How does the math work here? That is the question to keep in your pocket while reading the storage headlines of the last fortnight, because the numbers are good enough to make you stop being sceptical — and that is exactly when the ledger matters most. A company swinging to a profit of this size, on revenue growth of this size, deserves a close read, not a cheer. The quarterly rhythm tells you more than the press release, and the rhythm right now is a price cycle doing very heavy lifting. On the surface, the math works; the question is how much of it is the tide and how much is the boat.
The supercycle, in numbers
Start with the market itself, because the market numbers are the substrate everything else sits on. In the second quarter of 2026, general DRAM contract prices rose 58 to 63 percent quarter on quarter. NAND Flash contract prices rose 70 to 75 percent. That is not a healthy correction or a normal seasonal bounce; industry observers describe the first half as a once-in-fifteen-years super cycle, and the arithmetic supports the adjective. A 60 percent price rise in a quarter is the kind of move that rewrites every income statement in the industry at once.
Let me think about what a price rise of that size does to the math of any memory maker. Memory is a commodity with a brutal cost structure — enormous fixed investment, cyclical pricing, thin margins in the trough. When contract prices jump 60 to 75 percent, the incremental revenue goes almost straight to the bottom line. Fixed costs do not move. That is the first thing the ledger shows: in a supercycle, the profit curve is steeper than the revenue curve, and that steepness is a price effect, not a strategy effect. The same company looks dramatically more profitable in this quarter than in the last one, and the reason is the market, not a change in the business.
The balance sheet that turned around
Now read the flagship result in that context. CXMT, the leading domestic memory maker, reported first-half revenue of 150.3 billion yuan, up 873.64 percent year on year, and a net profit of 77.6 billion yuan — a swing from loss to a very large profit. Let me say plainly that this is a real result and a real number. The revenue growth is verified, the profit is stated in the half-year report, and the swing is the largest thing in the story.
But I want to be healthily sceptical about what the swing proves. An 873 percent revenue increase in a year is not a growth story in the normal sense; it is a price story with a volume component. Memory companies do not sell 873 percent more chips in a year; they sell somewhat more chips at dramatically higher prices. The distinction matters because it decides what you can extrapolate. If you read the headline as “this company grew ninefold”, you will draw the wrong conclusions. If you read it as “this company was fully exposed to a 60-75 percent price surge and captured it”, you are reading the ledger correctly.
There is a second line worth pulling out. The company’s own outlook expects global DRAM supply tightness to continue into the second half. That is a useful data point, and it is also, unavoidably, a company statement — a company deep in a supercycle has every incentive to expect more supercycle. I will take the expectation seriously and mark it as a forecast rather than a fact. The supply tightness is real in the present; its persistence is a wager on the future.
The price, not the product
Let me now ask the question the ledger forces. If most of the profit is price, what happens when price stops rising? The answer is that the income statement compresses as fast as it expanded — the same leverage that multiplied the profit will divide it. That is not a cynical reading; it is the definition of a cyclical business. The companies that matter are the ones that use the supercycle to buy something durable: capacity, process technology, customer contracts, share.
Let me correct my own first instinct here, because I want the record to be honest about the skepticism. When I first saw the 873 percent revenue figure, my reflex was to look for the catch — the accounting treatment, the one-off, the number that would dissolve on inspection. I did not find one, and I will not pretend I did. The profit is genuine, the swing is real, and the right question is not whether to believe the number but how much of it survives a change in the price environment. Correcting that instinct matters, because a skepticism that refuses to update is just bias wearing a ledger’s clothes.
And here the story gets genuinely interesting, because there is evidence of durable purchases being made. Samsung has raised prices by up to 15 percent on new orders across 4nm, 5nm and 8nm nodes, with its Pyeongtaek 4nm lines running full. SMIC has followed with price increases. Foundry pricing moving up across the board is the sign of an industry that is not just enjoying a spot price spike but is re-pricing its forward capacity — and capacity that gets re-priced is capacity that gets built. That is the part of the story that compounds, not just the part that spikes.
Where the money goes from here is the question the ledger answers last, and the answer decides the whole story. Into the things that let a cyclical player hold share when the cycle turns: leading-edge capacity, advanced packaging, process maturity. The CXMT balance sheet is a swing to profit, but the durable question is what the profit is being spent on while it lasts. A supercycle is the only time a memory maker has the cash to build for the next downturn, and the ones that treat it that way are the ones worth watching when the prices soften.
The global share shift
The other document in the pile is a share table, and it is the quietest proof of structural change. Counterpoint’s data shows the leading domestic NAND player reaching the global top three in shipment share for the first time in the second quarter, in a field where the top positions carry roughly 25, 22 and 14 percent shares. Let me read that carefully. It is not a claim of technological leadership; it is a claim of commercial mass. NAND is a market where volume determines cost position, and cost position determines survival through the trough. Getting into the top three is how a player buys a seat at the table for the next cycle.
I will be honest about what the share number does not prove. It does not prove the product is best-in-class on every metric. It does not prove the brand holds pricing power when supply loosens. It proves volume, and volume is the foundation of everything else in this industry. That is a solid foundation, and it is also — like the profit swing — partly a gift of the cycle. When contract prices rise 70 to 75 percent, every player’s share arithmetic benefits from the same wave; what the top-three entry shows is that this player converted the wave into a permanent seat.
Who is paying for the surge
There is a demand side to this ledger that the supply-side numbers tend to crowd out, and it is worth a paragraph on its own. The reason contract prices can rise 60 to 75 percent in a single quarter is not that memory companies got braver; it is that AI compute ate the market’s spare capacity. Training runs and inference clusters consume enormous amounts of fast memory, and every new data center ordered chips first and memory alongside them. When the biggest buyers on earth are building in parallel, the storage industry’s normally generous buffer of unused capacity disappears — and a commodity without a buffer is a commodity that prices violently.
Let me connect that to the numbers in front of us. The same supercycle that lifted DRAM contract prices by 58 to 63 percent and NAND by 70 to 75 percent is the demand story that made those price moves possible in the first place. The cyclical reading and the structural reading are not rivals; they are the same page of the ledger seen from two columns. The price spike is cyclical. The demand that supports it is structural. The investor who reads only one column will misprice the other.
The history lesson in the margin
I want to add a note on the fifteen-year framing, because the phrase does real work and deserves a sceptical look. Calling the current half-year a once-in-fifteen-years supercycle is a way of saying the magnitude is unusual without saying the direction is permanent. Memory has a documented habit of these cycles — deep troughs followed by violent recoveries, each one convincing a new cohort of observers that this time the cycle has been abolished. It never has been, and I have no reason to believe this cycle is the exception.
Let me be careful to state what is, and is not, being claimed. The market is genuinely tight today; the supply forecast for the second half is genuinely constrained; the quarterly results are genuinely large. None of that requires believing the cycle has ended. It requires the opposite discipline: treating the present strength as real while building a position that would survive the next downturn. That is the difference between reading the balance sheet for this quarter and reading it for the next decade.
And that discipline points back to the one durable line in the whole story. The top-three NAND seat, the rising foundry prices, the re-pricing of forward capacity — these are the purchases that survive the cycle. When the price spike fades, as price spikes do, the companies left standing will be the ones that spent the windfall on structure rather than on celebrating the windfall. That is the ledger line I would underline, and it is the line this quarter’s headlines tend to skip.
Fair enough — but read the columns
Let me sum the columns the way I would on the train, working through the ledger line by line. Revenue: extraordinary, and mostly price. Profit: real, and mostly the same price effect. Outlook: positive, and marked as a forecast. Foundry pricing: rising across the board, which is the most durable signal in the pile. Share: a first-time top-three position, which is structural. Balance the columns and the verdict is clear enough: the supercycle’s money has genuinely landed, the domestic champion has genuinely turned profitable, and the global table has genuinely changed. None of that is fiction.
But here is the sentence I would underline, in ink. A commodity price surge is not a business model, and a turnaround built on a 60 to 75 percent price move is not a turnaround until it survives a downturn. The industry has been here before — every supercycle in memory has ended with prices softer than anyone expected and capacity decisions made in the boom looking expensive in the bust. The companies that matter will be visible by what they built with this quarter’s cash, not by the quarter itself.
Fair enough — but don’t call it a turnaround yet. Call it what it is: a real, documented, cyclical windfall, captured well, with some genuine structural progress underneath. That is a good story. It just is not the story the headline suggests. The quarterly rhythm tells you more than the press release, and the rhythm says the music is loud right now. The interesting question is what the industry does with its voice while it has one.