One document is an accident; six are a pattern. The solar industry just produced a document, and the question is which category it belongs in. On August 6, 2026, eight leading polysilicon producers signed an “anti-involution” pact, pledging to price no lower than full cost and to hold back supply while the market recalibrates. Sina Finance and the photovoltaic industry portal both carried the announcement on August 14.
The receipts tell the story the press release won’t. The press release said the industry was uniting against irrational pricing. The receipts say prices actually moved.
The First Receipt: The Signing
Let me lay the primary facts down in order, because this column does not work from hearsay. On August 6, 2026, eight polysilicon leaders signed the anti-involution initiative. The core pledge: sell no lower than fully loaded cost, and halt sales while positions firm up. The signing was reported by Sina Finance on August 14, cross-checked with the industry portal’s same-day coverage — two independent lines, one consistent account.
That is the first receipt. An agreement among competitors on pricing floors is the kind of thing that usually stays whispered in hotel rooms; here it was signed, dated, and announced. Whatever else it is, it is documentation.
The Second Receipt: The Price Move
The second receipt is the market’s response, and it arrived fast. By August 14 — eight days after the signing — module quotes had risen by 1 to 4 cents per watt across the board. The wafer price for 183N silicon moved from 0.8 yuan per piece to 1.12 yuan per piece, an increase of about 40% in a single week. The figures come from TrendForce’s EnergyTrend pricing service, dated August 13, and the industry portal’s August 14 report.
Now I want to pause here, because this is the part that separates evidence from anecdote. A price hike that follows a signed pledge within eight days is not a coincidence; it follows the chain from announcement to quotation. The direction is documented, the magnitude is documented, and the timing is documented. The trail is clean.
But follow the trail far enough and you reach a fork. A 40% one-week wafer move is the kind of price action that can be a genuine correction or a coordinated spike. The receipts so far only prove the second — that prices moved. They do not yet prove the third — that the move will hold.
What the Receipts Do Not Yet Show
Here is where I hold back, and I want my readers to see the discipline. The anti-involution pact has produced two receipts: the signing and the price increase. What it has not yet produced is the third and decisive one: sustained order flow at the new prices.
To be honest, I have seen this pattern before in other industries — a floor is declared, prices firm up for a fortnight, and then demand at the higher level turns out thinner than hoped, and the floor quietly dissolves. The pact is only as durable as the next two quarters of shipment data. I would be overclaiming to call this a resolution of the overcapacity problem.
Let me correct one framing I nearly used: I almost said the industry has “ended” the price war. That is not supported by the evidence. What the evidence supports is narrower: a group of producers has announced a pricing floor and prices have responded. Whether the war is over is a question for the shipment ledger, not the press release.
The Ledger’s Verdict
So what does the paper trail actually establish? Three things. First, an eight-company pricing floor exists in writing, dated August 6, 2026. Second, module quotes rose 1-4 cents per watt within eight days, and wafer 183N rose about 40% in a week — documented by an independent pricing service. Third, the durability of both is unverified pending shipment data.
Follow the paper trail; it always leads somewhere. This trail leads to a clear verdict: the solar industry has moved from words to price discipline, and that is real progress. But the receipts are half the story — the other half is whether demand accepts the new floor. One document is an accident; six are a pattern. We have two receipts so far. The third, the shipment ledger, has not been written yet.
The History of Anti-Collusion Pacts
Follow the paper trail on the “anti-involution” initiative and the first thing you find is that this is not the first time an industry has tried to sign its way out of a price war. The eight polysilicon leaders committed in early August to price at no less than full cost and to hold supply — a classic output-restraint pact, dressed in the language of self-discipline. The receipts from previous industries are not encouraging, and that history is the first document in the file.
Steel did the same thing a generation ago, with capacity cuts and industry compacts, and the result was partial: the compacts slowed the bleeding but did not stop the cycle, because the fundamental overcapacity remained. Cement, glass, and a dozen other industries have cycles of “rational self-discipline” followed by quiet price cutting when demand weakens. The receipts tell the story the press release won’t: pacts hold when demand is stable and break when it is not. Polysilicon’s test will come at the first demand dip after the signing.
The mechanics of this pact are worth reading closely. Price at no less than full cost is a floor, not a target — it signals the majors are done funding the downturn, but it also signals they expect the downturn to continue. “Holding supply” and “sealing warehouses” (fengpan) are the operational receipts: the industry is choosing to carry the cost of idle capacity rather than sell below cost. One document is an accident; six are a pattern — and here the pattern is coordinated supply management by the largest eight.
The legal question hangs over the whole file: Chinese competition law has been applied to such pacts before, and “industry self-discipline” that becomes price-fixing is not automatically exempt. The receipts so far show an initiative; they do not yet show the fine print of how the floor will be monitored. The third document, the enforcement mechanism, has not been produced.
The Price Move That Followed
The second receipt in the file is the price action: by mid-August, module quotes were up 1–4 fen per watt, and the 183N silicon wafer jumped from 0.8 yuan per piece to 1.12 yuan — up about 40% in a single week. The move is real, and the receipts show it clearly. The question is whether it is the start of a recovery or the peak of a bounce.
Read the 40% wafer move carefully. A one-week jump of that size is not a market re-rating; it is a supply shock caused by the fengpan posture — sellers withdrew, and the remaining liquidity repriced fast. The same pattern appears at every trough of this industry’s history: a coordinated hold produces a sharp spike, then the market tests whether end demand can absorb the new level. The spike is the pact working; the absorption is the verdict.
The receipts do not yet show the third leg: the shipment ledger. Production numbers, inventory drawdowns, and new-order rates for the weeks after the price move will tell us whether the 1.12 yuan wafer holds. The two receipts we have — the signing and the spike — are necessary but not sufficient. The paper trail demands the third document, and it is not in the file yet.
What the ledger will show, when it is written: whether the majors actually held output while prices rose, or whether the rebound was front-loaded selling into strength. That distinction — between discipline and liquidation — is the whole difference between a genuine floor and a false dawn. Follow the chain; the third receipt settles it.
And the final entry in the receipts file, for now: the signing and the price move are two documents; the shipment ledger is the third, and it has not been written yet. The industry’s floor will be tested not by the initiative but by the absorption — whether end demand accepts the new price level and whether the majors hold output while it is tested. Follow the chain; the third receipt settles it. One document is an accident; six are a pattern. We have two, and the pattern is still open.
And the final entry in the receipts file: two documents in hand — the signing and the price move — and the third, the shipment ledger, still unwritten. The floor will be tested by absorption, not by initiative. Follow the chain; the third receipt settles it. One document is an accident; six are a pattern. The file is open, and the paper trail is the only honest way to close it.
And the final entry in the receipts file: two documents in hand, the signing and the spike, and the third — the shipment ledger — still unwritten. The floor will be tested by absorption, not by initiative, and absorption is measured in the shipment data. Follow the chain; the third receipt settles it. One document is an accident; six are a pattern. The file is open, and the paper trail is the only honest way to close it.
And one more line in the same file, carefully: the fengpan posture works only while the largest eight hold the line together. The history of such pacts is that they are tested at the first weak order book — and the test is usually private. The receipts that matter next are not the press statements but the shipment data and the actual transaction prices at the spot level. Follow the chain; the third receipt settles it. The file is open, and the paper trail is the only honest way to close it.
The Third Receipt, Expected Shortly
The next document in the file will be the one that matters: the shipment and pricing ledger for the weeks after the spike. If the transaction prices hold at the new level while volumes recover, the floor is real. If the prices drift back while volume stays thin, the spike was a withdrawal, not a recovery. Follow the chain — the third receipt settles it. The paper trail is the only honest way to read this industry, and the trail is still open.
And the final line on the file, carefully: the receipts so far — the signing and the spike — are real, but the ledger is judged on the third entry. The shipment data will tell the truth about whether the floor holds. Follow the chain; the third receipt settles it. The paper trail is the only honest way to close the file, and the file is still open.