Sep 13, 2026 · Weekend: Monday's SHFE trading day is already 38% traded after opening at 23,920. FedWatch resolves at 85.6%, not 90%.

一 · 行情表现

前置声明:本期性质、写稿时点与最重要的一条口径推论

本期写于 9月13日(周日)上午,沪伦两市均休市,且自上期以来两市均无任何新的行情数据。
⚑⚑ **本期新增的不是价格,而是口径、日期与日历:有明确日期的 9月11日 CME FedWatch 读数 85.6%(取代沿用的约 90%);自 8月11日起的完整月度定价链条;沿用六期的 EGA「25% 对 50%」标记的解决——且解决方式是认定其为本简报自身的口径错误;四家 2026 年全球平衡预测首次并列,离散度 12.14 倍;四家价格预测并列,离散度达现价的 44.29%,其中仅两家有明确日期;沿用 ING 条目的正确日期(2025年12月8日);一条流传中的中国 8月电解铝产量口径因算术上不可能而予以拒绝;以及周一与全周日历,其中一场落在周一上期所交易时段之内的能源事件。**
⊘ **无法取得:任何新的 Westmetall 行(9月11日仍为最后一行,本期已重新查询确认);9月11日夜盘之后的任何上期所数据;SMM 自身一手现货报价,连续第十三期;SMM 三地日度库存,连续第七期;上期所周度仓单一手口径;10月联邦基金利率条,连续第十六期;LME 成交量与持仓量。**

⚑⚑ 本页最重要的一条事实:周一的交易日已经开盘,而且已经成交了 38.3%

⚑⚑ **上期所 9月14日交易日并非从周一早晨开始,而是从周五 21:00 开始。这一点上期已写入口径说明,本期把它的后果算出来。**
⚑⚑ **因此周一交易日的以下价格已经印出,不是预测:开盘 **24,100、最高 24,100、最低 23,920、最新 24,000;AL2610 成交量 81,993 手,持仓 198,952(−12,438)。**
⚑⚑ **81,993 手相当于周五整个交易日 213,873 手的 38.3%。周一的日盘不是行情的开始,而是一场已过三分之一的行情的延续(derivation)。**
⊘⊘ **而周一唯一存在的价格 24,000 已经在 MA20 之下。任何以「周一从 24,235 起跳」为前提的推演,起点就是错的(interpretation)。**

⊘⊘ 六条均线中已有四条在周一唯一的价格上失守,而最后一道支撑已被逼近至 72.75 点

⚑⚑ **六条沪铝均线沿用且精确(基于 224 行序列,9月11日收盘基准):MA5 24,433.00 / MA10 24,311.00 / MA20 24,055.00 / MA30 24,029.83 / MA40 23,847.25 / MA60 23,601.25。**
⊘⊘ **对照周一已印出的 24,000:低于 MA5 433.00、低于 MA10 311.00、低于 MA20 55.00、低于 MA30 29.83;高于 MA40 152.75、高于 MA60 398.75。六条已失守四条。**
⊘⊘ **对照周一已印出的最低价 23,920:低于 MA20 135.00、低于 MA30 109.83,仅高于 MA40 72.75。周五夜盘已经把最后一道仍然完好的均线逼近到 72.75 点以内(derivation)。**
⊘ 23,920 **恰为 8月28日收盘价,亦为 9月1日以来的最低价位。**

⊘⊘ 事先写下的周一检验是「活的」,而它正落在自己的无结论区里

⚑⚑ **事先写下的检验原文:「9月14日日盘收在 24,055.00 之上,则夜盘破位为超跌;收在 23,920 之下,则本轮回调升级为趋势转折,下一参照为 MA40 23,847.25;收在 23,920–24,055 之间记为无结论区,不作任何定性」。**
⊘⊘ **周一唯一存在的价格 24,000 正落在该区间内:高于下沿 80.00,低于上沿 55.00,位于 135.00 点宽区间的 59.3% 处。**
⊘ **本简报不因此提前定性。检验写在数字之前,就必须等数字(derivation)。**

⚑⊘ 而持仓的两个口径互相矛盾,且要到周一日盘才能裁决

⊘⊘ **9月11日日盘是离场:全系持仓 −29,349 至 593,718,为本简报自设作废阈值的 1.96 倍;三期口径下离场以 34,381 对 12,624 压倒移仓;发生在本台账最重的成交量上(全系 414,471,+55.4%)。**
⚑ **9月11日夜盘(即周一交易日的第一段)却是移仓:六合约合计仅 −1,612,AL2610 的 −12,438 被 AL2611 +5,421、AL2612 +3,601、AL2701 +1,851、AL2702 +258 吸收;且 AL2611 持仓以 213,268 对 198,952 反超 AL2610。**
⊘ **两者性质相反,本期不予调和。周一日盘将裁决哪一个读数成立,这也是本期第二项关注清单的全部内容(interpretation)。**

⚑ 伦敦:沿用不变,而「沿用」本期是经过重新查询验证的,不是假设的

⚑ **本期重新查询 Westmetall,9月11日仍为最后一行——沿用经验证而非假设。**
⊘ 现货结算 **3,274.00 美元(−69.00,−2.06%),为本台账最大单日现货跌幅;三个月 3,250.00 美元(−68.00);backwardation +24.00 美元,五期以来首次收窄且仅收窄 1.00 美元;库存 244,100 吨(−250),低于 −500 的实质门槛,正确归类为噪音。**
⚑ **六条 LME 均线沿用且连续第五期精确:MA5 3,320.80 / MA10 3,293.80 / MA20 3,256.05 / MA30 3,263.48 / MA40 3,244.04 / MA60 3,217.94。**
⊘ **现货已失守 MA5(−46.80)与 MA10(−19.80);高于 MA20 仅 17.95、高于 MA30 仅 10.52——为本轮最窄的支撑余量。**
⚑ **平行口径,列示且不予调和:TradingEconomics 9月11日报 3,256.65 美元,日 −0.70%、月 −1.52%、年 +20.62%;与 Westmetall 现货相差 17.35 美元、与三个月相差 6.65 美元,符合不同合约参照而非数据冲突(derivation)。**

二 · 铝价及相关市场

⊘ 现货沿用至周五——而本期必须明确说明它对周一而言已经是陈旧报价

⚑ **SMM 镜像口径(9月11日):A00 24,240(−320);A00 佛山 24,430(−350);ADC12 24,300(−150)。**
⚑ **CNAL 五口径(9月11日):上海 24,230(−310);长江 24,240(−320);中原 24,240(−320);南海/佛山 24,710(−340);广东南储 24,440(−360)。**
⊘ **区域价差全线收窄:南海 − 上海 480(前 530);南储 − 上海 210(前 260);长江 − 上海 +10(不变)。华南跌幅 340–360 大于华东的 310–320。**
⚑ **9月11日收盘基差 +5(对收盘)、+50(对结算)。**
⊘⊘ **但对周一已经印出的 24,000,基差是 +240。+240 的基差不是现货偏紧的证据,而是一条等待周一早晨重新定价的陈旧报价。本简报明确记录:基差在周一 SMM 发布之前是「未知」,而非 +5(interpretation)。**
⊘ **SMM 一手报价、区域成交升贴水、出货与采购意愿指数、氧化铝指数与中铝五地挂牌连续第十三期登录墙,故本期仍不发布任何情绪指数,也不以代理指标替代。**

⊘⊘ 氧化铝:沿用于本台账最差的一张收盘表,而「活的」数字是夜盘那一个

⊘⊘ **AO2701 9月11日日盘:开 2,762,高 2,792,低 2,691,收 2,697,结算 2,731;−39(−1.41%)结算口径、−61(−2.21%)收盘口径。成交 275,018(+69.0%),持仓 220,066(+24,675,+12.63%)。全系持仓 462,497(+20,900)。**
⊘⊘ **AO2701 周五夜盘(即周一交易日的第一段):2,673(−58,−2.12%),成交 92,893,持仓 235,026(再 +14,960)。AO2610 2,651(−44)、AO2611 2,659(−45)、AO2612 2,672(−39)、AO2702 2,690(−55)。五个合约全部下跌,而持仓仍在增加。**
⚑⚑ **六条氧化铝均线沿用且精确(160 行 AO2701 序列):MA5 2,757.00 / MA10 2,753.10 / MA20 2,737.15 / MA30 2,738.67 / MA40 2,749.22 / MA60 2,781.45。**
⊘⊘ **周一已印出的 2,673 低于全部六条,依次低 84.00 / 80.10 / 64.15 / 65.67 / 76.22 / 108.45——每一条的幅度都比 9月11日收盘时(60.00 / 56.10 / 40.15 / 41.67 / 52.22 / 84.45)更宽(derivation)。**
⊘⊘ **成本口径一(2,718)已连续两日失守,幅度 51 点;余量链条 +39 → +19 → +2 → −3 → −51,结算价 2,695 亦低该线 23 点。**
⊘⊘ **成本口径二(2,826.21)在整条曲线上无处定价,最高收盘为 AO2708 的 2,815,仍低 11.21。**
⊘ **首尾 contango +148(占前端 5.55%),第三次收窄;无日期的中国氧化铝现货口径 2,707 仍高于 AO2701 收盘价 10 点,盘面对其自身现货参照仍为贴水。**

⊘ 冶炼利润代理:夜盘再压缩 188.68

⊘ **主力配对(AL2610/AO2701,1.93 吨/吨,假设先行声明):周一夜盘价格下 24,000 − 1.93 × 2,673 = 18,841.11,对照 9月11日收盘的 19,029.79,再 −188.68(derivation)。**
⊘ **配对口径(AL2610/AO2610):24,000 − 1.93 × 2,651 = 18,883.57。**
⊘ **两个口径同向:夜盘的下跌由冶炼利润吸收,而非由氧化铝端吸收(interpretation)。**

三 · 库存

⊘ 全部沿用,而周一的数字将承担两周的信息量

⊘ **SMM 全国铝锭社会库存沿用 796,000 吨(9月10日)。链条:8/24 860,000 → 8/27 852,000 → 8/31 837,000 → 9/3 815,000 → 9/7 802,000 → 9/10 796,000。**
⊘ **沿用判定不变:周四对周四 −19,000,为此前 36,000 速度的 52.8%、8月自身约 23,935 的 79.4%;周一至周四仅 −6,000。去库减半而非停止;自 7月初约 113 万吨高点累计 −334,000 吨(−29.6%)。**
⚑ **Mysteel 沿用:8/31 809,000 → 9/3 789,000 → 9/7 774,000 → 9/10 768,000;SMM 减 Mysteel 为 28,000 / 26,000 / 28,000 / 28,000,恒定定义性偏移结论维持。**
⚑ **LME 244,100 吨(9月11日,−250),本期重新查询确认未变。**

⊘ 连续第七期没有区域分解——而周一的全国数字也回答不了这个问题

⊘ **SMM 三地日度库存连续第七期不可得,沿用至 9月3日:641,000 吨(广东 143,000、无锡 300,000、巩义 198,000)。**
⊘ **上期所周度仓单一手口径再次不可得;第三方称上期所铝库存周环比约 −3%,列示且不赋权。**
⊘ **必须明确写下的一句:周一的全国数字会告诉我们减速是否继续,但不会告诉我们它发生在哪里。区分「华东到货集中」与「真实需求停滞」,仍然需要那份连续第七期缺失的三地数据(interpretation)。**

解读

⚑ **伦敦仓单十七个区间中十个为零,而 backwardation 在现货跌 69 美元时只让了 1 美元——两者互相印证:周五抛售铝的不是实货近端市场(interpretation)。**
⊘ **但国内一侧本期完全是空白的,而在一个盘面已经在 MA20 之下开盘的周一,没有库存数据本身就是一种暴露(interpretation)。**

四 · 宏观因素

⚑⚑ FedWatch 口径落地:85.6%,而非上期公布的约 90%

⚑⚑ **有明确日期的 9月11日 CME FedWatch 读数为 85.6%,对应 9月16日加息 25 个基点。上期公布的约 90% 予以修正。**
⚑⚑ **这项修正一半对本简报不利、一半有利:重定价是真实的,但幅度比公布值小 4.4 个百分点;而事先写下的升级阈值 95% 因此距离 9.4 个百分点而非 5 个百分点——该阈值比上期所暗示的要难触发得多(derivation)。**
⚑ **完整月度链条首次可还原:48.4%(8月11日)→ 约 66%(8月31日)→ 约 69.3%(CPI 前)→ 85.6%(CPI 后);31 天 +37.2 个百分点,其中 +16.3 发生在一个下午。**
⚑ **本周沿用的 56% / 59% / 60.6% 三口径「冲突」就此撤销——它们从来不是冲突,而是同一条链条上的中段快照,予以退役(derivation)。**
⊘ **10月联邦基金利率条连续第十六期缺失,联储路径期限结构仍不可得。**

⊘⊘ 但利率这一条腿本身是本页最差的:十年期站上 4.97%

⊘⊘ **美国十年期国债收益率约 4.97%–4.98%(9月11日),为 2023年10月以来最高;触发点是财政部扩容后的回购操作实际购买量低于预期。**
⊘ **美国 8月 CPI:整体 +0.4% 环比、+3.4% 同比;核心 +0.3% 环比、+2.4% 同比;汽油 +3.9%。**
⊘ **EY-Parthenon 由「按兵不动」改判「加息 25 个基点」,目标区间将升至 3.75%–4.00%。**
⚑ **美元几乎没有跟上:DXY 99.095(+0.047,+0.05%),月度 −0.92%。在联储路径大幅重定价的同一天,美元只动了 5 个基点——本简报关于「日元/日本央行通道仍占主导」的判断继续成立(interpretation)。**

⊘⚑ 能源:两个 Brent 口径,差 0.18%,结论相同

⊘ **结算口径:Brent 104.61 美元,日 −2.8%,周 +8.7%。**
⊘ **TradingEconomics 口径:104.42 美元,日 −2.98%,周约 +9%。**
⚑ **两者相差 0.19 美元(0.18%),并列且不予偏取;结论两个口径完全一致:连续第三日收在 100 美元之上,抬高冶炼成本底部(偏多)同时使 9月16日的决定更难(偏空)(interpretation)。**
⊘⊘ **而周一有一场直接作用于这条腿的事件:海湾合作委员会与伊朗的外交官将于周一在阿曼会面,讨论霍尔木兹海峡通航的临时安排。该会面落在上期所周一交易时段之内(见第九节)。**

⊘ 中国:周二的两个需求数字在公布之前就已经是弱的

⊘ **9月15日(周二)公布 8月规模以上工业增加值、能源生产、固定资产投资、商品房销售、社会消费品零售总额与 70城房价。**
⊘ **市场预期:社零同比约 +0.6%,与 7月持平;固定资产投资累计同比预计进一步走弱至约 −7%;基建投资被描述为「弱」,房地产投资被描述为「低迷」。**
⊘ **对铝而言这三条直接对应建筑、电网与家电渠道,预期集合在数字落地之前就是偏空的(interpretation)。**
⚑ **唯一的对冲项是出口,且本期已完整注明日期:8月未锻轧铝及铝材出口 626,000 吨,同比 +17.2%(2025年8月为 534,000 吨),但环比 −2.6%(7月为 643,000 吨);1–8月累计 467 万吨,同比 +16.7%——8月比八个月均值 583,750 吨/月高出 7.24%(源文日期:9月8日)(derivation)。**

⊘ 日本:9月17–18日日本央行会议仍是本简报认定的主导通道

⊘ 日本央行 **9月17–18日议息,加息定价持续上升;审议委员田村直树(Takata)称应「灵活地」加息,并暗示节奏可能快于或幅度大于以往的半年一次;行长植田和男保留加息空间。**
⚑ 9月3日日元一度升至约 **155.28,单日涨逾 2%,为一个月盘中高位。**
⊘ **对铝的传导是美元指数而非直接渠道,但在联储路径几乎不推动美元的当下,它是解释美元为何不动的那个变量(interpretation)。**

五 · 基本面

⚑⚑ 本期解决一项沿用六期的标记——而错误是本简报自己的

⚑⚑ **EGA「25% 对 50%」冲突不存在:50% 是 Al Taweelah 氧化铝厂的爬产目标,2026年7月10日宣布复产时称「数日内」可达,并预计 2026年底恢复全部产能;18% 是 Al Taweelah 电解铝厂的电解槽复产进度(8月11日,1,262 槽中的 227 槽)。两项是不同的资产。**
⚑⚑ **标记退役,错误如实记录为本简报自身的口径混淆(derivation)。**
⊘ **而这项条目的实质是偏空的、且落在氧化铝一侧:Al Taweelah 氧化铝厂 2025年产量 240 万吨,接近 EGA 自身氧化铝需求的一半;一座这个量级的氧化铝厂在 2026年底之前由零走向满产,与本页氧化铝栏的其余所有内容站在同一侧(interpretation)。**

⊘⊘ 2026 年全球平衡不是一个数字,而是一个 12.14 倍的区间

⊘⊘ **四家预测首次并列,且刻意不予调和:AL Circle 约 14 万吨短缺(2026年,年内无精确日期);ING 约 20 万吨(若 Mozal 关停则扩至 60 万吨),日期为 2025年12月8日;Wood Mackenzie 约 90 万吨,日期为 2026年8月4日;摩根大通 170 万吨,日期为 2026年8月21日。**
⊘⊘ **最高比最低 = 12.14 倍;摩根大通比 ING = 8.5 倍;Wood Mackenzie 比 AL Circle = 6.43 倍(derivation)。**
⊘ **必须写下的一句:任何不注明来源就引用「2026 年短缺 X 万吨」的说法,引用的是一个十二倍区间中的一点(interpretation)。**

⊘⊘ 价格预测的离散度达现价的 44.29%,而四家中仅两家有日期

⊘ **基准:9月11日 LME 现货结算 3,274.00 美元。**
⚑ **摩根大通(2026年8月21日):Q3 2026 3,800(+16.07%)、Q4 2026 3,700(+13.01%)、2026 全年均值 3,587(+9.56%);Q1 2027 3,450、Q2 3,200、Q3 2,900、Q4 2,750(−16.00%)、2027 全年均值 3,075(−6.08%)。**
⚑ **TradingEconomics 模型:Q3 3,267.98(−0.18%)、十二个月 3,489.17(+6.57%)。**
⊘ **ING(2025年12月8日):2026 全年均值 2,900(−11.42%)。**
⊘ **高盛:2,350(−28.22%),无日期,列示且不赋权。**
⊘⊘ **首尾相差 1,450 美元 = 现价的 44.29%(derivation)。**
⊘ **而最偏多的那条同时是一条「见顶」路径:摩根大通的 3,800 是其整条八个季度路径的最高点,Q4 2027 的 2,750 较之低 27.63%,2027 全年较 2026 全年低 14.27%。本页最偏多的、有日期的预测,说的是周期高点就在正在结束的这个季度(interpretation)。**
⚑ **其论据落在本简报自己的供应台账上:中东冶炼厂停产抹去全球供应 6% 以上、霍尔木兹海峡运输受阻、中国 4,500 万吨产能上限,并明确提示若中国限制铝材出口则存在显著上行风险。**

⊘ 一条流传中的中国产量口径予以拒绝,而非列示

⊘ **某一口径给出中国 8月电解铝产量 762 万吨。**
⊘ **拒绝理由(算术):4,500 万吨/年上限对应 375 万吨/月,762 万吨为其 2.03 倍,年化 9,144 万吨,算术上不可能。**
⊘ **且它与 8,800 万吨/年氧化铝口径对应的月度值 733.3 万吨仅相差 3.91%——几乎可以确定是一条被错标为电解铝的氧化铝数据。故予以拒绝,不作为数据发布(derivation)。**

⊘ 氧化铝供应:新增一项产能条目

⊘ **新增:广西防城港 240 万吨/年氧化铝厂,被列为投向本已过剩的国内氧化铝市场的 2026 年新增产能之一。**
⊘ **AL Circle 口径:2026 年冶金级氧化铝需求约 1.452 亿吨,广义市场接近 1.47 亿吨。**
⊘ 沿用:CERA 预计 2026 年全球氧化铝过剩 **179 万吨;中国 1–6月氧化铝进口 227.7 万吨(+749.1%);一个口径称 2026 年国内计划新增氧化铝产能约 860 万吨、全年过剩超 1,000 万吨(无日期,不赋权)。**

⊘ 下游与贸易:沿用,无新数据

⊘ 铝型材产能利用率 **37.05% 周期底部;龙头下游加工企业开工率 59.9%(8月13日);再生铝合金开工率 49.1%。**
⊘ 沿用:墨西哥对来自中国与美国的铝型材征最终反倾销税(9月10日);欧盟拟议废铝出口禁令;加拿大 25%→50% 并点名铝杆。
⚑ 沿用:Mysteel 8月 A00 月均价——华东 **23,869(+3.34%)、华南 24,010(+3.79%)、华中 23,762.4(+3.42%);8月价格在涨而开工率没有跟上。**

六 · 电解铝现货市场

⊘ 本期无新的现货成交信息,以下全部为 9月11日沿用值

⚑ **五地报价(9月11日,CNAL 口径):上海 24,210–24,250,均 24,230(−310);长江 24,220–24,260,均 24,240(−320);中原 24,240(−320);南海/佛山 24,660–24,760,均 24,710(−340);广东南储 24,410–24,470,均 24,440(−360)。**
⚑ **SMM 镜像口径:A00 24,240(−320);A00 佛山 24,430(−350);ADC12 24,300(−150)。**
⊘ **区域结构:南海 − 上海 480;南储 − 上海 210;长江 − 上海 +10。四日走阔之后的一次全线回吐,华南领跌。**

⊘⊘ 出货与采购意愿:本期不给出任何指数,并说明为什么连推断都要降级

⊘ **SMM 出货意愿指数与采购意愿指数连续第十三期不可得,本简报不发布、也不以代理指标替代。**
⊘⊘ **上期作出的定性推断本期必须降级:上期据「A00 单日 −320 而基差仍收敛至 +5」推断「现货跟随而未领跌」。该推断在 9月11日收盘口径下成立,但周一交易日已经把盘面价格带到 24,000,而现货报价停在周五。在 SMM 周一发布之前,持货与接货意愿的任何判断都是对一个已经过期的价差作出的,本简报因此明确不作推断(interpretation)。**
⚑ **唯一仍可成立的一句:五个区域升水全线收窄,是持货方出货意愿上升的证据;而它是否被下游接货意愿抵消,要等周一。**

⚑ 基差与交割月结构

⚑ **9月11日收盘基差 +5(对收盘 24,235)、+50(对结算 24,190)。**
⊘⊘ **对周一已印出的 24,000,基差为 +240——记为陈旧报价而非现货升水;链条 +70 → +55 → −20 → −20/0 → +70 → +5 → 未知(待周一)(derivation)。**
⚑⚑ **主力换月:周五夜盘 AL2611 持仓 213,268 已超 AL2610 的 198,952(多 14,316);日盘 AL2610 仍以 3,543 领先。本简报将在日盘确认后切换主力口径并于当期明确标注。**

七 · 关键价格驱动因素

⚑ 偏多因素(16 项)

⚑⚑ **事先写下的 95% 阈值比上期所暗示的更难触发:实际 FedWatch 为 85.6% 而非约 90%,距离 9.4 个百分点而非 5 个。**
⚑⚑ **伦敦现货跌 69 美元,backwardation 只让了 1.00 美元(+25.00 → +24.00),实货近端未参与抛售。**
⚑⚑ **LME 库存 244,100 吨,十七个区间中十个恰为零,近端实货未见松动;−250 低于 −500 实质门槛。**
⚑⚑ **在联储路径大幅重定价的同一天 DXY 只涨 0.05%,美元渠道传导明显不足。**
⚑⚑ **9月11日点名的 24,075 守得分毫不差,且日间时段买回了夜盘 450 点中的 110 点。**
⚑⚑ **周五夜盘是移仓而非离场:六合约合计仅 −1,612,AL2610 的 −12,438 被 AL2611 等吸收。**
⚑⚑ **沪市跑赢伦敦 1.02 个百分点(比值 7.3257 → 7.4023),宏观冲击在上海只被部分接受。**
⚑⚑ **摩根大通(8月21日)是本台账上首条有日期且高于现价的预测:Q4 2026 3,700(+13.01%)、2026 均值 3,587(+9.56%),并预计 2026 年短缺 170 万吨。**
⚑ **MA40(23,847.25)与 MA60(23,601.25)在周一唯一存在的价格上仍然完好。**
⚑ **9月11日现货跟随而未领跌,A00 −320、基差 +5,且算术回溯证明上期镜像行为真实报价。**
⚑ **Brent 连续第三日站上 100 美元(104.61 / 104.42 两口径),抬高冶炼成本底部。**
⚑ **中国 8月铝材出口 626,000 吨(+17.2%),比八个月均值高 7.24%;1–8月 +16.7%。**
⚑ **中国 8月出口总额 +25%、贸易顺差 1,190.9 亿美元,外需侧未见转弱。**
⚑ 沿用:LME 库存处 36 年低位附近;Wood Mackenzie 2026 年短缺约 **90 万吨(8月4日)。**
⚑ 沿用海外供应扰动:力拓 Yarwun(10月起 −120 万吨/年)、美铝指引下调 20–30 万吨、Qatalum 约 60% 名义产能、Alba 1–3 号线、几内亚出口上限、俄铝 Friguia(8月24日)、7月 GCC 产量同比 **−44%。**
⚑ **摩根大通明确提示的上行风险:若中国政策转向限制铝材出口,价格存在「显著上行」空间。**

⊘ 偏空因素(31 项)

⊘⊘ **周一交易日已经开盘,且已在 MA20 之下:开 24,100、高 24,100、低 23,920、最新 24,000。**
⊘⊘ **周一唯一存在的价格已使六条均线失守四条(低于 MA5 433.00、MA10 311.00、MA20 55.00、MA30 29.83)。**
⊘⊘ **周五夜盘低点 23,920 已把最后一道完好均线 MA40(23,847.25)逼近至 72.75 点以内。**
⊘⊘ **周一已成交 81,993 手,为周五全交易日 213,873 手的 38.3%——破位发生在流动性充足的时段,不是薄成交的夜盘噪音。**
⊘⊘ **9月11日日盘全系持仓 −29,349,为自设作废阈值的 1.96 倍,「移仓」定性已作废。**
⊘⊘ **三期口径下离场以 34,381 对 12,624 压倒移仓,且发生在本台账最重的成交量上(+55.4%)。**
⊘⊘ **FedWatch 9月16日加息定价 85.6%,月内 +37.2 个百分点(8月11日 48.4%)。**
⊘⊘ **美国十年期约 4.97%–4.98%,为 2023年10月以来最高,且触发点是财政部回购操作低于预期。**
⊘⊘ **LME 现货 −69.00 美元(−2.06%)为本台账最大单日跌幅,失守 MA5 与 MA10,MA20 仅余 17.95、MA30 仅余 10.52。**
⊘⊘ **AO2701 周一已印出的 2,673 低于全部六条均线,且每一条的幅度都比收盘时更宽。**
⊘⊘ **氧化铝持仓在下跌中继续上升:日盘 AO2701 +24,675、全系 +20,900,夜盘再 +14,960——新空入场。**
⊘⊘ **成本口径一(2,718)连续两日失守且扩大至 51 点,结算价亦低 23 点。**
⊘⊘ **成本口径二(2,826.21)在十二个合约上全部未被定价。**
⊘⊘ **2026 年全球平衡的四口径离散度达 12.14 倍,「短缺」这一论据本身不具备可引用的精确度。**
⊘⊘ **价格预测离散度达现价的 44.29%,且四家中两家无日期或已陈旧。**
⊘⊘ **最偏多的有日期预测同时是见顶路径:摩根大通 Q3 2026 的 3,800 为其八个季度路径最高点,2027 全年低于 2026 全年 14.27%。**
⊘⊘ **沿用的 ING 条目实为 2025年12月8日,已陈旧九个月,其 2,900 美元的 2026 均值较周五现货低 11.42%;该条目移出偏多清单。**
⊘⊘ **EGA Al Taweelah 氧化铝厂(2025年产 240 万吨,约占 EGA 自身需求一半)预计 2026年底恢复满产——氧化铝供应端的又一项增量。**
⊘ **广西防城港 240 万吨/年氧化铝新产能投向已过剩的国内市场。**
⊘ **周二中国 8月社零预期约 +0.6%(与 7月持平)、固投累计预期走弱至约 −7%,基建「弱」、地产「低迷」。**
⊘ **中国 8月铝材出口环比 −2.6%(626,000 对 7月 643,000)。**
⊘ **美国 8月核心 CPI +0.3% 环比(前值 +0.2%),整体 +3.4% 同比。**
⊘ **EY-Parthenon 改判加息 25 个基点,目标区间将至 3.75%–4.00%。**
⊘ **五个区域现货升水全线收窄,华南领跌(−340 至 −360);南海 − 上海 480(前 530)。**
⊘ **主力配对利润代理在夜盘再压缩 188.68 至 18,841.11。**
⊘ **首尾 contango 第三次收窄至 +148,市场把过剩定价到远期。**
⊘ **无日期的氧化铝现货口径 2,707 仍高于 AO2701 收盘价,盘面贴水于自身现货参照。**
⊘ **SMM 三地日度库存连续第七期缺失,周一的全国数字也无法回答「减速发生在哪里」。**
⊘ **SMM 一手报价与意愿指数连续第十三期缺失,本简报仍无法发布任何情绪指标。**
⊘ **上期所周度仓单一手口径再次缺失;10月联邦基金利率条连续第十六期缺失。**
⊘ 沿用:铝型材产能利用率 **37.05% 周期底部,龙头下游开工率 59.9%;墨西哥对中美双来源铝型材反倾销;高盛 2,350 美元(无日期,不赋权),较周五现货低 28.22%。**

八 · 一句话观点

本期没有任何市场发生变动,因此本简报能够增加的不是价格而是三件事——一条被上期自己写下却没有算出后果的口径、三项自我修正,以及一次把预测并列起来的算术;而其中第一件就是本页最重要的事实:上期所 9月14日交易日并非从周一早晨开始,它从周五 21:00 就已开始,所以周一「已经开盘,并且已经成交了 38.3%」——其开盘 24,100、最高 24,100、最低 23,920、最新 24,000,AL2610 已成交 81,993 手,对照周五整个交易日的 213,873 手;这意味着任何以「周一从 24,235 起跳」为前提的推演,起点就是错的,而真实的起点 24,000 已经在 MA20(24,055.00)之下 55.00、在 MA30(24,029.83)之下 29.83,六条均线已失守四条,且周五夜盘的 23,920 低点(恰为 8月28日收盘价)已把最后一道完好的 MA40(23,847.25)逼近至 72.75 点以内;更关键的是事先写下的周一检验此刻正落在它自己的无结论区里——24,000 高于下沿 23,920 80.00、低于上沿 24,055.00 55.00,位于 135.00 点宽区间的 59.3% 处,本简报据此拒绝提前定性;三项自我修正依次为:FedWatch 的正确口径是 85.6% 而非上期公布的约 90%(修正一半不利一半有利——重定价真实但比公布值小 4.4 个百分点,而事先写下的 95% 阈值因此距离 9.4 个百分点而非 5 个,比上期所暗示的难触发得多),同时完整链条首次可还原为 48.4%(8/11)→ 约 66%(8/31)→ 约 69.3%(CPI 前)→ 85.6%(CPI 后),31 天 +37.2 个百分点,本周沿用的 56/59/60.6 三口径「冲突」随之作为「同一链条的中段快照」退役;沿用六期的 EGA「25% 对 50%」标记并非数据冲突,而是本简报把两项资产混为一谈——50% 是 Al Taweelah 氧化铝厂的爬产目标(7月10日,数日内,2026年底满产,2025年产 240 万吨、约占 EGA 自身需求一半),18% 是 Al Taweelah 电解铝厂的电解槽复产(8月11日,1,262 槽中的 227 槽),标记退役、错误归属本简报,且其实质落在偏空的氧化铝一侧;沿用作为偏多项的 ING 条目实为 2025年12月8日、已陈旧九个月,其 2,900 美元的 2026 均值较周五现货低 11.42%,移出偏多清单;而把预测并列之后得到的是本期最不舒服的两个数字——2026 年全球平衡不是一个数字而是一个 12.14 倍的区间(AL Circle 约 14 万吨、ING 约 20 万吨、Wood Mackenzie 约 90 万吨、摩根大通 170 万吨),价格预测的首尾相差 1,450 美元、等于周五现货的 44.29%,且最偏多的那条有日期预测(摩根大通 8月21日:Q3 3,800、Q4 3,700、2026 均值 3,587)同时是一条见顶路径,其 3,800 为整条八季度路径的最高点、2027 全年较 2026 全年低 14.27%;此外一条流传中的「中国 8月电解铝产量 762 万吨」因相当于 4,500 万吨上限对应月度值的 2.03 倍、年化 9,144 万吨而予以拒绝(且与 8,800 万吨氧化铝口径的月度值仅差 3.91%,几乎可以确定是错标);其余全部沿用且沿用经过重新查询验证——伦敦 9月11日仍为最后一行(现货 3,274.00、backwardation +24.00、库存 244,100,现货在 MA20 之上仅余 17.95),氧化铝沿用于本台账最差的一张表且周一已印出的 2,673 低于全部六条均线、每一条的幅度都比收盘时更宽、主力配对利润代理再压缩 188.68 至 18,841.11,国内库存沿用 796,000 吨、「去库减半而非停止」的判定不变而三地分解连续第七期缺失,现货沿用至周五并必须加注:人人将引用的 +5 基差是对一个已不再是活价的收盘价算出的,对周一已印出的 24,000 基差是 +240,那不是现货偏紧而是一条等待重新定价的陈旧报价;综合而言,未来 48 小时内有四项事先写下的检验落地、其中一项(周一在阿曼举行的海合会—伊朗霍尔木兹通航临时安排磋商)直接落在上期所周一交易时段之内,而周二中国 8月社零(预期 +0.6%,与 7月持平)与固投(预期走弱至约 −7%)在公布之前就已是偏空的;最诚实的表述是:结构目前仅在收盘口径上完好,而决定它的那场交易已经进行了三分之一以上,本期没有改变任何一个价格(interpretation)。**

九 · 下周一(9月14日)日盘与夜盘关注清单

⚑⚑ 第一,MA20(24,055.00)——而且必须从「盘中已在其下」而非「等待开盘」开始看。 周一交易日已印出开 24,100、高 24,100、低 23,920、最新 24,000,已成交 81,993 手(周五全日的 38.3%)。判别沿用上期原文且不修改:日盘收在 24,055.00 之上,则周五夜盘破位为超跌;收在 23,920 之下,则升级为趋势转折,下一参照为 MA40 23,847.25;收在 23,920–24,055 之间仍记为无结论区。参考位:24,311.00(MA10)/ 24,235(9月11日收盘)/ 24,100(周一开盘兼最高)/ 24,075(已守住的关键位)/ 24,055.00(MA20)/ 24,029.83(MA30)/ 24,000(周一最新)/ 23,920(周一最低)/ 23,847.25(MA40)/ 23,601.25(MA60)。
⚑⚑ 第二,日盘与夜盘的持仓口径互相矛盾,周一日盘裁决哪一个成立。 9月11日日盘是离场(全系 −29,349,为自设阈值的 1.96 倍,三期口径下离场 34,381 对移仓 12,624);夜盘是移仓(六合约仅 −1,612,AL2611 反超 AL2610 达 14,316)。判别带(沿用并收紧):周一全系持仓再降 10,000 以上,则离场为趋势,应下调对本轮全部反弹的权重;回升 5,000 以上,则周五日盘属一次性平仓潮;±5,000 以内记为待定。同时观察 AL2611 日盘持仓是否超过 AL2610(周五日盘 AL2610 仍领先 3,543)——一旦确认,本简报当期切换主力口径并明确标注。
⚑⚑ 第三,周一的 SMM 全国铝锭库存,它将承担两周的信息量。 沿用链条 9/7 802,000 → 9/10 796,000,周四对周四 −19,000(此前速度的 52.8%)。判别带沿用上期重写后的版本,且完整覆盖数轴:≤ 785,000 = 去库重新加速;785,001–792,000 = 维持当前减速节奏;792,001–799,000 = 进一步减速;≥ 799,001 = 停滞或累库。同时观察 Mysteel 同日数字与 SMM 的差值是否仍在 26,000–28,000(若偏离,则「恒定定义性偏移」结论需重新审视),以及三地分解能否在连续第七期缺失后补发——没有它,周一的全国数字告诉我们减速是否继续,但不会告诉我们它发生在哪里。
⚑⚑ 第四,周一在阿曼举行的海合会—伊朗磋商,这是本期唯一落在上期所交易时段之内的外部事件。 议题为霍尔木兹海峡通航的临时安排。判别(本期新写,先于数字):若会后出现明确的临时通航安排且 Brent 跌破 100 美元,则能源成本底部一项应下调权重、而 9月16日的通胀约束同步放松,对铝净影响偏空(成本端让步大于宏观端受益);若磋商无果或 Brent 重新站上 108 美元(周四盘中高点),则成本底部抬升但宏观压力加大,记为双向且不予净定性;若无任何可核实的会议结果,则不作任何记录,不以传闻替代。
⚑ 第五,9月16日 FOMC 之前的定价,基准已由约 90% 修正为 85.6%。 判别带(据新基准重写):定价升至 95% 以上且 DXY 站上 99.5,则美元渠道开始传导,对铝进一步偏空;定价维持 85%–92% 而 DXY 仍在 99 附近横盘,则「日元/日本央行 9月17–18日会议为主导通道」的判断继续成立,宏观压力被部分隔离;定价回落至 80% 以下,则本轮宏观冲击属过度反应,对铝偏多。10月联邦基金利率条若终于可得,应立即取回以还原期限结构。
⚑ 第六,周二中国 8月工业增加值、固定资产投资、社零与房价——本期唯一能提前写下判别带的需求侧事件。 预期:社零约 +0.6%(与 7月持平),固投累计约 −7%。判别带(本期新写):社零 ≥ +1.5% 或固投累计好于 −6.0%,则「旺季未兑现」的判定应部分撤回,对铝偏多;社零 +0.3% 至 +1.0% 且固投在 −7.5% 至 −6.5%,则符合预期、维持现有判定;社零 ≤ 0% 或固投差于 −8.0%,则下游需求一项应整体移出多头清单,对铝明确偏空。同时观察房价与商品房销售,它们直接对应铝型材 37.05% 的利用率底部。
⚑ 第七,氧化铝的新空是继续加仓还是获利了结,以及现货报价何时追上盘面。 周一已印出 AO2701 2,673,低于全部六条均线 64.15–108.45,且持仓仍在增加。判别带(沿用):持仓继续上升且价格继续下跌,则空头趋势成立,成本支撑一项应完全移出多头清单;持仓下降而价格反弹,则周五属空头挤压前的最后一段;AO2610 若重新收于 2,718 之上需连续两日方可回撤成本口径一的降级(与采纳降级时的对称条件一致)。另观察最高收盘价能否重新越过 2,826.21,以及无日期的 2,707 现货口径能否更新——它目前仍高于 AO2701 收盘价,盘面贴水于自身现货参照的状态需要一个有日期的数字来确认或证伪。

免责声明

本简报仅供内部晨会参考。数据来源包括上海期货交易所公开日度与夜盘数据及其第三方镜像、Westmetall、TradingEconomics、CME FedWatch、美国劳工统计局、路透、CNBC、雅虎财经、福布斯、CNAL 世铝网、AL Circle、Mysteel、SMM(上海有色网)、中国海关总署(经转述)、Mining.com / Mining Weekly / Kitco、CERA、Wood Mackenzie、ING、摩根大通、EY-Parthenon 以及公开报道的高盛观点。标注「⚑ / ⊘」者为事实数据;标注「(interpretation)」或「(derivation)」者为本简报的判断或计算,二者在全文中明确区分。来源冲突时,所有口径平行列示且不予调和,除非一手来源证伪其中之一。依赖假设系数的推导(如 1.93 吨氧化铝/吨铝)先声明假设再计算。本期作出四项自我更正并予以明确记录:其一,上期公布的 FedWatch「约 90%」修正为有明确日期的 85.6%,该修正使事先写下的 95% 阈值距离由 5 个百分点变为 **9.4 个百分点;其二,沿用六期的 EGA「25% 对 50%」冲突被认定为本简报自身的口径错误(氧化铝厂爬产目标对电解铝厂电解槽复产,两项不同资产),标记退役;其三,沿用作为偏多项的 ING 条目实为 2025年12月8日、已陈旧九个月,予以正确注明日期并移出偏多清单;其四,上期关于现货「跟随而未领跌」的定性推断予以降级,因为周一交易日已把盘面带到 24,000 而现货报价停在周五,在 SMM 周一发布之前该推断所依据的价差已经过期。此外,本期拒绝一条流传中的「中国 8月电解铝产量 762 万吨」口径,理由是它相当于 4,500 万吨产能上限对应月度值的 2.03 倍、年化 9,144 万吨,算术上不可能;拒绝而非列示,是因为该数字并非口径差异而是类别错误。本简报不构成投资建议,亦不对据此作出的任何决策承担责任。**

1 · Market Performance

Statement of writing time, the nature of this edition, and the one caliber inference that matters most

Written the morning of Sunday, September 13, with BOTH markets closed and NO new market data of any kind in either market since last edition.
⚑⚑ WHAT IS NEW THIS EDITION IS NOT PRICE — IT IS CALIBER, DATING AND CALENDAR: a dated September 11 CME FedWatch reading of 85.6% superseding the approximately 90% carried; the full month-long FedWatch chain from August 11; the RESOLUTION of the six-edition EGA 25%-vs-50% flag as a caliber error of this brief’s own making; four 2026 global balance forecasts placed side by side for the first time, spreading 12.14x; four price forecasts spreading 44.29% of spot, of which only two carry dates; the correct date on the carried ING item (December 8, 2025); the REJECTION of a circulating China August output figure as arithmetically impossible; and the Monday and week-ahead calendar, including one energy event that lands INSIDE Monday’s SHFE session.
⊘ NOT RETRIEVABLE: any new Westmetall row (September 11 remains the last, re-queried and confirmed this edition); any SHFE print after the September 11 night session; SMM’s own primary quotations for a THIRTEENTH consecutive edition; SMM three-region daily inventory for a SEVENTH; the SHFE weekly warrant primary caliber; the October fed funds strip for a SIXTEENTH; LME volume and open interest.

⚑⚑ The most important fact on this page: Monday’s trading day is already open, and already 38.3% traded

⚑⚑ The SHFE September 14 trading day did NOT begin on Monday morning. It began at 21:00 on FRIDAY. The caliber note was established two editions ago; this edition computes its consequence.
⚑⚑ These Monday trading-day prices are therefore ALREADY PRINTED, not forecast: open 24,100, high 24,100, low 23,920, last 24,000; AL2610 volume 81,993, open interest 198,952 (−12,438).
⚑⚑ 81,993 lots is 38.3% of Friday’s ENTIRE trading-day volume of 213,873. Monday’s day session does not start the price action; it continues a session already more than a third complete (derivation).
⊘⊘ And the only Monday price that exists, 24,000, is ALREADY BELOW MA20. Any reasoning that starts Monday at 24,235 starts from the wrong number (interpretation).

⊘⊘ Four of six averages are already lost on that price, and the last support was approached to within 72.75

⚑⚑ The six SHFE averages, carried exact off the 224-row series at the September 11 close: MA5 24,433.00 / MA10 24,311.00 / MA20 24,055.00 / MA30 24,029.83 / MA40 23,847.25 / MA60 23,601.25.
⊘⊘ Against the 24,000 Monday has already printed: below MA5 by 433.00, below MA10 by 311.00, below MA20 by 55.00, below MA30 by 29.83; above MA40 by 152.75 and above MA60 by 398.75. FOUR OF SIX LOST.
⊘⊘ Against the 23,920 low Monday has already printed: below MA20 by 135.00, below MA30 by 109.83, and above MA40 by only 72.75. Friday night brought the last average still standing to within 72.75 points (derivation).
⊘ 23,920 is EXACTLY the August 28 close and the lowest tick since September 1.

⊘⊘ The pre-written Monday test is live — and it is sitting in its own no-conclusion zone

⚑⚑ The pre-written test, verbatim: close above 24,055.00 (MA20) equals oversold; close below 23,920 equals trend reversal with MA40 23,847.25 as the next reference; between 23,920 and 24,055 equals NO CONCLUSION, with no ruling made.
⊘⊘ The only Monday price in existence, 24,000, lies inside that band: 80.00 above the floor, 55.00 below the ceiling, at 59.3% of a 135.00-point band.
⊘ This brief does not pre-judge it. A test written before the number must wait for the number (derivation).

⚑⊘ And the two open-interest calibers contradict each other; Monday’s day session decides

⊘⊘ The September 11 DAY was EXIT: complex open interest −29,349 to 593,718, 1.96x this brief’s own void threshold; over three sessions exit outweighs roll 34,381 to 12,624; on the heaviest volume on this ledger (complex 414,471, +55.4%).
⚑ The September 11 NIGHT — the first portion of Monday’s trading day — was a ROLL: six contracts −1,612 in total, AL2610’s −12,438 absorbed by AL2611 +5,421, AL2612 +3,601, AL2701 +1,851, AL2702 +258; and AL2611 overtook AL2610 on open interest, 213,268 to 198,952.
⊘ The two readings are opposite and are NOT reconciled here. Monday’s day session decides which survives (interpretation).

⚑ London: carried unchanged — and this edition VERIFIED the carry rather than assuming it

⚑ Westmetall was re-queried this edition and September 11 remains the last published row. The carry is verified, not assumed.
⊘ Cash settled USD 3,274.00 (−69.00, −2.06%), the largest one-day cash fall on this ledger; 3-month USD 3,250.00 (−68.00); backwardation +USD 24.00, the first narrowing in five and by only USD 1.00; stocks 244,100 mt (−250), below the −500 substance threshold and correctly classified as noise.
⚑ The six LME averages carried exact for a FIFTH consecutive edition: MA5 3,320.80 / MA10 3,293.80 / MA20 3,256.05 / MA30 3,263.48 / MA40 3,244.04 / MA60 3,217.94.
⊘ Cash has LOST MA5 (by 46.80) and MA10 (by 19.80); it holds MA20 by only 17.95 and MA30 by only 10.52 — the narrowest support margin of this run.
⚑ Parallel caliber, listed and NOT reconciled: TradingEconomics reads USD 3,256.65 at September 11, −0.70% d/d, −1.52% m/m, +20.62% y/y — a USD 17.35 gap to the Westmetall cash settlement and USD 6.65 to the 3-month, consistent with a different contract reference rather than a data conflict (derivation).

⊘ Spot is carried at Friday — and this edition states plainly that it is STALE against Monday

⚑ SMM mirror caliber (September 11): A00 24,240 (−320); A00 Foshan 24,430 (−350); ADC12 24,300 (−150).
⚑ CNAL five-caliber set (September 11): Shanghai 24,230 (−310); Changjiang 24,240 (−320); Zhongyuan 24,240 (−320); Nanhai/Foshan 24,710 (−340); Guangdong Nanchu 24,440 (−360).
⊘ Every regional premium narrowed: Nanhai − Shanghai 480 (from 530); Nanchu − Shanghai 210 (from 260); Changjiang − Shanghai +10, unchanged. South China fell 340–360 against East China’s 310–320.
⚑ Basis at the September 11 close: +5 against the close, +50 against the settlement.
⊘⊘ But against the 24,000 Monday has already printed, the basis is +240. A +240 basis is not evidence of spot tightness — it is a stale quotation awaiting Monday morning’s reprice. This brief records the basis as UNKNOWN until SMM publishes on Monday, not as +5 (interpretation).
⊘ SMM’s own primary quotations, regional transaction premia, shipment and purchasing sentiment indices, alumina index and Chalco’s five listings remain login-walled for a THIRTEENTH consecutive edition, so no sentiment index is published and no proxy is substituted.

⊘⊘ Alumina: carried at the worst close on this ledger, and the LIVE number is the night one

⊘⊘ AO2701, September 11 day: open 2,762, high 2,792, low 2,691, close 2,697, settlement 2,731; −39 (−1.41%) on settlement, −61 (−2.21%) close-to-close. Volume 275,018 (+69.0%), open interest 220,066 (+24,675, +12.63%). Complex open interest 462,497 (+20,900).
⊘⊘ AO2701 in the Friday night session — already part of Monday’s trading day: 2,673 (−58, −2.12%), volume 92,893, open interest 235,026 (a further +14,960). AO2610 2,651 (−44) / AO2611 2,659 (−45) / AO2612 2,672 (−39) / AO2702 2,690 (−55). Five of five down, with position still building.
⚑⚑ The six alumina averages, carried exact and live off the 160-row AO2701 series: MA5 2,757.00 / MA10 2,753.10 / MA20 2,737.15 / MA30 2,738.67 / MA40 2,749.22 / MA60 2,781.45.
⊘⊘ The 2,673 Monday has already printed is below ALL SIX, by 84.00 / 80.10 / 64.15 / 65.67 / 76.22 / 108.45 — wider on every single average than the 60.00 / 56.10 / 40.15 / 41.67 / 52.22 / 84.45 recorded at the September 11 close (derivation).
⊘⊘ Cost caliber one (2,718) has failed twice and by fifty-one: AO2610 closed 2,667; clearance chain +39 → +19 → +2 → −3 → −51; the settlement of 2,695 is 23 below the line as well.
⊘⊘ Cost caliber two (2,826.21) is UNPRICED across the entire curve out to August 2027; the highest close is AO2708 at 2,815, still 11.21 below.
⊘ Front-to-tail contango +148 (5.55% of the front), a third consecutive narrowing; and the undated 2,707 Chinese alumina spot caliber still sits ABOVE the AO2701 close of 2,697, so the board still trades at a discount to its own physical reference.

⊘ Smelter margin proxy: a further 188.68 of compression in the night

⊘ Main pair (AL2610/AO2701, at the stated 1.93 t/t assumption): at Monday’s already-printed prices, 24,000 − 1.93 × 2,673 = 18,841.11, against 19,029.79 at the September 11 close — a further −188.68 (derivation).
⊘ Matched pair (AL2610/AO2610): 24,000 − 1.93 × 2,651 = 18,883.57.
⊘ Both calibers agree: the night’s decline was absorbed by smelter margin, not by the alumina leg (interpretation).

3 · Inventory

⊘ Entirely carried — and Monday’s print carries two weeks of information

⊘ SMM national ingot social inventory carried at 796,000 mt (September 10). Chain: Aug 24 860,000 → Aug 27 852,000 → Aug 31 837,000 → Sept 3 815,000 → Sept 7 802,000 → Sept 10 796,000.
⊘ Carried ruling unchanged: Thursday-to-Thursday −19,000, which is 52.8% of the prior 36,000 pace and 79.4% of August’s own approximately 23,935; Monday-to-Thursday only −6,000. The destock HALVED, it did not stop. Cumulative from the early-July peak of approximately 1.13 million mt: −334,000 mt (−29.6%).
⚑ Mysteel carried: Aug 31 809,000 → Sept 3 789,000 → Sept 7 774,000 → Sept 10 768,000; SMM minus Mysteel reads 28,000 / 26,000 / 28,000 / 28,000 and the constant definitional offset conclusion holds.
⚑ LME 244,100 mt (September 11, −250), re-queried and confirmed unchanged this edition.

⊘ A SEVENTH consecutive edition without a regional breakdown — and Monday’s national print will not supply one

⊘ SMM three-region daily inventory unavailable for a SEVENTH edition, carried at September 3: 641,000 mt (Guangdong 143,000, Wuxi 300,000, Gongyi 198,000).
⊘ The SHFE weekly warrant primary caliber is again unavailable; a third-party note has SHFE aluminium stock down approximately 3% week-on-week, listed and not weighted.
⊘ The sentence that has to be written down: Monday’s national number will tell us whether the deceleration continued, but not where it happened. Separating an East China arrival cluster from a genuine demand stall still requires the three-region series that has now been missing for seven editions (interpretation).

Interpretation

⚑ Ten of seventeen LME stock intervals are exactly zero, and the backwardation gave up only USD 1.00 on a USD 69.00 fall in cash. The two corroborate each other: whatever sold aluminium on Friday, it was not the physical prompt market (interpretation).
⊘ But the domestic side of this column is entirely blank this edition, and on a Monday whose session has already opened below MA20, the absence of inventory data is itself an exposure (interpretation).

4 · Macro Factors

⚑⚑ The FedWatch caliber resolves: 85.6%, not the approximately 90% published last edition

⚑⚑ A dated September 11 CME FedWatch reading puts a 25 bp September 16 hike at 85.6%. The approximately 90% carried last edition is corrected.
⚑⚑ The correction cuts both ways: the repricing was real, but it was 4.4 points smaller than published — and the pre-written escalation threshold of 95% is therefore 9.4 points away rather than 5, which makes it materially harder to trigger than last edition implied (derivation).
⚑ The full month-long chain is recoverable for the first time: 48.4% (August 11) → approximately 66% (August 31) → approximately 69.3% (pre-CPI) → 85.6% (post-CPI). That is +37.2 points in 31 days, of which +16.3 came in a single afternoon.
⚑ The 56% / 59% / 60.6% three-way caliber “conflict” carried all week is RETIRED — it was never a conflict, it was a mid-chain snapshot of the same series (derivation).
⊘ The October fed funds strip is missing for a SIXTEENTH consecutive edition, so the term structure of the Fed path remains unavailable.

⊘⊘ But the rates leg itself is the worst thing on this page: the 10-year at 4.97%

⊘⊘ The US 10-year yield stands at approximately 4.97%–4.98% (September 11), the highest since OCTOBER 2023, with the trigger an expanded Treasury buyback operation that delivered weaker-than-expected purchases.
⊘ US August CPI: headline +0.4% m/m, +3.4% y/y; core +0.3% m/m, +2.4% y/y; gasoline +3.9%.
⊘ EY-Parthenon switched from HOLD to a 25 bp HIKE, which would take the target range to 3.75%–4.00%.
⚑ The dollar has not kept up: DXY 99.095 (+0.047, +0.05%), month −0.92%. On the day the Fed path repriced sharply, the dollar moved five basis points — and this brief’s carried judgement that the yen and BoJ channel remains dominant continues to hold (interpretation).

⊘⚑ Energy: two Brent calibers, 0.18% apart, identical conclusion

⊘ Settlement caliber: Brent USD 104.61, −2.8% on the day, +8.7% on the week.
⊘ TradingEconomics caliber: USD 104.42, −2.98% on the day, approximately +9% on the week.
⚑ The gap is USD 0.19, or 0.18%. Both are listed, neither is preferred, and the conclusion is identical on either: a third consecutive close above USD 100 raises the smelter cost floor (bullish) and makes September 16 harder (bearish) (interpretation).
⊘⊘ And Monday carries an event acting directly on this leg: GCC and Iranian diplomats are due to meet in OMAN on Monday to discuss a temporary arrangement for shipping through the Strait of Hormuz. That meeting lands INSIDE the SHFE Monday session (see section 9).

⊘ China: the two demand numbers landing Tuesday are weak before they arrive

⊘ Tuesday September 15 brings August industrial production, energy production, fixed asset investment, home sales, total retail sales and the 70-city house price index.
⊘ Expectations: retail sales approximately +0.6% y/y, UNCHANGED from July; cumulative fixed asset investment expected to deepen to approximately −7% year-to-date; infrastructure investment characterised as weak and real estate investment as muted.
⊘ For aluminium these three map directly onto construction, grid and appliance channels, and the expectation set is bearish before the print (interpretation).
⚑ The one offset is exports, and this edition dates it fully: August unwrought aluminium and product exports 626,000 mt, +17.2% y/y (August 2025: 534,000 mt), but −2.6% m/m (July: 643,000 mt); January–August 4.67 million mt, +16.7% y/y — so August ran 7.24% ABOVE the eight-month average of 583,750 mt/month (source article dated September 8, 2026) (derivation).

⊘ Japan: the September 17–18 BoJ meeting remains this brief’s designated dominant channel

⊘ The Bank of Japan decides on September 17–18 with a hike increasingly priced. Board member Takata said the bank should hike “nimbly” and suggested moves faster or larger than its recent semiannual pace; Governor Ueda kept the door open.
⚑ On September 3 the yen reached approximately 155.28, up more than 2% on the day, a one-month intraday high.
⊘ The transmission to aluminium is through the dollar index rather than directly — but with the Fed path barely moving the dollar, it is the variable that explains why the dollar is not moving (interpretation).

5 · Fundamentals

⚑⚑ This edition resolves a six-edition flag — and the error was this brief’s own

⚑⚑ The EGA 25%-vs-50% conflict does not exist. The 50% is the ramp target for the Al Taweelah ALUMINA REFINERY, stated on July 10, 2026 as reachable within days, with full capability expected by end-2026; the 18% is the Al Taweelah SMELTER pot restart (227 of 1,262 reduction cells as at August 11). Two different assets.
⚑⚑ The flag is RETIRED and the error is recorded as this brief’s own caliber confusion (derivation).
⊘ And the substance of the item is BEARISH, and it sits on the alumina side: Al Taweelah produced 2.4 million mt of alumina in 2025, close to half of EGA’s own alumina requirement. A refinery of that size moving from zero toward full capability by end-2026 stands on the same side of the ledger as everything else in this brief’s alumina column (interpretation).

⊘⊘ The 2026 global balance is not a number — it is a 12.14x spread

⊘⊘ Four forecasts placed side by side for the first time and deliberately NOT reconciled: AL Circle approximately 140,000 mt deficit (2026, undated within the year); ING approximately 200,000 mt (600,000 if the Mozal smelter closes), dated December 8, 2025; Wood Mackenzie approximately 900,000 mt, dated August 4, 2026; J.P. Morgan 1.7 million mt, dated August 21, 2026.
⊘⊘ Highest over lowest = 12.14x; J.P. Morgan over ING = 8.5x; Wood Mackenzie over AL Circle = 6.43x (derivation).
⊘ The sentence that follows: anyone quoting “a 2026 deficit of X” without naming a source is quoting one point inside a twelvefold range (interpretation).

⊘⊘ The price forecasts spread 44.29% of spot, and only two of four carry dates

⊘ Benchmark: the September 11 LME cash settlement of USD 3,274.00.
⚑ J.P. Morgan (dated August 21, 2026): Q3 2026 USD 3,800 (+16.07%), Q4 2026 USD 3,700 (+13.01%), FY2026 average USD 3,587 (+9.56%); Q1 2027 3,450, Q2 3,200, Q3 2,900, Q4 2,750 (−16.00%), FY2027 average 3,075 (−6.08%).
⚑ TradingEconomics model: Q3 USD 3,267.98 (−0.18%), 12-month USD 3,489.17 (+6.57%).
⊘ ING (dated December 8, 2025): FY2026 average USD 2,900 (−11.42%).
⊘ Goldman Sachs: USD 2,350 (−28.22%), UNDATED — listed and NOT weighted.
⊘⊘ Top to bottom is USD 1,450 = 44.29% of spot (derivation).
⊘ And the most bullish of them is simultaneously a PEAK call: J.P. Morgan’s 3,800 is the high of its entire eight-quarter path, Q4 2027 at 2,750 is 27.63% below it, and FY2027 is 14.27% below FY2026. The most bullish datable forecast on this page says the cycle high is in the quarter now ending (interpretation).
⚑ Its reasoning runs on this brief’s own supply ledger: Middle East smelter shutdowns removing over 6% of global supply, Strait of Hormuz disruption, China’s 45 million mt cap — with an explicit upside risk if Chinese policymakers curtail aluminium exports.

⊘ One circulating China output caliber is REJECTED rather than listed

⊘ One caliber returns China August primary aluminium output of 7.62 million mt.
⊘ Grounds for rejection (arithmetic): the 45 million mt/yr cap implies 3.75 million mt per month; 7.62 is 2.03x that and annualises to 91.44 million mt — arithmetically impossible.
⊘ And it sits within 3.91% of the 7.333 million mt monthly ALUMINA run rate implied by an 88 million mt/yr caliber — so it is almost certainly an alumina figure mislabelled. Rejected, not published as data (derivation).

⊘ Alumina supply: one new capacity item

⊘ NEW: a 2.4 million mt/yr alumina refinery at Fangchenggang, Guangxi, named among the Chinese capacity coming online into an already-surplus domestic alumina market.
⊘ AL Circle caliber: 2026 smelter-grade alumina demand approximately 145.2 million mt, broader market near 147 million mt.
⊘ Carried: CERA’s 1.79 million mt 2026 global alumina surplus; China’s January–June alumina imports 2.277 million mt (+749.1% y/y); one caliber putting 2026 planned new domestic alumina capacity at approximately 8.6 million mt against full-year oversupply above 10 million mt (undated, not weighted).

⊘ Downstream and trade: carried, no new data

⊘ Aluminium extrusion capacity utilisation at a 37.05% cyclical floor; leading downstream processors’ operating rate 59.9% (August 13); secondary alloy operating rate 49.1%.
⊘ Carried: Mexico’s definitive anti-dumping duties on extrusions from China AND the United States (September 10); the EU’s planned scrap export ban to non-OECD countries; Canada’s 25%→50% step-up with aluminium rod named.
⚑ Carried: Mysteel’s August A00 monthly averages — East China 23,869 (+3.34% m/m), South China 24,010 (+3.79%), Central China 23,762.4 (+3.42%). Price rose in August and operating rates did not follow.

6 · Primary Aluminum Spot Market

⊘ No new spot transaction information this edition; everything below is carried from September 11

⚑ Five-region quotations (September 11, CNAL caliber): Shanghai 24,210–24,250, avg 24,230 (−310); Changjiang 24,220–24,260, avg 24,240 (−320); Zhongyuan 24,240 (−320); Nanhai/Foshan 24,660–24,760, avg 24,710 (−340); Guangdong Nanchu 24,410–24,470, avg 24,440 (−360).
⚑ SMM mirror caliber: A00 24,240 (−320); A00 Foshan 24,430 (−350); ADC12 24,300 (−150).
⊘ Regional structure: Nanhai − Shanghai 480; Nanchu − Shanghai 210; Changjiang − Shanghai +10. A full give-back after four sessions of widening, led by South China.

⊘⊘ Shipment and purchasing sentiment: no index, and this edition DOWNGRADES even the qualitative inference

⊘ SMM’s shipment willingness and purchasing sentiment indices are unavailable for a THIRTEENTH consecutive edition. This brief publishes neither and substitutes no proxy.
⊘⊘ The qualitative inference made last edition is downgraded. Last edition inferred from “A00 fell 320 while the basis still converged to +5” that spot followed rather than led. That held at the September 11 close — but Monday’s trading day has already taken the board to 24,000 while spot quotations stopped on Friday. Until SMM publishes on Monday, any judgement about shipment or purchasing willingness is being made on an expired spread, so this brief declines to make one (interpretation).
⚑ The one statement that still stands: all five regional premia narrowed, which is evidence of rising willingness to ship; whether downstream buying willingness offsets it has to wait for Monday.

⚑ Basis and delivery-month structure

⚑ Basis at the September 11 close: +5 (against 24,235) and +50 (against the 24,190 settlement).
⊘⊘ Against the 24,000 Monday has already printed, the basis is +240 — recorded as a stale quotation, not a spot premium. Chain: +70 → +55 → −20 → −20/0 → +70 → +5 → UNKNOWN, pending Monday (derivation).
⚑⚑ The main-contract handover is underway: in the Friday night session AL2611 open interest 213,268 exceeded AL2610’s 198,952 by 14,316; on the day session AL2610 still led by 3,543. This brief will switch its main-contract caliber once a day session confirms it, and will label the switch in that edition.

7 · Key Price Drivers

⚑ Bullish factors (16)

⚑⚑ The pre-written 95% threshold is harder to trigger than last edition implied: actual FedWatch 85.6%, not approximately 90% — 9.4 points away, not 5.
⚑⚑ London cash fell USD 69.00 and the backwardation gave up only USD 1.00 (+25.00 → +24.00). The physical prompt did not participate in the selling.
⚑⚑ LME stocks 244,100 mt with TEN of seventeen intervals exactly zero; the −250 draw sits below the −500 substance threshold.
⚑⚑ On the day the Fed path repriced sharply, DXY rose only 0.05% — dollar-channel transmission is plainly incomplete.
⚑⚑ The named 24,075 held to the tick on September 11, and the daytime hours bought back 110 of the prior night’s 450.
⚑⚑ The Friday night session was a ROLL, not an exit: six contracts −1,612 in total, with AL2610’s −12,438 absorbed down the curve.
⚑⚑ Shanghai out-performed London by 1.02 percentage points (ratio 7.3257 → 7.4023); the macro shock was only partly accepted in Shanghai.
⚑⚑ J.P. Morgan (August 21) is the first DATED forecast on this ledger sitting ABOVE spot: Q4 2026 3,700 (+13.01%), FY2026 average 3,587 (+9.56%), on a 1.7 million mt 2026 deficit.
⚑ MA40 (23,847.25) and MA60 (23,601.25) are still intact on the only Monday price that exists.
⚑ Spot followed rather than led on September 11 (A00 −320, basis +5), and the arithmetic retroactively proved last edition’s mirror rows genuine.
⚑ Brent held above USD 100 for a third consecutive session (104.61 / 104.42 on two calibers), raising the smelter cost floor.
⚑ China’s August aluminium exports 626,000 mt (+17.2% y/y), 7.24% ABOVE the eight-month average; January–August +16.7%.
⚑ China’s total August exports +25% with a USD 119.09 bn surplus — no weakening on the external demand side.
⚑ Carried: LME stocks near a 36-year low; Wood Mackenzie’s approximately 900,000 mt 2026 deficit (August 4).
⚑ Carried overseas supply disruption: Rio Tinto Yarwun (−1.2 million mt/yr from October), Alcoa’s 200,000–300,000 mt guidance cut, Qatalum at approximately 60% of nameplate, Alba lines 1–3, Guinea’s export cap, Rusal Friguia (August 24), GCC output −44% y/y in July.
⚑ J.P. Morgan’s explicit upside risk: a Chinese policy turn toward curtailing aluminium exports would put prices significantly higher than its own path.

⊘ Bearish factors (31)

⊘⊘ Monday’s trading day is ALREADY OPEN and already BELOW MA20: open 24,100, high 24,100, low 23,920, last 24,000.
⊘⊘ That single existing price has already lost FOUR of six averages (below MA5 by 433.00, MA10 by 311.00, MA20 by 55.00, MA30 by 29.83).
⊘⊘ The Friday night low of 23,920 brought MA40 (23,847.25), the last average standing, to within 72.75 points.
⊘⊘ 81,993 lots are already traded on Monday’s book, 38.3% of Friday’s whole trading day — the break happened in a liquid session, not in thin night noise.
⊘⊘ September 11 day-session complex open interest −29,349, 1.96x this brief’s own void threshold; the roll explanation is dead.
⊘⊘ Over three sessions exit outweighs roll 34,381 to 12,624, on the heaviest volume on this ledger (+55.4%).
⊘⊘ FedWatch at 85.6% for a September 16 hike, +37.2 points within the month (48.4% on August 11).
⊘⊘ The US 10-year at approximately 4.97%–4.98%, the highest since October 2023, triggered by a Treasury buyback that under-delivered.
⊘⊘ LME cash −USD 69.00 (−2.06%), the largest one-day fall on this ledger, losing MA5 and MA10 with only 17.95 and 10.52 of clearance on MA20 and MA30.
⊘⊘ AO2701’s already-printed 2,673 is below ALL SIX averages, and wider on every one of them than at the close.
⊘⊘ Alumina open interest kept RISING into the fall: AO2701 +24,675 on the day, complex +20,900, a further +14,960 in the night — new shorts entering.
⊘⊘ Cost caliber one (2,718) failed twice and the margin widened to 51; the settlement is 23 below the line too.
⊘⊘ Cost caliber two (2,826.21) is unpriced on all twelve alumina contracts.
⊘⊘ The four-caliber 2026 balance spread is 12.14x — the “deficit” argument itself lacks quotable precision.
⊘⊘ The price forecast spread is 44.29% of spot, and two of the four are undated or stale.
⊘⊘ The most bullish datable forecast is simultaneously a peak call: J.P. Morgan’s Q3 2026 3,800 is its eight-quarter high, and FY2027 is 14.27% below FY2026.
⊘⊘ The carried ING item is dated December 8, 2025 — nine months stale — and its USD 2,900 2026 average is 11.42% BELOW Friday’s cash. Moved off the bullish list.
⊘⊘ EGA’s Al Taweelah refinery (2.4 million mt in 2025, approximately half of EGA’s own requirement) is expected at full capability by end-2026 — another alumina supply addition.
⊘ A 2.4 million mt/yr alumina refinery at Fangchenggang, Guangxi is entering an already-surplus domestic market.
⊘ Tuesday’s China expectations are weak before the print: retail sales approximately +0.6% (unchanged from July), fixed asset investment deepening to approximately −7%, infrastructure weak and property muted.
⊘ China’s August aluminium exports fell 2.6% month-on-month (626,000 against July’s 643,000).
⊘ US August core CPI +0.3% m/m (from +0.2%), headline +3.4% y/y.
⊘ EY-Parthenon switched to a 25 bp hike call, target range to 3.75%–4.00%.
⊘ All five regional spot premia narrowed, South China leading the fall (−340 to −360); Nanhai − Shanghai 480 from 530.
⊘ The main-pair margin proxy compressed a further 188.68 in the night to 18,841.11.
⊘ Front-to-tail alumina contango narrowed a third time to +148 — the surplus is being priced forward.
⊘ The undated 2,707 alumina spot caliber still sits above the AO2701 close, so the board is at a discount to its own physical reference.
⊘ SMM three-region daily inventory missing for a SEVENTH edition; Monday’s national print cannot answer where the deceleration happened.
⊘ SMM primary quotations and sentiment indices missing for a THIRTEENTH edition; no sentiment indicator can be published.
⊘ The SHFE weekly warrant primary caliber is again unavailable; the October fed funds strip is missing for a SIXTEENTH.
⊘ Carried: extrusion capacity utilisation at the 37.05% floor and downstream operating rates at 59.9%; Mexico’s dual-origin anti-dumping duties; Goldman’s undated USD 2,350, 28.22% below Friday’s cash.

8 · One-Sentence Market View

Nothing traded this edition, so what this brief can add is not a price but three things — one caliber it wrote down last edition without computing its consequence, three self-corrections, and one piece of arithmetic that comes from placing the forecasts side by side; and the first of these is the most important fact on the page: the SHFE September 14 trading day did not begin on Monday morning, it began at 21:00 on Friday, so Monday is ALREADY OPEN AND ALREADY 38.3% TRADED — open 24,100, high 24,100, low 23,920, last 24,000, with 81,993 lots of AL2610 already on its book against Friday’s whole trading-day total of 213,873, which means any reasoning that starts Monday at 24,235 starts from the wrong number, and the real starting point of 24,000 is already 55.00 below MA20 (24,055.00) and 29.83 below MA30 (24,029.83), with four of six averages lost, while Friday night’s 23,920 low — exactly the August 28 close — brought MA40 (23,847.25), the last average still standing, to within 72.75 points; more to the point, the pre-written Monday test is sitting inside its own no-conclusion zone, since 24,000 lies 80.00 above the 23,920 floor and 55.00 below the 24,055.00 ceiling, at 59.3% of a 135.00-point band, and this brief therefore declines to pre-judge it; the three self-corrections are, in order: the correct FedWatch caliber is 85.6%, not the approximately 90% published last edition — a correction that cuts against this brief in that the repricing was 4.4 points smaller than printed, and in its favour in that the pre-written 95% escalation threshold is therefore 9.4 points away rather than 5, materially harder to trigger than implied — while the full chain becomes recoverable for the first time as 48.4% (Aug 11) → approximately 66% (Aug 31) → approximately 69.3% (pre-CPI) → 85.6% (post-CPI), +37.2 points in 31 days, retiring the 56/59/60.6 three-way “conflict” as a mid-chain snapshot of one series; the six-edition EGA 25%-vs-50% flag was never a data conflict but this brief confusing two assets — the 50% is the Al Taweelah ALUMINA REFINERY ramp target (July 10, within days, full capability by end-2026, 2.4 million mt produced in 2025, close to half of EGA’s own requirement) and the 18% is the Al Taweelah SMELTER pot restart (227 of 1,262 cells, August 11) — flag retired, error owned, and the substance lands on the bearish alumina side; and the ING item carried as bullish is dated December 8, 2025, nine months stale, with a USD 2,900 2026 average 11.42% below where cash actually settled on Friday, so it comes off the bullish list; placing the forecasts side by side then produces the two most uncomfortable numbers here — the 2026 global balance is not a number but a 12.14x spread (AL Circle approximately 140,000 mt, ING approximately 200,000, Wood Mackenzie approximately 900,000, J.P. Morgan 1.7 million) and the price forecasts run USD 1,450 top to bottom, equal to 44.29% of Friday’s cash, with the most bullish datable path (J.P. Morgan, August 21: Q3 3,800, Q4 3,700, FY2026 average 3,587) simultaneously a peak call, its 3,800 being the high of the entire eight-quarter path and FY2027 sitting 14.27% below FY2026; a circulating “China August primary aluminium output 7.62 million mt” is rejected rather than listed, being 2.03x the 3.75 million mt/month implied by the 45 million mt cap and annualising to 91.44 million mt, while sitting within 3.91% of the alumina monthly run rate — almost certainly a mislabelled alumina figure; everything else is carried, and the carries were verified by re-query rather than assumed — London’s September 11 row is still the last (cash 3,274.00, backwardation +24.00, stocks 244,100, cash holding MA20 by only 17.95), alumina is carried at the worst close on this ledger with Monday’s already-printed 2,673 below all six averages and wider on every one of them than at the close and the main-pair margin proxy compressing a further 188.68 to 18,841.11, domestic inventory is carried at 796,000 mt with the halved-destock ruling unchanged and the three-region breakdown missing for a seventh edition, and spot is carried at Friday with an explicit warning attached: the +5 basis everyone will quote is measured against a close that is no longer the live price, and against the 24,000 Monday has already printed the basis is +240, which is not tightness but a stale quotation awaiting reprice; on the calendar, four pre-written tests land inside 48 hours and one of them — the GCC–Iran meeting in Oman on a temporary Strait of Hormuz arrangement — lands INSIDE Monday’s SHFE session, while Tuesday’s China retail sales (expected approximately +0.6%, unchanged) and fixed asset investment (expected to deepen to approximately −7%) are bearish before they print; the honest position is that the structure is intact only on the closing caliber, the session that decides it is already more than a third over, and nothing in this edition changes a price (interpretation).

9 · Watchlist for Monday’s Day and Night Sessions (September 14)

⚑⚑ First, MA20 (24,055.00) — and it must be read from “already below it intraday”, not from “waiting for the open”. Monday has already printed open 24,100, high 24,100, low 23,920, last 24,000, on 81,993 lots (38.3% of Friday’s whole day). The test is carried VERBATIM and unamended: a day-session close above 24,055.00 makes Friday night’s break oversold; a close below 23,920 escalates this pullback to a trend reversal with MA40 23,847.25 as the next reference; a close between 23,920 and 24,055 remains a NO-CONCLUSION zone. Reference ladder: 24,311.00 (MA10) / 24,235 (Sept 11 close) / 24,100 (Monday open and high) / 24,075 (the level that held to the tick) / 24,055.00 (MA20) / 24,029.83 (MA30) / 24,000 (Monday last) / 23,920 (Monday low so far) / 23,847.25 (MA40) / 23,601.25 (MA60).
⚑⚑ Second, the day and night open-interest calibers contradict each other, and Monday’s day session decides. September 11 day was EXIT (complex −29,349, 1.96x the void threshold; exit 34,381 against roll 12,624 over three sessions); the night was a ROLL (six contracts −1,612 in total, AL2611 overtaking AL2610 by 14,316). Bands, carried and tightened: complex open interest down a further 10,000 or more makes exit a trend and requires downgrading every rebound on this leg; up 5,000 or more makes Friday’s day session a one-off liquidation; within ±5,000 is undecided. Watch whether AL2611 day-session open interest passes AL2610 (AL2610 still led by 3,543 on Friday’s day); once confirmed, this brief switches its main-contract caliber in that same edition and labels the switch.
⚑⚑ Third, Monday’s SMM national ingot inventory print, which carries two weeks of information. Carried chain Sept 7 802,000 → Sept 10 796,000, Thursday-to-Thursday −19,000 (52.8% of the prior pace). Bands carried from last edition’s rewrite, and they cover the number line completely: ≤ 785,000 = the destock re-accelerated; 785,001–792,000 = the current decelerated pace held; 792,001–799,000 = further deceleration; ≥ 799,001 = stall or build. Also watch whether Mysteel’s same-day figure stays 26,000–28,000 below SMM (a deviation would reopen the constant-offset conclusion) and whether the three-region breakdown finally reappears after seven editions — without it, Monday’s national number says whether the deceleration continued but not where it happened.
⚑⚑ Fourth, the GCC–Iran meeting in Oman on Monday — the only external event this edition can identify that lands INSIDE an SHFE session. The subject is a temporary arrangement for shipping through the Strait of Hormuz. Bands written before the number: a concrete interim transit arrangement plus Brent breaking below USD 100 would downgrade the energy cost-floor argument while easing the September 16 inflation constraint — net bearish for aluminium, because the cost-side concession outweighs the macro-side relief; no agreement, or Brent reclaiming USD 108 (Thursday’s intraday high), raises the cost floor but worsens the macro pressure and is recorded as two-sided with no net ruling; no verifiable outcome means nothing is recorded at all — rumour does not substitute for a result.
⚑ Fifth, the pre-FOMC pricing into September 16, now rebased from approximately 90% to 85.6%. Bands rewritten against the corrected base: pricing above 95% together with DXY above 99.5 means the dollar channel has begun to transmit — further bearish for aluminium; pricing holding 85%–92% with DXY still flat near 99 means this brief’s yen/BoJ-is-the-dominant-channel judgement continues to hold and the macro pressure stays partly insulated; pricing falling below 80% would mark this macro shock as an overreaction — bullish for aluminium. If the October fed funds strip finally becomes retrievable, take it immediately to restore the term structure.
⚑ Sixth, Tuesday’s China August industrial production, fixed asset investment, retail sales and house prices — the only demand-side event this edition can band in advance. Expectations: retail sales approximately +0.6% (unchanged from July), cumulative fixed asset investment approximately −7%. Bands written this edition: retail sales ≥ +1.5% or fixed asset investment better than −6.0% would partially withdraw the “peak season has not shown up in the data” ruling — bullish; retail sales +0.3% to +1.0% with fixed asset investment between −7.5% and −6.5% is in line and leaves the current ruling standing; retail sales ≤ 0% or fixed asset investment worse than −8.0% takes downstream demand off the bullish list entirely — clearly bearish. Watch house prices and home sales alongside, since they map directly onto the 37.05% extrusion utilisation floor.
⚑ Seventh, whether alumina’s new shorts keep adding or take profit, and when the spot quotation catches up to the board. Monday has already printed AO2701 at 2,673, below all six averages by 64.15–108.45, with open interest still rising. Bands carried: open interest rising with price still falling establishes the short trend and removes cost support from the bullish list entirely; open interest falling on a price rebound makes Friday the last leg before a short squeeze; AO2610 reclaiming 2,718 requires TWO consecutive closes before the cost caliber one downgrade can be withdrawn (the symmetric condition to the one used to adopt it). Also watch whether the highest close can retake 2,826.21, and whether the undated 2,707 spot caliber gets refreshed — it still sits above the AO2701 close, and a board trading at a discount to its own physical reference needs a dated number to confirm or falsify it.

Disclaimer

This brief is for internal morning-meeting reference only. Sources include Shanghai Futures Exchange public daily and night-session data and third-party mirrors, Westmetall, TradingEconomics, CME FedWatch, the US Bureau of Labor Statistics, Reuters, CNBC, Yahoo Finance, Forbes, CNAL, AL Circle, Mysteel, SMM, China’s General Administration of Customs (as reported), Mining.com / Mining Weekly / Kitco, CERA, Wood Mackenzie, ING, J.P. Morgan, EY-Parthenon, and publicly reported Goldman Sachs views. Items marked ⚑ / ⊘ are factual data; items marked (interpretation) or (derivation) are this brief’s judgements or calculations, and the two are distinguished throughout. Where sources conflict, all calibers are listed in parallel and NOT reconciled unless a primary source falsifies one of them. Derivations that rely on an assumed coefficient (such as 1.93 t alumina per tonne of aluminium) state the assumption before the calculation. FOUR SELF-CORRECTIONS ARE MADE AND RECORDED THIS EDITION: first, the FedWatch figure of approximately 90% published last edition is corrected to a dated 85.6%, which moves the pre-written 95% threshold from 5 points away to 9.4 points away; second, the six-edition EGA 25%-vs-50% conflict is identified as this brief’s own caliber error — an alumina refinery ramp target set against a smelter pot restart, two different assets — and the flag is retired; third, the ING item carried as a bullish line is correctly dated to December 8, 2025, nine months stale, and is moved off the bullish list; fourth, last edition’s qualitative inference that spot followed rather than led is DOWNGRADED, because Monday’s trading day has already taken the board to 24,000 while spot quotations stopped on Friday, so the spread the inference rested on has expired pending Monday’s SMM publication. In addition, this edition REJECTS rather than lists a circulating China August primary aluminium output figure of 7.62 million mt, on the grounds that it is 2.03x the monthly rate implied by the 45 million mt cap and annualises to 91.44 million mt; it is rejected rather than listed because it is a category error rather than a caliber difference. This brief does not constitute investment advice and accepts no responsibility for any decision made on the basis of it.