The only tension that matters
Longsys is not a fab. It buys DRAM and NAND wafers from the upstream giants, packages and controls them, and sells finished product under Lexar, Zhitai, and an enterprise and embedded line. I said in the Q1 write-up that this model is the story of the print, and the half-year report makes it undeniable. Revenue of RMB 24.088B (~US$3.35B at ~7.18 CNY/USD, approximate) rose 136.26% against RMB 10.196B. Net profit attributable to shareholders was RMB 10.577B (~US$1.47B) versus RMB 14.77M a year earlier — the +71,528.66% that made every headline, and the least informative number in the document, because the denominator was effectively zero. The figure that actually carries weight is smaller and stranger: net profit ex-non-recurring items of RMB 10.047B, meaning roughly 95% of the bottom line came from operations rather than one-offs. Basic EPS was RMB 25.20. Weighted average ROE was 79.03%. Against FY2025's full-year net profit of RMB 1.423B, this half earned 7.43x the entire prior year.
And in the same six months, operating cash flow was −RMB 3.151B, against +RMB 693M a year ago. Cash paid for goods and services was RMB 24.469B while cash received from sales was RMB 23.352B. Inventory went from RMB 11.678B to RMB 25.777B. The standard reading of that pairing is that the profit is somehow less real than it looks. I think that reading is lazy, and it is exactly the kind of place where the house edge lives — the gap between the narrative and what the data supports. The profit is real and it was earned in cash terms by the previous purchase decision; what is unfinished is the conversion. Longsys did not report a quarter. It reported the mark-to-market on a position it is still carrying.
Read the profit and the cash side by side
Three figures tell the whole structure of this half, and stringing them through a sentence is exactly the thing readers can't hold. The chart does it in one look.
Now decode it. The company's own language for the inventory is "strategic stockpiling" (战略性备货), executed through long-term agreements and MOUs with upstream suppliers because, in its framing, AI demand has tilted wafer capacity toward enterprise parts and left commodity supply structurally short. Strip the corporate register and that is a plain statement of intent: we deliberately went long inventory ahead of a price move we expected. Inventory turnover fell from 1.8x to 1.2x and inventory days extended from 99 to 152. Read conventionally, slowing turns are a demand warning. Read against the stated strategy, 152 days is the holding period of the trade — you cannot simultaneously praise a company for locking supply ahead of a shortage and mark it down for holding what it locked.
The rest of this dive is for paid subscribers.
The headline numbers are above. The 9 sections below carry the mechanism, the peer read, the valuation work, and the dated tests that decide it.
- A 59% gross margin at a company that owns no fab
- The line nobody printed
- Who funded the warehouse
- The strongest case against me
- How the rest of the stack reads it
- Valuation: paying for a holding-period margin
- What decides it from here
- The trade (analysis, not advice)
- Bottom line
Already a subscriber? Sign in.