YMTC Pre-IPO — One Quarter Nearly Cleared the Deficit, and the Tax Holiday Hasn't Started

The setup

YMTC’s prospectus is being read as a margin story. It is better read as a reset. In the first three months of 2026 the company earned RMB 33.38B (~US$4.96B) of attributable net profit on RMB 47.04B (~US$6.99B) of revenue, and that one quarter took the accumulated deficit from RMB 36.78B (~US$5.47B) to RMB 3.40B (~US$0.51B). Our test is the next financial update in the review, the first one with a period-end after March 31: the quarters after Q1 should average more than Q1’s RMB 47.04B of revenue. If they average less, the inventory YMTC built in its best quarter was not being held for price. It was product nobody took.

How big is the quarter? Bigger than any year the company has had. Full-year 2025 attributable profit was RMB 14.21B (~US$2.11B); Q1 alone was 2.35x that, and 1.59x the RMB 20.98B YMTC made across both of its profitable years combined. It is 1.74x the RMB 19.18B (~US$2.85B) the company lost in 2023, the worst year in the series. Q1 revenue was 74.45% of all of 2025. What the quarter is not is a run rate: on the filing’s own count, the average NAND selling price in Q1 was 172.72% above the 2025 full-year average.

The filer is the holding company, 长江存储控股股份有限公司 (CCSH), which applied to the STAR Market on August 21, 2026. The exchange moved the file to “inquiry issued” on September 3, and as of October 16 it had not moved again. The plan is 1.98–2.43B new shares, 10–12% of the post-offer total, for RMB 33.0B (~US$4.90B) of proceeds: RMB 20.8B (~US$3.09B) to upgrade the production lines and RMB 12.2B (~US$1.81B) for R&D. There is no controlling shareholder. Hubei Changsheng holds 26.54%, Xinfei Technology 25.35%, and the two Big Fund vehicles 11.97% and 11.38%. The Wall Street Journal, as carried by Yahoo Finance, reports that the company expects to list in 2027.

The numbers

Read the ingredients, not the level. The two growth years were different animals. In 2024 NAND volume rose 25.63% and price 125.11%. In 2025 revenue still grew 39.78%, but NAND volume rose 64.74% while price fell 13.76%: a volume year. Q1 2026 is a price year, and the volume lever is gone. Capacity utilization was 98.02%. The plant is full, so whatever the next quarters add has to come from price or from the warehouse.

The company explains the margin in one sentence, and it is the most useful sentence in the margin section: a high share of NAND cost is labour, depreciation and utilities that are “relatively stable in the short term,” so the part of the price above the mostly-materials variable cost “forms new gross profit” (产品售价高于以材料成本为主的变动成本的部分形成新增毛利). Decode: the 78.73% NAND margin is operating leverage on price, and operating leverage works in both directions. The 2023 column shows the same machine running in reverse. The reported NAND margin that year was 1.00%, but only because earlier inventory write-downs were released through cost of sales. Without the release it was −96.72%, and the year carried RMB 11.35B of inventory and contract-cost impairments. The prospectus is selling the top of a curve the company itself says is fixed-cost-driven.

What’s actually driving it

Three things the headline margin does not show, and each one is in the filing.

The warehouse. Sell-through (sales volume over production volume) was 82.40% in Q1, against 97.48% in 2025. Inventory rose RMB 1.97B, or 11.86%, to RMB 18.59B (~US$2.76B). A maker that sells 82.40% of its output in a quarter when its average price is 172.72% above last year’s is doing one of two things: holding product back for a higher price, or failing to place it. The thesis library calls the first one the warehouse is the position. We coined it for module houses, where gross margin is a holding-period spread. Here the maker is running it on its own output. Reports citing TrendForce put second-quarter contract increases at roughly 60%, and if that is right, every unit held back out of Q1 was repriced on the way out.

The series supports the holding reading over the stranding one. The CXMT miss taught us that a one-period change is a phase until you have the series, and this series has a pattern. Sell-through was 108.37% in 2023 (drawing down stock into a falling market), 89.18% in 2024 (building while price rose 125.11%), and 97.48% in 2025 (selling the build in a volume year). YMTC has built inventory when price was rising and released it when volume was the lever. Q1 matches the 2024 pattern. The update tells us whether Q2 matched 2025.

The export book. In Q1, 63.28% of NAND revenue, RMB 28.63B (~US$4.25B), was booked outside mainland China, up from 43.36% in 2024. That is 3.53x the 2025 quarterly average of overseas sales and 1.66x all of 2024’s. YMTC has been on the US Entity List since December 2022. The restriction governs what it can buy, not whom it can sell to, and the growth went abroad. The prospectus names the exposure itself: domestic NAND is “a key area of international trade controls” in equipment, parts and materials “and overseas sales” (境外销售). The filing does not split 境外 any further. It is outside the mainland, which can include Hong Kong and offshore entities of Chinese buyers, so do not read 63.28% as 63.28% sold to foreign customers.

The tax clock. Q1 pretax profit was RMB 33.52B (~US$4.98B). Income tax was RMB 37.97M (~US$5.64M), an effective rate of 0.11%. In most turnarounds a near-zero rate means loss carryforwards that run out, and the after-tax margin falls as they do. This one is built the other way. The filing says YMTC qualifies as an IC manufacturer at 28nm or below, which earns full income-tax exemption for years one through ten, “counted from the profit year” under tax law. It says the company “has not yet actually enjoyed” the exemption because that year “had not been reached” in the reporting period. With the book deficit down to RMB 3.40B, the clock looks close to starting. That is our reading: taxable losses and the book deficit are not the same number, and the filing does not give the tax figure. Call it the holiday starts after the windfall. The near-zero tax line is not a turnaround artifact on its way out. It is the opening of a decade-long exemption.

None of this removes the bill underneath. The carrying cost is senior to the shareholder: over the period YMTC spent RMB 96.39B (~US$14.32B) on long-term assets and booked RMB 50.95B (~US$7.57B) of depreciation and amortization, and at March 31 it carried RMB 88.81B (~US$13.20B) of borrowings against RMB 51.99B (~US$7.73B) of cash. Q1 capex of RMB 7.87B (~US$1.17B) was 86.98% of 2025’s quarterly average of RMB 9.05B. CXMT taught us not to call that restraint, because the series has capex up 45.49% in 2024 and 16.70% in 2025, and the RMB 20.8B line-upgrade project is waiting on the raise. Q1 capex used 29.43% of the quarter’s RMB 26.74B (~US$3.97B) of operating cash flow.

How peers read this

Kioxia and SanDisk: the growth is landing where they sell. Counterpoint ranks YMTC third in NAND shipments for Q2 2026 at 16.5%, and has its revenue share going from 9% in Q1 2025 to 14% in Q1 2026. Put that next to the 63.28% export share and the competitive read is direct: YMTC’s increment is going abroad, not into a protected domestic market. On the prospectus table, YMTC’s NAND margin sits 14.62 points above Kioxia’s and within half a point of SanDisk’s. In this cycle gross margin does not separate a sanctioned supplier from an unsanctioned one. Access to tools does, and that shows up in the next build, not in this quarter.

Longsys and Biwin: the supplier is also the competitor. Memory chips (wafers and dies sold as such) were 50.84% of Q1 main-business revenue, and their average price was 218.00% above the 2025 average. That is the module houses’ input cost. The filing adds that chip volume grew more slowly than total NAND volume because YMTC put more of its own dies into its own smart-terminal and SSD products, which were 35.53% and 10.05% of the quarter. For Longsys and Biwin, a supplier that keeps more die for itself in a shortage tightens the lane twice.

CXMT: the twin, one step ahead. CXMT’s 79.16% margin is within half a point of YMTC’s NAND line, and its listing sets the precedent for how STAR prices this. Its plan was RMB 29.5B, and the book-built raise came in at RMB 57.92B (~US$8.61B), 1.96x the plan (CXMT IPO dive).

Micron: the litigation now runs both ways. Micron sued YMTC in federal court in Idaho on September 15, alleging it hired Micron engineers and patented their know-how, and asking for ownership of four US patents. Three days later a Munich court found Micron infringed two YMTC utility models and enjoined the products in Germany; Micron is appealing. The prospectus counts 9 patents in ownership disputes and 36 in invalidation proceedings at June 30. The filing’s counsel calls the ownership-case risk low. The size of YMTC’s overseas book is what makes the question expensive (Micron Q4 FY26 dive).

Valuation reality check

There is no price yet. What the filing fixes is the share count: 1.98–2.43B new shares into a post-offer total of 19.80–20.25B. If the raise equals the RMB 33.0B plan, the offer price works out to RMB 13.58–16.67 (~US$2.02–2.48) and the post-offer value to RMB 275–330B (~US$40.86–49.03B), which is 2.06–2.47x annualized Q1 profit of RMB 133.52B (~US$19.84B). That is plan arithmetic, not a forecast. CXMT raised 1.96x its plan and debuted near RMB 3.3T. Market reports put recent secondary transfers of YMTC shares at about RMB 400B. The planned figure fixes the minimum money the company wants. Book-building fixes the price.

Annualizing Q1 is the error the filing itself warns about. Its price-risk section says the length of the shortage is uncertain and a supply response can bring prices back, and the 2023 column shows what that looks like. The better denominator is the cycle. Q1 EBITDA of RMB 38.75B (~US$5.76B) was 1.15x all of 2025’s RMB 33.83B. A buyer at the listing is paying for how many quarters like that the next build allows, with a tax line the filing says has not yet started its holiday.

The best case against this. A bear reads the same balance sheet as a channel that has stopped absorbing. Inventory up, receivables up 139.45% in one quarter to RMB 22.84B (~US$3.39B), sell-through at the low of the series, and five customers taking 54.34% of revenue: at the top of a price spike, that is what it looks like when buyers take product on terms and the maker parks the rest. If Q2 contract prices held but volumes did not, the warehouse is stranded stock carried at cost, and a reversal turns it into the 2023 impairment line again. That case is serious, and it is why the call is on revenue, not margin. We still land the other way for two reasons. Annualized receivables turnover rose to 11.62 from 8.43 in 2025, so collections got faster relative to sales, not slower. And the last time sell-through dropped below 90%, in 2024, the next year was the volume year.

What we’d watch from here

• Revenue per quarter after Q1 vs RMB 47.04B. The next update in the review will carry a later period-end. If the quarters after March 31 average above RMB 47.04B, the Q1 warehouse was held for price. If they average below, it was demand that was not there. This is the call.
• Sell-through vs 82.40%. A move back toward 2025’s 97.48% means the build was sold. Another period below 90% at full utilization turns the holding reading into a stocking one.
• The profit year under tax law. If the update says the ten-year exemption has started, the 0.11% tax line becomes structural, not transitional.
• The export share vs 63.28%. Any new US action aimed at sales rather than tools would hit the part of the book that grew.
• The inquiry response and the raise. The file has sat at “inquiry issued” since September 3. The final raise against the RMB 33.0B plan will show how far STAR book-building takes the price.

The trade

There is nothing to trade in YMTC itself until it lists, and the company reportedly expects that in 2027. The tradeable part of this filing is the read-through. For holders of Kioxia and SanDisk, the 63.28% export share and the 16.5% shipment share make YMTC a direct competitor outside China. That is a share question for the next build, not a margin question for this quarter. For the module houses, it is an input-cost and allocation question while the shortage lasts. Analysts covering the China memory complex should treat the listing as a capital event for supply: RMB 20.8B is earmarked for the lines, and CXMT shows the raise can come in well above plan. We have not priced strikes here.

The risk to that read is the same one in the steelman. If the next update shows the warehouse did not sell, YMTC’s fixed-cost machine is pointed the wrong way, and its competitors get a stranded-inventory supplier in a market that is turning, which is a different problem from a capacity build.

Bottom line

The prospectus sells a 78.73% margin. The filing documents something more durable: one quarter cleared 90.75% of the deficit the company had built since 2016, the income-tax holiday it qualifies for has not yet begun, and the growth is being booked outside mainland China despite the Entity List. The weak point is the warehouse, an 82.40% sell-through in the best pricing quarter the company has had. The call: in the next financial update, revenue per quarter after Q1 averages above Q1’s RMB 47.04B. Wrong if it averages below.