode chemistries for lithium-manganese-oxide and NMC batteries. The steel cycle is still running. The battery buildout has not paused. The demand machine is running at rate. The supply machine is standing in a railyard waiting for a locomotive that hasn't been serviced in two years.

The vessel queue at Port Elizabeth has been running at 15 to 20 ships waiting at anchor. Each day at anchor costs a producer around $20,000–$35,000 in demurrage. This is not a rounding error — it is quietly destroying the economics of smaller manganese producers whose contracts were priced on the assumption that the rail actually worked. Several have already suspended export commitments to Chinese buyers for Q3. The buyers don't have an alternative source that scales. Gabon produces manganese but at a fraction of South African output. Australia's Groote Eylandt operation is already running at full tilt. There is no surge capacity sitting idle anywhere on earth.

Here's what the steel analysts are not modelling: the Transnet repair timeline.

Transnet has been in a state of managed crisis since a ransomware attack in 2021 compounded years of governance failures and capital underspending. The South African government has announced — and re-announced — a turnaround plan multiple times. Each announcement involves new locomotive procurement from Chinese suppliers, new management appointments, and a target date for restoring rail capacity that quietly gets pushed back once the cameras leave. The current plan calls for 220 new locomotives to be operational by end of 2026. As of now, fewer than 60 are in-country and serviceable. I've seen this movie before with other state infrastructure entities. It does not end with the target being hit.

The deeper structural problem is the maintenance backlog on track infrastructure itself. Locomotive availability is what gets reported. What doesn't get reported is that sections of the Northern Cape manganese corridor have not had scheduled track maintenance in over three years. Speed restrictions have been imposed on three segments due to rail degradation. Even when Transnet adds locomotives, the track condition caps the throughput below design capacity. You can't solve a track problem with locomotives.

Now add the paper market overlay. Manganese ore prices have been drifting lower this year on a China steel demand narrative — the usual suspects citing property sector weakness, construction starts, the standard dashboard reading. The CTAs are net short. The positioning data tells you the market is treating this as a demand problem. It is not a demand problem. It is a supply extraction problem, which is a completely different machine with a completely different repair timeline.

A demand problem resolves when demand recovers. A supply extraction problem resolves when the infrastructure is fixed — which, in South Africa's case, is on a timeline measured in years, not quarters. The market is short a commodity whose physical availability is structurally constrained, while modelling a scenario where the constraint is temporary and cyclical. That's the gap. It isn't subtle.

The sequence isn't complicated. It's just slow — slow enough that the market keeps talking itself out of it quarter after quarter until it isn't slow anymore.

The Chinese mills are the first transmission point. They have been carrying larger port inventory buffers specifically because the South African rail situation has been erratic since 2022. Those buffers are not infinite. Industry estimates put average Chinese port manganese ore inventory at roughly 4–5 weeks of consumption. If South African mine output cuts accelerate through Q3 — which the stockpile math strongly suggests — the buffers run down. And then the mills are in the spot market, competing with battery cathode procurement teams who have their own supply problems and their own deadlines.

The battery angle is the one nobody is modelling yet. Manganese in NMC and LMFP cathode chemistries is not the same product as manganese ore for steelmaking — but the two markets source from overlapping supply chains, and when ore prices spike, the entire complex reprices. The battery buildout has been quietly running down its own manganese inventories on the assumption that South African export volumes would normalise. They are not normalising. They are getting worse.

The timing is the honest uncertainty here. Transnet has been in crisis for three years without a clean resolution trigger. That's exactly the pattern that keeps the paper market complacent — there's always been a next announcement, a next commitment, a next six-month window to assess. But the stockpile math now has a hard floor. Mines cannot keep producing indefinitely into storage capacity they don't have. The production cuts are coming whether the market wants to price them or not. Physical reality has never particularly cared about the consensus timeline.