Jukan转述韩媒/设备业高管口径称长鑫存储HBM3 8-Hi试产良率仅约25%~30%、TSV工艺不成熟,四个里有三个不良
直接主体长鑫科技(688825.SH)已上市且流动性充足,但本条为匿名单一信源、经社媒转述、无法追溯韩国首发原文,且文章自身给出反例(三星HBM4半年内良率从<60%提到80%)明确对冲结论;HBM3仍为试产/送样、无收入贡献,不改变公司当期以标准DRAM为主的盈利驱动。更关键的是时点已失效:推文发布于2026-09-08 22:25 UTC(=北京9月9日06:25,A股开盘前),9月9日688825.SH高开58.90→收56.87、-2.6%的高开低走已完成定价,9月10日+0.58%、9月11日-0.82%走平,三个交易日累计仅-2.2%;读通标的SK海力士同日+3.51%亦已反应。做空红线两条均不满足(非确凿重大负面、置信低),做多则事件方向为负且事件后回落不构成多头催化。二阶设备/材料链方向自相矛盾且无确认订单。故全腿不做。
展开详情(成交明细 / 研究档案 / 原文全文 / 决策记录链)
研究档案
结论摘要
事实层:一条经Jukan转述的韩媒/设备业高管口径称长鑫HBM3 8-Hi综合良率约25%(前道约30%×后道约70%),根因是TSV积累不足,样品小量供平头哥与寒武纪。影响链层:HBM对长鑫仍是期权而非收入,其当期盈利由标准DRAM超级周期驱动(LPDDR6量产、Q2份额约10%、拒绝苹果压价),故该口径不触及未来1–10个交易日的盈利/现金流,未通过风险库三重测试;对SK海力士/三星是既有共识的再确认(内部KB中SemiAnalysis已量化长鑫成本劣势与ASP驱动),属超大市值软催化。价格层:688825.SH在推文后首个交易日高开低走-2.6%(58.90开→56.87收)、随后两日走平,000660.KS同日+3.51%,反应已发生并消退;标的距52周高仅-8.2%、20日仍+2.8%,无剩余空间。证据层:本轮Web检索通道明显降级(多次查询返回无关结果),无法完成first-print溯源、公司回应核查与历史先例面板构建,先例按证据薄处理。结论:无可执行正期望,全腿AVOID。
消息传播链
- 2026-09-09首发韩国媒体/半导体设备业高管口径(未追溯)非一手正文引用『9日受访的半导体设备公司高层』与Nomad Semi分析,指向一份2026-09-09(KST)的韩国产业报道;本轮4次Web检索(中韩英多组关键词)均返回无关结果,未能定位首发媒体与原文链接,故首发节点无法坐实,只能标记为未核验
- 知名半导体分析师账号发布英文全文转述,标题化处理为『Three out of four are defective』;该账号 historically 能引发股价波动,扩散力强,但仍为二手转述而非首发,且发布于A股9月9日开盘前约6.5小时
- 2026-09-09市场反应internal_market_api / yfinance 日线非一手首个可反应交易日:长鑫科技688825.SH开盘58.90(前收58.39,小幅高开)、盘中高59.08、低56.47、收56.87,-2.6%的高开低走;SK海力士000660.KS同日+3.51%(收1,856,000)。两腿同向反应说明消息当日即被市场消化市场反应:688825.SH -2.6%(高开低走);000660.KS +3.51%
- 2026-09-10市场反应internal_market_api / yfinance 日线非一手反应消退:688825.SH +0.58%(收57.20),000660.KS -0.16%(收1,853,000),无二次下跌,说明该口径未被市场当作持续性利空市场反应:688825.SH +0.58%;000660.KS -0.16%
- 2026-09-11市场反应internal_market_api / yfinance 日线非一手688825.SH -0.82%(收56.73,开56.00/高57.00/低54.80),000660.KS -2.21%(收1,812,000);三个交易日累计688825.SH约-2.2%,距52周高61.80仅-8.2%,20日仍+2.8%,事件冲击已被吸收市场反应:688825.SH -0.82%;000660.KS -2.21%
关键理由
- 该口径为匿名单一信源经社媒转述,无公司披露或第二独立源,且文章自身用三星HBM4良率半年内<60%→80%作反例,结论被明确对冲
- HBM3对长鑫科技仍处试产/送样阶段、无收入贡献,公司当期盈利由标准DRAM超级周期驱动(LPDDR6量产、Q2全球份额约10%),本事件不改变未来1–10日盈利基线
- 定价已完成:推文后首个A股交易日688825.SH高开低走-2.6%、随后两日走平,SK海力士同日+3.51%,5天后无任何剩余edge
主要风险
- 若后续出现公司层面确认、更多独立信源或HBM客户验证/订单延后的硬证据,负面读数可能升级为对『国产HBM溢价』的实质下修,届时需重估
- 本轮Web检索通道降级,未能追溯韩国首发原文与公司回应,first-print与传播范围判断存在残余不确定性
反向证据
- 做空前查反向 · 来源 kb_search · 日期 2026-06-25 · 影响 weakens
内部首席观点引SemiAnalysis:长鑫1Q26营收73亿美元、同比+700%、毛利率由-4.7%升至37.8%,但比特出货环比仅+11%、ASP环比+57%,结论是『长鑫难掀DRAM桌子,关注有HBM护城河企业』——即『中国HBM落后』早已是市场共识,本条口径缺乏surprise
检索:长鑫存储 CXMT HBM3 良率 TSV 国产HBM 进展(kb_search) 原文 ↗ - 做空前查反向 · 来源 web · 日期 2026-09-08 · 影响 weakens
自家案例库显示同一标的近两周被多条正面消息刷屏:LPDDR6量产并首发搭载小米18 Fold(8/29宣布、9/1与9/7扩散)、Counterpoint称Q2全球DRAM份额达10%、拒绝苹果压价要求不低于三星/SK海力士、2027年前产能基本售罄——这些正面事实与HBM负面口径并存,净方向不支持做空
检索:长鑫科技 近期利好 LPDDR6 量产 份额 苹果(search_case_library) 原文 ↗ - 做空前查反向 · 来源 query_modeling_kb · 日期 2026-03-12 · 影响 weakens
SemiAnalysis口径HBM缺口2025约5%→2026约6%→2027约9%,美光称仅能满足关键客户50%–67%需求,SK海力士称未来三年需求远超供给;行业处于供不应求超级周期,做空存储链条的宏观逆风明显
检索:HBM 供需缺口 与 存储超级周期(query_modeling_kb) - 做空前查反向 · 来源 filing · 日期 2026-09-11 · 影响 invalidates
9月9日高开58.90→收56.87(-2.6%)、当日低56.47,抛压一次性释放;9月10日+0.58%、9月11日-0.82%,未延续下跌,20日累计仍+2.8%、距52周高仅-8.2%、量比0.72(缩量)。事件后回落不构成做空edge,也不构成做多催化
检索:688825.SH 事件后真实日线形态(stock_quote full) - 做多前查反向 · 来源 filing · 日期 2026-09-11 · 影响 invalidates
读通已在9月9日兑现(+3.51%),20日累计+13.77%,9月11日回吐-2.21%且当日开1,773,000低于前收、高开后回落,显示短线拥挤与获利了结,做多读通腿无剩余空间
检索:SK海力士 000660.KS 读通反应与拥挤度(stock_quote)
先例如何影响了这次决策
对本次决策的影响:先例缺失未改变结论:可观测的真实反应已足以判定price-in完成,且形态显示追空edge为负,最终全腿AVOID、不设tp/sl
方向:无可用外部先例面板;但本事件的价格反应可直接观测——直接主体在推文后首个交易日高开低走-2.6%后走平,竞争对手读通主体同日+3.51%,两者均在3个交易日内消化完毕,方向与幅度无需借助先例推断
涨幅量级:未用先例校准幅度;改用真实日线:688825.SH 9/9 -2.6%、9/10 +0.58%、9/11 -0.82%(累计约-2.2%),000660.KS 9/9 +3.51%、9/10 -0.16%、9/11 -2.21%
形态:688825.SH 9月9日为典型高开低走(开58.90高于前收58.39、收56.87接近当日低点56.47),属见光死/派发形态;此后两日未延续下跌,说明抛压在当日一次性释放,追空胜率差
工具执行结果 · 成功 14 · 降级 1 · 失败 0 · 取消 0
- search_processed_events 成功 stock_tools HTTP · 20 ms
{"result": "search_processed_events:agent query=`changxin memory hbm3 yield tsv defective | cxmt hbm 样品 阿里 平头哥 寒武纪 | cxmt hbm3 良率 25% tsv | sk海力士 hbm 竞争 长鑫 落后 | 长鑫存储 hbm 良率 不良`;结构化索引覆盖完整 72 小时;旧记录兼容扫描近 0.0/72 小时、0 条本臂决策(完整),返回 0 条候选。未命中只表示这些 agent query 表达式没有召回候选,不能据此证明消息首发或不是重复;按当前事件继续正常研判。"} - fetch_decision_experience 成功 stock_tools HTTP · 10 ms
{ - web_search 成功 Bing RSS · 249 ms
Bing RSS 搜索结果(原始查询:CXMT 长鑫存储 HBM3 良率 TSV 세 개 중 네 개가 불량) - web_search 成功 Bing RSS · 257 ms
Bing RSS 搜索结果(原始查询:ChangXin Memory CXMT HBM3 yield 25% TSV September 2026 report) - query_modeling_kb 成功 stock_tools HTTP · 62 ms
{"result": "modeling_kb 命中 12 条 (库快照 2026-07-27 · AI 算力链 19 页 · 本次命中证据最新 2026-06-12 · 其中 ⟨预测⟩ 9 条) —— 这是**定期重生成的快照**,不是活数据;比快照更晚的事实以你手上的信号/kb_search 为准。\n\n• [A] HBM 反推上限(定性) · T6 · 并行约束 — CoWoS / HBM / N3 前道,每年谁 binding · CY2027E = HBM 每比特耗晶圆 ~3x DRAM(HBM4)→~4x(HBM4E);Rubin HBM 容量较 Blackwell +50% ( - kb_search 成功 stock_tools HTTP · 2,234 ms
{ - normalize_ticker 成功 stock_tools HTTP · 20 ms
未归一到唯一代码;请用 stock_quote 的 candidates 消歧,或直接给代码。 - stock_quote 成功 stock_tools HTTP · 74 ms
{ - web_search 成功 Bing RSS · 640 ms
Bing RSS 搜索结果(原始查询:CXMT HBM3 수율 TSV 불량 조선비즈 2026년 9월 창신메모리) - stock_quote 降级 stock_tools HTTP · 850 ms
{ - web_search 成功 Bing RSS · 264 ms
Bing RSS 搜索结果(原始查询:长鑫存储 HBM3 良率 25% TSV 报道 寒武纪 平头哥 样品) - search_case_library 成功 stock_tools HTTP · 203 ms
{"result": "search_case_library「688825.SH」—— 与这些标的相关的自家决策案例(本页 12 条,该标的全历史相关记录共 314 条;ENTER 只按该标的自己的腿统计):\n◆ 688825.SH:相关历史决策 314 条 · 其中该标的 ENTER 0 条(0.0%) · 我们从没在这只票上 ENTER 过\n【对所查标的真 ENTER 过的:0 条 —— 相关案例里没有该标的自己的 ENTER 腿】\n【最近 12 条其它决策(多为「不做」,用来看这票最近在被什么消息刷屏)】\n• 2026-09-08 UTC 不做 · 事件:长鑫存储官宣 LPDD - web_search 成功 Bing RSS · 309 ms
Bing RSS 搜索结果(原始查询:"CXMT" HBM yield "three out of four" defective TSV Korea media The Elec Chosun) - web_search 成功 Bing RSS · 266 ms
Bing RSS 搜索结果(原始查询:CXMT HBM yield problems report SK hynix Samsung shares reaction memory stocks) - web_search 成功 Bing RSS · 262 ms
Bing RSS 搜索结果(原始查询:长鑫科技 688825 HBM 良率 传闻 公司回应 澄清 9月)
原文全文
展开原文 · 4,929 字
ChangXin Memory Technologies (CXMT), China's largest DRAM maker, has begun trial production of fourth generation high bandwidth memory (HBM3), but initial yields are barely improving. Yields are reported to have stalled at 25%, roughly one third of the so called "golden yield" of 80% that the semiconductor industry treats as the threshold for volume production. The gap is stark compared with SK hynix, which has been mass producing the same product since 2022 and has secured yields above 90%. Industry sources point to the gap in maturity of through-silicon via (TSV) technology, the core process for stacking and connecting multiple DRAM layers, as the root cause.
◇ "DRAM has caught up, but HBM stacking is a different problem"
According to a senior official at a semiconductor equipment company familiar with CXMT's situation on the 9th, the yield of CXMT's HBM3 8-High product is stuck at around 30% in the front end process. Of the products that survive that stage, only about 70% are recognized as final good units after passing through the back end process. In simple terms, if 100 HBM3 units are started, close to 80 of them fail the final test.
CXMT is reported to be supplying the small volumes of HBM samples it produces this way to Chinese companies such as Alibaba's T-Head and Cambricon while continuing its yield improvement work. The problem is that the issue does not lie in the fine process technology of the DRAM itself. A semiconductor equipment industry official explained, "There is no major problem with the standard DRAM that CXMT makes on its 'G4' (17nm class) process used for HBM. However, the DRAM dies used for HBM are larger in area than standard products and have more demanding electrical specifications, so even on the same process they are much harder to pass the acceptance criteria."
In other words, CXMT's fundamental capability in making standard DRAM has risen to a considerable level, but a bottleneck is emerging at the stage of converting it into the high performance product that is HBM. At the heart of that bottleneck, according to industry sources, is the TSV process.
◇ The real hurdle is TSV... "Impossible to catch up without years of accumulated know how"
TSV stands for "Through Silicon Via" and refers to the microscopic copper wiring that passes vertically through each layer to carry electrical signals when DRAM is stacked in multiple layers, as in HBM. It is a highly demanding process in which a DRAM wafer is thinned down to several tens of micrometers, a fraction of the thickness of a human hair, after which thousands of tiny holes are drilled through that thin silicon plate and filled completely with copper. A single hole that is misaligned or not properly filled can cause the entire layer to be rejected, making it one of the semiconductor processes with the most stringent precision requirements.
The consensus in the industry is that this is the process where the technology gap between CXMT and the leading companies is widest. According to analysis by semiconductor research firm Nomad Semi, Samsung Electronics' HBM2 (second generation HBM) has more than 5,000 TSVs per die and SK hynix's HBM3 has more than 8,000, while CXMT's is understood to have only around 3,000. A smaller number of TSVs means sacrificing bandwidth (data processing speed) in exchange for lower process difficulty, yet even so CXMT's yields still fall far short of Samsung and SK hynix. TSV is a process that is challenging even for the industry leader: SK hynix itself publicly disclosed in 2024 that the yield of the standalone TSV process was only 40 to 60% at the time.
On top of this, yield losses also occur in the back end (stacking and bonding) stage where the dies are actually stacked and joined. If even one of the eight dies is misaligned, if a microscopic void forms at a bonding interface, or if a layer warps during the thermocompression bonding process (warpage), the entire stack is scrapped. Because the number of possible failure points grows with each additional layer, the difficulty rises exponentially. Given that CXMT is already showing such poor yields at 8-High, some expect it to face even greater difficulties when moving to higher stacks such as 12-High.
A semiconductor industry official explained, "The TSV process is an area that only stabilizes after years of accumulated wafer handling know how. Chinese companies have rapidly closed the gap in the fine process technology of DRAM itself, but back end know how such as TSV and bonding is difficult to catch up on in a short period." He added, "That said, Samsung Electronics also had initial HBM4 (sixth generation HBM) production yields below 60% in February this year and raised them to 80% within six months, so it is too early to declare CXMT's 25% yield a 'failure.'"
决策记录链 · 只追加
每次领取产生一条记录,重新决策追加一条并以 supersedes 指向被替代的记录(ADR 0014)。
这条决策早于决策账本(2026-10-08 起逐条追加),没有追加记录。