Follow the paper trail; it always leads somewhere. The story of the storage supercycle is written in three documents: a contract-price list for the second quarter, a half-year earnings statement released on August 31, and a market-share table from Counterpoint. One document is an accident; six are a pattern. Three documents from three independent sources, all pointing the same direction — that is enough to stop reading press releases and start reading the trail itself.
The receipts tell the story the press release won’t. Start with the contract prices, because they are the most direct evidence and the least flattering to every narrative about “AI-driven demand.” In Q2 2026, general-type DRAM contract prices rose 58-63% quarter-over-quarter. NAND Flash contract prices rose 70-75%. On its own, one quarter of sharp pricing is noise; a trade can spike for any number of reasons. But two consecutive quarters of this magnitude is not a coincidence, and the industry itself is calling it a once-in-15-years supercycle. That phrase appears in the record — attributed to industry sources cited by China Economic Net on August 20.
Now the second document, and this is where the trail gets interesting. ChangXin Memory, the domestic DRAM leader, published its half-year results on August 31: revenue of 150.31 billion yuan, up 873.64% year-over-year, and a net profit of 77.605 billion yuan — swinging from a loss to a profit in a single reporting period. The company also expects tight DRAM supply to persist through the second half. A 77.6 billion yuan swing is not a rounding error. It is the financial consequence of a market in which the marginal supplier sets the price, and in a tight market, the marginal supplier has pricing power over everyone.
The market-share table is the third document, and it changes the story
Counterpoint’s Q2 data adds the structural layer: a domestic storage leader’s NAND shipment share entered the global top three for the first time — 25%, 22%, and 14% being the shares of the leading players in the table. The receipts here are unambiguous. A domestic producer, for the first time in the NAND segment, sits among the top three by shipment share. This is the same supplier family that, a few years ago, the market would not have expected to appear anywhere near this table. The documents do not say the industry is “catch up” — they say it has arrived at a position where it is one of the people setting the prices.
Cross-check the three documents against each other and the chain of custody is consistent. Contract prices spike 58-75% in a single quarter; the leading domestic supplier swings to a 77.6 billion yuan profit; that same supplier’s NAND share enters the global top three. The trail does not support the version of the story where foreign suppliers keep all the pricing power. It supports a version where the market has structurally reorganized — and the domestic producer is inside the pricing loop, not outside it.
But the receipts also show who is paying
No investigative reading stops at the winners. Follow the chain further and ask who pays for a 70-75% NAND price increase. The answer, in the receipts, is downstream: server makers, phone makers, and ultimately the consumer. That is the part of the trail the celebratory press releases skip. A supercycle is not a neutral event. It transfers wealth from buyers to sellers in a hurry, and every transfer produces a matching set of receivables and complaints.
And the trail has a complication worth noting. Samsung raised prices on new 4nm/5nm/8nm orders by up to 15%, and SMIC has followed with its own increases — that is the foundry side, not memory, but it is the same paper trail showing capacity constraints propagating through the semiconductor chain. When foundries and memory makers are both raising prices in the same window, the story is not one company’s pricing strategy. It is an industry-wide supply constraint, and the constraint is the signal.
Let me think about what a 15-year cycle actually means operationally. Memory is the most cyclical commodity in technology, and a genuine supercycle is rare precisely because capacity takes years to build and years to retire. Fabs are not switchable assets; a wafer that runs DRAM cannot quickly become a wafer that runs logic. So when demand swings, supply cannot respond in quarters — it responds in capital cycles. The last comparable episode was roughly fifteen years ago, which is why the industry uses the phrase “15-year supercycle.” The pricing data, the earnings, and the share table all sit inside that long-dated reality. This follows the chain of causation from capacity to price to earnings in exactly the order you would expect.
What the documents do not show
I want to be careful about what the documents do not show. They do not show how long the cycle lasts. Fifteen-year supercycles, by definition, come with fifteen-year dry spells between them, and the timing of the bust is never visible in the boom’s receipts. The half-year statement says supply tightness persists through H2; it does not and cannot date the turning point. That uncertainty is part of the trail, and a reader who ignores it is reading selectively.
There is also the question of what the share table is measuring. NAND shipment share is a volume metric, not a revenue metric and not a profit metric. A producer can hold a top-three shipment share while earning less per gigabyte than its rivals, particularly if it is buying its way up the curve on price. The receipts show the position; they do not show the margin structure under it. That is a real gap in the evidence, and it argues for reading the market-share document as a footprint rather than a verdict.
Let me also check the timing, because dates matter in this kind of record. The contract-price data covers Q2 2026 and was reported around August 20. The ChangXin results were released August 31. The Counterpoint share table covers Q2 and was cited around August 21. Three documents, three publication windows, one consistent picture. If the sources disagreed on timing or direction, the whole analysis would collapse. They do not. That is what a pattern looks like when it is real — not a single explosive number, but multiple records converging on the same conclusion from independent directions.
What the boom does not explain
There is one more layer of the trail worth pulling before closing the file, and it is the layer the celebratory reading tends to skip. A supercycle explains the price increase, the earnings swing, and the share table. It does not explain where the demand came from in the first place, and the answer to that question matters for how long the pattern persists. The demand driver is not a single product refresh; it is the scale-out of AI compute, which has pulled memory capacity into allocation for training and inference clusters. That is a structural demand story, not a one-quarter stocking event — and the receipts, read carefully, point to it: DRAM tightness persisting into H2, foundries at full load, NAND at record pricing. Structural demand is the reason the cycle can run for fifteen-year intervals; it is also the reason the eventual correction, when it comes, will be sharp. Both halves of that sentence are in the record.
And there is a final receipt to add: the counterfactual. The documents would look very different if this were a normal cycle. Prices would move in single digits, a domestic producer would not swing 77.6 billion yuan in a single half, and a first-time top-three NAND position would not land in the same quarter as 70%+ contract-price increases. The fact that all three arrived together is the record’s own argument for calling it a supercycle. The trail does not need the word “supercycle” to be true; the numbers support the label on their own. That is what a well-documented turning point looks like.
How the documents would be challenged
Let me stress-test the trail the way a skeptical reader would, because a receipt only means something if it survives challenge. The strongest challenge to the price data is that contract prices are not transaction prices — contracts are negotiated, and not every buyer pays the same. That is true, and it is exactly why the earnings document matters: a domestic producer swinging 77.6 billion yuan into profit in one half is a transaction-level confirmation, not a contract-level one. Money earned is the most audited receipt in the file. The second challenge is to the share table: top-three NAND share by shipments is a volume measure, and volume leadership is not the same as value leadership. That is fair, and it is why the documents are read as a set — the earnings confirm the value, the share table confirms the position, and the price data confirms the environment. No single document carries the argument; the set does.
The third challenge is the one I would take most seriously: cycles end, and documents written at the top of a cycle look foolish at the bottom. The way to meet that challenge is not to deny it but to date the evidence. The documents cover Q2 2026 and the reports published in late August — they are records of a current state, not prophecies. Nothing in them claims the supercycle is permanent; the industry’s own phrase, “15-year supercycle,” implies the arithmetic of its own ending. The honest reading is a real turning point with an unspecified timer. That is the position the evidence supports, and it is the position this file will stand on.
The chain resolves to a single sentence
Here is where the paper trail ends up. The storage industry is in a genuine, industry-acknowledged supercycle — the first in roughly fifteen years — and this time the domestic producer is not on the sidelines watching foreign suppliers take the pricing. It is in the table, in the top three, with the earnings to prove it. The receipts tell that story, and they tell it better than any press release could.
Follow the chain one step further, and the practical question for anyone reading: the cycle that rewards the sellers will also test them. The margins on the way up are easy; the discipline on the way down is what the records will judge.
The way to read it going forward is to keep the three documents on separate shelves and check them on a schedule. The contract-price list refreshes quarterly and will tell you when the cycle peaks. The earnings statement refreshes half-yearly and will tell you when the profitability under the boom turns. The share table refreshes quarterly and will tell you whether the top-three position holds or was a one-season artifact. Three documents, three cadences, one trail. Watch the cadences, and the trail will never surprise you — it will just keep leading somewhere.
For now, the documents are clear, the pattern is real, and the position of the domestic supplier has changed permanently. That is the story the trail tells — and it always leads somewhere.