Tearing apart the Chinese aluminum power curtailment

Aluminum.

The financial press has spent two weeks running variations of the same headline: Chinese demand is soft, trade war overhang, LME aluminum under pressure. If you've been watching the physical side of this market, you're reading a completely different story. The newsroom is pointing at the price chart. The smelters are pointing at the power grid.

Aluminum smelting is electricity made solid. Every tonne of primary aluminum requires roughly 14,000 kilowatt-hours to produce — about what the average American home consumes in sixteen months. That is not a rounding error. That is the cost structure. Which means when China's grid tightens, aluminum smelters are the first valve to close. They always have been. This summer is shaping up to be the tightest grid since 2021.

Yunnan province runs its smelters on hydropower. That was the whole pitch — cheap, clean, politically defensible electricity piped from the mountains into the pots. It works nine months out of twelve. The three months it doesn't work are May, June, and July, when the rainy season hasn't arrived yet and the reservoirs are at their annual floor. The smelters and the grid operators both know this. It happens every year. And every year, the financial market acts mildly surprised when curtailment notices start appearing.

The mainstream read is: China is slowing down, aluminum demand is soft, the market is responding rationally to macro headwinds. The physical read is: you are about to lose 4-plus million tonnes of annualized smelter output during the window when you can least afford it, and nobody on the demand side has priced that yet.

Numbers. No narrative.

Yunnan's hydropower reservoir levels are publicly reported by China's National Energy Administration. As of mid-May they sit at 34% capacity — below the seasonal average of 41% and tracking the lowest pre-rainy-season level since 2019. The rainy season typically restores power surplus starting late July. That's a ten-week window where the grid is tight and the smelters are first in line to get shut off.

Yunnan hosts approximately 9.3 million tonnes of annual smelting capacity. In the 2021 curtailment cycle, roughly 35–40% of that went offline during the peak constraint period. Even if this cycle is milder — call it 25% curtailment — that's still over 2 million tonnes of annualized output removed from the market for two to three months. On a global production base of around 70 million tonnes per year, that is not a rounding error. It moves the balance sheet.

The flow runs like this:

The 4-to-6 week lag is the critical piece. Curtailment orders typically start appearing in late May. The market won't see it clearly in the inventory data until late June or early July. By the time it's undeniable, the position has already moved against you.

Here's what the market is missing. It's not just the curtailment volume — it's where the curtailment hits in the production chain. Yunnan's smelters are disproportionately weighted toward high-purity aluminum and aluminum alloys for automotive and electronics applications. The commodity-grade material that China exports by the boatload runs from different provinces. When Yunnan cuts, you don't lose generic ingot. You lose the feedstock for the value-added end of the market. That's a different shortage, with a different buyer base, and a different premium structure.

The second fault line: the managed money position. CTAs have been building net short positions in LME aluminum for six weeks, reading the demand-softness narrative and extrapolating it mechanically into the futures strip. Speculative net shorts are sitting at multi-year highs heading into the peak curtailment window. I've watched this setup before. It's a coiled spring on the wrong side of a supply shock.

There's also the grid infrastructure angle nobody in the financial press has touched. Yunnan's hydropower dependency was sold as a feature — clean energy, low cost, sustainable. What it actually created is a smelting base that runs on a seasonal power budget that nobody controls. The province can't build its way out of this with new capacity. You can't add hydropower supply in May. You either have water in the reservoir or you don't.

The market is treating this as a demand story because that's what the price chart looks like from a distance. Up close it's a supply story wearing demand's clothes.

The physical market is already whispering what the financial market will be shouting in six weeks. The Rotterdam physical premium is ticking up. The Shanghai spot premium over futures is widening. Those are not demand signals. Those are logistics signals. Somebody who needs aluminum is starting to pay up to make sure they actually get it.

The sequence is mechanical once the curtailment orders go out. It's run this way in 2021, in 2019, and in 2016. The actors change. The choreography doesn't.

Where does capital go? Not into diversified materials ETFs. Those are diluted with steel, with lithium names that are still working through their own oversupply hangover, with miners that hedged forward during the 2024 price spike and will sit out a spot rally entirely. The edge — if there is one — is narrow.

Pure-play Western aluminum producers outside China's grid constraints are the levered play. Companies like Norsk Hydro or Alcoa that have smelting capacity in Norway and the US Pacific Northwest — hydropower operations, yes, but not on Yunnan's interruptible contract structure. Their cost of production is actually somewhat competitive when LME premiums blow out, because they're the marginal suppliers buyers call when Chinese material gets rationed. I traded a version of this setup in 2021. It worked, then it gave back half on the demand-side correction when China's grid recovered. Nothing is clean here.

The timing risk is real. Curtailment could be milder than 2021 if Beijing decides to prioritize industrial output over grid stability — they've done it before, and they'll do it again when the politics demand it. There's also the demand side: if downstream Chinese aluminum consumption stays genuinely weak, even a supply cut might not move the price cleanly. Both of those are legitimate risks. Neither of them changes the structural setup.

Paper market says the macro is soft. Physical market says the reservoirs are low. The reservoirs are not reading the macro. In my experience, the water in the reservoir always wins the argument. It just takes until June to prove it.