The Crack in ChangXin’s Blockbuster Earnings

When ChangXin Technology released its third-quarter 2026 earnings report, the market produced two readings of the same document.

The first reading looks at year-on-year numbers. Revenue for the first three quarters was RMB 269 billion, up more than sevenfold from a year earlier. Net profit was RMB 132 billion, compared with a loss in the same period last year. On those numbers alone, the report is clean. It is a textbook case of a domestic memory leader cashing in on a super-cycle.

The second reading looks at quarter-on-quarter numbers. Q3 net profit was RMB 54.435 billion. Q2 net profit was RMB 52.843 billion. The sequential increase was 3.01%. In Q2, the sequential growth rate was 113.4%. The shift from doubling to standing still took one quarter.

The market chose the second reading. ChangXin fell 7.77% on October 19. From September 30 to October 30, it fell 24.75% over the month. A report with explosive year-on-year growth met a sustained share-price decline. The disagreement sits in the sequential data. That data is closer to the company’s current position than any single quarter’s year-on-year figure.

1. Sequential profit growth has already peaked before prices did

Q3 revenue was RMB 118.7 billion, up 19.3% from Q2. Net profit was RMB 54.435 billion, up 3.01%. Nearly RMB 20 billion of extra revenue produced less than RMB 1.6 billion of extra net profit. Revenue kept expanding. The marginal conversion from revenue to profit fell sharply.

“Growth slowed” is too gentle for that change. The engine that drove ChangXin’s profit surge in the first half of 2026 had lost most of its marginal thrust by Q3. Profit doubled from Q1 to Q2. From Q2 to Q3, it almost stopped moving. The profit curve did not continue rising. It climbed steeply, then flattened.

One explanation points to the base effect. The first two quarters pushed margins to a high level. When price increases normalized, profit growth naturally converged. That explanation works arithmetically. It misses the main point. The base itself was built on unsustainable price increases. Here the phrase “high base” carries direction. Year-on-year data still describes the past twelve months. Sequential data already points to what comes next.

2. The price engine is losing thrust

ChangXin’s profit explosion did not come from linear shipment growth. It came from the steep rise in memory prices during the super-cycle. From Q1 to Q2, DRAM prices rose fast under supply contraction and AI demand. ChangXin, a mature-node manufacturer with a cost advantage, saw its profit elasticity to price increases amplified to the extreme. In Q3, that elasticity failed.

Jiangbolong’s concurrent announcement is worth reading word for word: “The price increases for DRAM and NAND Flash have narrowed significantly compared with the earlier period, and the marginal contribution of memory price increases to the company’s profitability has slowed markedly.” That statement comes from inside the supply chain. It is more direct than any external estimate. “Marginal contribution has slowed markedly” means the same price increase now produces far less incremental profit than it did in the early and middle stages of the cycle. ChangXin’s profit is concentrated in the original manufacturer segment. A slowdown in the price engine becomes a slowdown in the profit engine.

The slowdown reflects a structural shift in the cycle position. It will not disappear next quarter. Memory prices rose steeply from the cycle bottom. That rise naturally flattens as the supply gap narrows and the base rises. The Q3 profit stall is not “this quarter happened to be bad.” It is the financial expression of a simple fact: the easiest stretch of road has been traveled. Price increases narrowed from the Q1 magnitude to the Q3 magnitude. From there, sequential profit growth can converge toward zero or turn negative. That conclusion needs no extra assumptions.

3. Apple: technical qualification without a commercial deal

ChangXin once carried a vivid narrative. It would enter Apple’s supply chain with LPDDR5X and DDR5. It would open a gap among Samsung, SK Hynix, and Micron. “Domestic substitution” would move from a stock game in China to incremental certification by the world’s top customer.

The technical basis for that narrative is real. ChangXin’s 17nm DDR5 yield has exceeded 90%. Samsung’s same-generation yield is 92% to 93%. The gap is two to three percentage points. LPDDR5X speed has reached 10,667 Mbps. That matches the core parameters of Samsung and SK Hynix’s same-generation consumer products. By technical metrics, ChangXin could sit at Apple’s negotiating table.

It did not sit down. Apple did ask ChangXin about LPDDR5X. It wanted a procurement price below Samsung, SK Hynix, and Micron. ChangXin refused the price pressure. It insisted on quoting at least on par with Samsung and SK Hynix, and higher for some specifications. It chose to prioritize domestic DRAM demand.

Refusing Apple’s price pressure makes commercial sense. Domestic demand is full. Capacity utilization is high. Supplying Apple below the international giants would trade scarce capacity for a certification label with compressed margins. That is not an attractive trade. The investment implication still needs a clear statement. The Apple substitution expectation was valid at the technical level. It has not materialized at the commercial level. Any valuation premium tied to “entering Apple” has lost its path to realization. A story that did not happen cannot stay in the price.

4. The buyback signal

ChangXin’s RMB 1 billion to RMB 2 billion buyback plan split the market. People looking at the absolute amount call it small. People looking at the ratio call it large.

The absolute amount is small against profit. Net profit for the first three quarters was RMB 132 billion. A buyback of at most RMB 2 billion equals about 1.5% of that profit. Total share capital is 67.884 billion shares. Even a RMB 2 billion buyback at the upper limit would cover about 0.04% of total shares. The ratio is large against the float. Tradable shares are only 6.63% of total share capital. RMB 1 billion to RMB 2 billion equals 0.5% to 1.0% of free-float market capitalization. That is three to five times the 0.1% to 0.3% typical of single buybacks by A-share hard-tech companies with market caps above RMB 100 billion. That is real support, though limited.

Both numbers are true. They answer different questions. The ratio answers a short-term question: can this money create a chip-contraction effect in an extremely thin float? Yes, to a limited extent. The absolute amount against profit answers a management question: at a profit peak, with industry price momentum fading, how much money will management spend to express a view on its own shares?

A buyback of at most RMB 2 billion against RMB 132 billion of profit is 1.5%. The share price slid from RMB 49 toward RMB 41. The signal reads closer to caution than confidence. A company at the top of the cycle with large cash balances has room to act if it believes its shares are deeply undervalued. It could make a move far more convincing than 1.5%. It chose not to. That choice is information.

5. What the market is pricing

Together, these four facts explain the share-price action.

ChangXin’s year-on-year data remains strong. The market does not price the sum of the past twelve months. It prices the marginal direction of the next quarter. Q3 sequential profit growth collapsed from 113% to 3%. An internal supply-chain announcement confirmed narrowing memory price increases. The Apple substitution story failed at the commercial level. The buyback is too small to change the balance between chips and confidence. The marginal direction points to deceleration. Valuation must move from a scarcity premium to steady-state pricing after earnings realization.

Zhou Xiaochuan’s remarks at the Bund Summit became a narrative catalyst at this node. He said AI risks need continuous observation. His remarks are not the cause. The cause sits in the sequential lines of the earnings report, in Jiangbolong’s wording, in ChangXin’s refusal to accept Apple’s price pressure, and in a buyback worth 1.5% of profit.

Year-on-year figures tell you how fast the company ran. Sequential figures tell you where it is heading. ChangXin’s third-quarter report puts the past in the year-on-year columns and the future in the sequential columns. “All the good news is out” lives in the gap between them.

For writers, this report offers an observational frame. A company at the top of the cycle can deliver a report with clean year-on-year numbers and stalled sequential numbers. The market’s decline reprices the future. The useful question is not “are the results good?” The useful question is “has the marginal direction changed?” ChangXin’s third-quarter report answers both questions on two sides of one sheet.