供应链分析
产业链结构、上下游与瓶颈环节研究 · 共 1315 条 · 滚到底自动加载更多
-
AXTI为光子学最上游标的，ETF映射Semivision所述中游瓶颈。
@cronked $AXTI 是我的个人最爱，但它在光子学领域已经处于最上游了。上面的 ETF 只是将瓶颈名称映射到 Semivision 描述的主要瓶颈上，而这些瓶颈更多涉及中游公司。
英文原文
@cronked $AXTI is my personal favorite but it’s as upstream as you can go for photonics. The ETF above is just bottleneck name mapping to what Semivision described as their top bottlenecks, which were more midstream companies.
-
A开头半导体股因订单与瓶颈大涨，暗示供应链瓶颈改善。
哇…… $AEHR 今日因超大规模云服务商订单上涨30%+。 $ACMR 本周因中国存储（长鑫存储+长江存储）上涨29%。 $AXTI 本周因光子学(CPO)产能爬坡及谷歌资本支出上涨42%。 $AMKR 本周因先进封装瓶颈上涨19%。 也许以字母A开头的瓶颈是个好兆头？
英文原文
Wow… $AEHR up 30%+ today on hyperscaler orders. $ACMR up 29% this week on China memory (cmxt + ymtc) $AXTI up 42% this week on photonics ramp and Google capex spend $AMKR up 19% this week w/ advanced packaging bottlenecks. Maybe the bottlenecks starting with letter A are a good sign?
-
揭秘监管与保证金调整如何重创加密市场，看好比特币及持牌机构反弹。
我觉得伯恩斯坦(Bernstein)和机构错过了比特币暴跌的许多原因。华尔街日报(WSJ)和媒体也是如此，这就是为什么散户一无所知。 两点原因： 1. “清晰法案(Clarity Act)”加上“天才法案(Genius Act)”的一部分，是银行和政府为了加强美元而对加密货币投下的战术核弹。“安全”目的纯属胡扯。例如（稳定币储备必须与美元/国债保持1:1）。 据CEO去年会议透露，Tether持有10万+枚比特币，且2025年是净买入方。随着该法案实施，像Tether这样的主要买家可能会在2026-2027年被迫卖出，因为其约1800亿美元市值的稳定币中持有大量南美农场等流动性差的资产，而其10万+枚比特币具有高流动性。 -> 稳定币收益率 如果交易所和金融科技(Fintech)公司无法通过稳定币和其他存款提供收益，这将极具破坏性。巨额流动性将从加密货币中流出。 此外，关于美元抵押品、资金流中的银行以及国家CFTC许可证的不同条款，可能会摧毁许多交易所/去中心化交易所(DEX)/金融科技公司的头部，并伤害山寨币。有太多毒丸条款会扼杀任何非银行金融科技/加密货币公司。例如，使银行成为有效的垄断者。但这可以在另一篇帖子中解释。 2. 白银流动性核弹。这是CME和交易所将保证金提高到不合理水平以迫使清算级联，并以“鹰派”美联储主席作为叙事借口。 其他风险投资(VC)声称某亚洲公司爆仓并被迫清算来自白银交易的比特币，但这都是多方面的，很可能源于上述两点。 主流媒体几乎未报道百分比变化和6次以上的加息。只是责怪美联储主席。 这也蔓延到股市中任何高贝塔(high beta)资产，但看起来像是一次孤立暴跌。 基本上，没人能报道《清晰法案》是摧毁加密货币的毒丸，因为银行游说，以及CME保证金上调（以迫使白银清算）。如果报道了，贝森特(Bessent)和美国政府就会追究他们。所以散户一无所知，并被煤气灯操纵(gaslit)认为“清晰法案”对加密货币是好事。 实际上，美国政府只是在通过游说加强美元+银行垄断，同时在这个过程中摧毁加密货币。（如果我是政府，这很有道理，我只是引用诚实的后果） 话虽如此，比特币在当前水平感觉像是免费的地皮，我预计该资产以及申请银行特许状的特定加密货币公司将恢复。
英文原文
I feel like Bernstein and institutions missed a lot on why Bitcoin crashed. Same with WSJ + media, which is why nobody in retail knows. Two things: 1. “Clarity Act” + part of the Genius Act is a tactical nuke on crypto by banks + Gov to strengthen the dollar. “Safety” purposes is just BS. -> For example (Stablecoin reserves must be kept 1:1 with USD/treasuries) Tether has 100k+ Bitcoin (as per CEO last year conference) and were a net buyer in 2025. With the act the main buyers like Tether would probably be a forced sellers 2026-2027 with their $180b-ish stablecoin MC since they have a lot of illiquid assets like farms in South America while their 100k+ Bitcoin is liquid. -> Stablecoin yield This is extremely destructive if exchanges and fintechs were not able to give yield off stablecoin and other deposits. An enormous of liquidity would flow out of crypto. There’s also different clauses around exchange rules regarding USD collateral + banks in flow of funds + national cftc licenses that would likely blow up a lot of exchanges/dexs/fintechs tops and harm altcoins. There’s just way too many poison pill clauses that just kills off any non-bank fintech/crypto company. Eg. Making banks an effective monopoly. But can explain in another post. 2. Silver liquidity nuke. This was CME + exchange margin hikes to unreasonable levels to force liquidation cascades, with a “hawkish” fed chair as a narrative excuse. Other VCs were claiming some Asian firm blew up and were forced to liquidate a bitcoin from the silver trade but it’s all multifaceted, likely stemming from above two. Almost zero mainstream media reporting over changes to percentages and 6+ hikes. Just blaming the fed chair. This also trickled into anything high beta in the stock market but looks like an isolated crash. But basically, nobody can report that the Clarity Act is a poison pill nuking crypto because of bank lobbying and that the CME margin hikes (to force Silver liquidations). Bessent and the Us gov would just go after them if they did lol. So nobody in retail knows and are being gaslit that “Clarity Act” is a good thing for crypto. When in reality US gov is just doing everything to strengthen the dollar + bank monopoly from their lobbying while blowing up crypto in the process. (Which makes sense if I’m the government, I’m just citing the honest repercussions) That being said Bitcoin at these levels feels like free real estate and I expect the asset + crypto companies that applied for bank charters in specific to recover.
-
回顾光子学供应链上游 $AXTI 的20倍涨幅及持有难度。
当时没多少散户真正了解光子学的极端扩张。甚至 $LITE 都极不引人注目。因此，一路追溯到 $AXTI，在他们仍受出口管制禁令时买入并全程持有，需要极大的信念。但错过了那20倍的涨幅确实很遗憾。
英文原文
not many too many retail really knew about the extreme photonics ramp back then. Even $LITE was extremely under the radar. So going all the way upstream to $AXTI while they were still export control banned + holding through it all would take a lot of conviction. But yeah sucks to miss out on that 20x
-
AEVA凭借确定性速度数据优势，降低世界模型计算负载。
世界*模型。显然，如果世界上最先进的世界模型需要多普勒级别的精度，与$OUST相比，$AEVA是少数能做到这一点的公司之一？如果你给Cosmos提供“概率性”深度数据（来自$OUST），模型必须消耗大量计算周期来计算物体的移动位置。$AEVA直接输入确定性速度数据。“如果模型必须从连续深度帧中推断运动，那比直接接收速度作为输入需要更多的计算。”
英文原文
World* models. Apparently if the world's most advanced World Models require Doppler-level precision, $AEVA is one of the only companies that can do it compared to companies to $OUST here? If you give Cosmos "probabilistic" depth data (from $OUST ) the model has to spend massive amounts of compute cycles calculating where things are moving. $AEVA just feeds in deterministic velocity data. “If a model has to infer motion from sequential depth frames, that’s additional computation versus receiving velocity as a direct input”
-
日本在上游半导体瓶颈环节具垄断优势，支撑美国终端产品。
@wheelertweets 日本在上游瓶颈环节处于领先地位。其中许多是独家供应商。例如住友电木(InP substrates)的磷化铟衬底、铠侠(Kioxia)的存储制造、爱德万测试(Advantest)的HBM良率、揖斐电(Ibiden)的IC载板等。最终产品通常是你所见的美国公司，如$NVDA的GPU。
英文原文
@wheelertweets Japan are leaders in bottlenecks upstream. And many are sole source. For example Sumitomo for InP substrates, Kioxia for memory manufacturing, Advantest for HBM yield, Ibiden for IC substrates, etc. The end product are usually American companies like $NVDA GPUs that you see
-
日本7.2万亿刺激计划利好AI半导体，MU获补贴建厂。
是的，最大的顺风因素是日本股市的支出。不久前提到，刺激计划包括“为AI、半导体和量子技术等17个优先领域提供7.2万亿日元”。$MU正在广岛建设一座价值96亿美元的HBM工厂，但这更多是间接的（例如，他们仍从日本获得33亿美元的补贴）。
英文原文
Yeah biggest tailwind is spend for Japanese stocks. Not too long ago they detailed that the stimulus includes “¥7.2 trillion for 17 priority sectors like AI, semiconductors, and quantum tech.” $MU is setting up a $9.6B HBM plant in Hiroshima but it’s more indirect (eg. They still get $3.3B in subsidies from Japan).
-
日韩股市大涨，AI板块因资本支出增加而反弹。
日本日经225指数(Nikkei 225)飙升逾5%，创下历史新高。 韩国KOSPI指数今日也上涨4.2%，其中存储芯片如SK海力士(SK Hynix)上涨5.4%。 市场认为以刺激政策为主的日本政策总体利好，并为从$MU到Advantest的股票增加了流动性。 尽管如此，这一政策具有细微差别：像比特币这样的资产更青睐“印钞机”（货币刺激），而非政府支出的财政刺激，因为后者可能导致债券收益率飙升。 我们开始看到由资本支出(capex)增加引发的特定板块（尤其是AI交易）反弹，而其他板块则处于熊市。
英文原文
Japan’s Nikkei 225 skyrockets over 5% to hit record high. Korea’s KOSPI is also up 4.2% today with memory like SK Hynix up 5.4%. A stimulus heavy Japan was seen as net positive and increases liquidity for stocks from $MU to Advantest. Despite this, this policy is nuanced as assets like Bitcoin love the money printer (monetary stimulus) vs. fiscal stimulus (government spend), when there’s a potential spike in bond yields. We’re starting to see sector specific rallies (especially the AI trade) from increased capex spend, but a bear market in others.
-
披萨订单作为算法信号，反映委内瑞拉局势对AVAV和NOC的影响。
不过这是真实的流量，因为恶搞者确实会下大量披萨订单，并造成等待自取人群的积压。只是我觉得没人试过，毕竟浪费披萨。但对于 $AVAV 到 $NOC 而言，披萨订单是算法用于委内瑞拉的合法信号。
英文原文
It's live traffic though since a troll would be actually placing a lot of Pizza orders + creating backlog from people waiting for pickup. Just dont think people have tried since it's a waste of Pizza. But for $AVAV to $NOC, Pizza orders are a legitimate signal algos used for Venezuela.
-
对比供应链瓶颈巨头与未获大厂认证的投机性小盘股。
@B38B37 这些都是供应链上已知的大型瓶颈供应商，如 $TSM 或 $GLW。$POET 是一家市值仅数亿美元的投机性标的，尚未获得超大规模云服务商（Hyperscaler）的资质，因此目前不完全符合该描述。但它属于一种“登月式”（Moonshot）的选股。
英文原文
@B38B37 These are all big known bottleneck suppliers on the supply chain $TSM or $GLW. $POET is a few hundred million MC speculative pick without hyperscaler qualification, so they don’t quite fit that description yet. But it is a moonshot type pic
-
马斯克指出Optimus三大瓶颈：智能、手部及制造，暗示美韩主导核心部件。
$TSLA Optimus 瓶颈 TLDR - Elon Musk 采访： “人形机器人只有三件难事”： 1. 现实世界智能（例如 Dojo，$NVDA，三星机电） 2. 手部（~ 可能是 $VPG） 3. 规模化制造（~ 浙江华正，宁波拓普，Everwin） 关于手部： Elon 声称 $TSLA Optimus 的人手自由度比其他机器人更多。 特斯拉“设计了定制执行器、齿轮、功率电子、控制、传感器。这方面没有供应链。” Elon 在采访中基本上强调手部是机器人最涉及知识产权且最复杂的部分。Elon 重申特斯拉解决了中国其他人形机器人今天尚未解决的问题。 这使得西方制造商参与这一特定 BOM 部分的可能性略微增加（相比此前报道的中国供应链）。 关于现实世界智能方面： Elon 聚焦于：“惯性测量、GPS 信号、其他数据，将其与视频结合，主要是视频，然后输出控制命令” - 训练 ~ Cortex / Dojo 供应商，$NVDA。 - 视频与视觉：三星机电？ - IMU + GPS：$STM 或博世？ 规模化制造： 正如朝鲜日报报道，身体部分可能来自中国供应链。 从推理来看似乎是： 大脑 + 高端组件 = 美国/韩国 身体是中国（三花，拓普等）。 但再次强调，这并未确认，只是可能的供应链映射。 无论如何，Elon 为投资者指出了人形机器人量产过程中的三个瓶颈投资方向。
英文原文
$TSLA Optimus Bottleneck TLDR - Elon Musk Interview: "There's only three hard things for humanoid robots": 1. Real World Intelligence (eg. Dojo, $NVDA, Samsung Electro-Mechanics) 2. The Hand (~ possibly $VPG) 3. Scale Manufacturing ( ~ Zhejiang Huazheng, Ningbo TuoFu, Everwin) For the hand: Elon claimed that $TSLA Optimus has the more degrees of freedom with a human hand than other robotic humanoid. Tesla "designed custom actuators, gears power electronics, controls, sensors. There is no supply chain for this." Elon basically focused on the hand is the most IP-sensitive and complex part of the robot during the interview. Elon reiterated that Tesla solved issues with the hand that no other humanoids in China have today. This makes things slightly more probable that Western manufacturers are involved (compared to Chinese supply chains from earlier reports), with this part of this specific BOM. For the Real World Intelligence Aspect: Elon focused on: "Inertial measurements, GPS signals, other data, combining that with video, primarily video, and then outputting the control commands" - Training ~ Cortex / Dojo suppliers, $NVDA. - Video & Vision: Samsung Electro-Mechanics? - IMUs + GPS: $STM or Bosch? Scale Manufacturing: Likely Chinese supply chains as reported, by Chosun for the body. From inference looks like: Brain + high-end components = US/Korean Body is Chinese (Sanhua, Tupou, etc). But again, this is not confirmed, just possible supply chain mapping. Regardless, Elon gave investors three bottlenecks to invest in for the humanoid ramp.
-
梳理AI半导体供应链瓶颈，看好存储、光子及台积电，指出中游为定价关键。
以下是我对 Semivision 瓶颈总结的 TLDR（太长不看版）及其映射到投资框架的内容： HBM（高带宽内存）： HBM4 (16Hi) - 三星，SK 海力士，$MU HBF（高带宽闪存）- $SNDK，Kioxia Base Die（基础晶圆）- $TSM，三星（内部） CPO（共封装光学）/光子学： 玻璃基板：$GLW，$INTC，Ibiden 光学：$LITE，$AVGO，$COHR，$MRVL 电力传输： 电压：$MPWR，$VICR 散热：$VRT，$NVT，$MOD 电网：$ETN，$SBGSY SiC/GaN（碳化硅/氮化镓）：$ON，$IFNNY 机架：$APH N2 产能： $TSM，$AMD（数据中心先行者），$QLCM，联发科，$NVDA 先进封装： 良率：$CAMT，$ONTO，$KLAC OSATs（外包半导体组装与测试）：$AMKR $BESIY，Disco Semivision 的总结： 1. “先进封装产能与良率控制” 2. “HBM 生态系统协调” 3. “电力传输创新（SiC、GaN PMICs、机架级电源架构）” 4. “CPO/光子学集成能力” 5. “数据中心基础设施作为半导体收入实现的‘隐性限制因素’” 正如你可能知道的，我对上述的存储/光子学领域（如 SK 海力士/三星，$SNDK，$MU，$LITE）以及 $TSM 最为看好。 我可能会稍微向上游延伸，比如 $AXTI（磷化铟前驱体），但中游玩家是瓶颈所在，并掌控大部分定价权。 但为了简化普通散户投资者的理解，我添加了相关公司的评论。
英文原文
Here's my TLDR + mapped into investment framework from Semivision bottleneck summary: HBM: HBM4 (16Hi) - Samsung, Sk Hynix, $MU HBF - $SNDK, Kioxia Base Die - $TSM, Samsung (internal) CPO/photonics; Glass Substrate: $GLW, $INTC, Ibiden Optical: $LITE, $AVGO, $COHR, $MRVL Power Delivery: Volatage: $MPWR, $VICR Thermal: $VRT, $NVT, $MOD Grid: $ETN, $SBGSY SiC/GaN: $ON, $IFNNY Rack: $APH N2 Volume: $TSM , $AMD (First mover dc), $QLCM, Mediatek, $NVDA Advanced Packaging: Yield: $CAMT, $ONTO, $KLAC OSATs: $AMKR $BESIY, Disco Semivision's summary: 1. "advanced packaging capacity and yield control" 2. "HBM ecosystem coordination" 3. "power delivery innovation (SiC, GaN PMICs, rack-level power architectures)" 4. "CPO/photonics integration capability" 5. "data center infrastructure as a “hidden limiter” to semiconductor revenue realization" As you probably know, I'm probably most bullish on memory/photonics like Sk Hynix/Samsung, $SNDK, $MU, $LITE above. And $TSM. I probably go a bit more upstream like $AXTI for InP precursors, but midstream players are the chokepoint + control most of the pricing. But just added commentary of related companies to topics to make things simpler for the regular retail investor.
-
对比法美AI投入，指出法国资金规模远不及美巨头。
@ticker_thoughts 我确实关注了。法国2030年3000万欧元的投入，与$GOOGL和$AMZN约3800亿美元的年度资本支出(capex)相比，对法国AI领域的影响肯定微乎其微。他们并没有打破刻板印象。https://t.co/qXdcTRvmVd
英文原文
@ticker_thoughts I did. That whole 30M euros for 2030 is definitely going to make a large impact on France’s AI sector compared to $GOOGL and $AMZN ~$380B 1 year capex spend. They’re not really breaking stereotypes here. https://t.co/qXdcTRvmVd
-
日本5500亿投资计划聚焦半导体散热、数据中心能源及石油终端。
一份新发布的报告显示日本正在投资： 5500亿美元（86万亿日元）投入多个领域。 第一轮投资流向以下三个方向： 1. 半导体用合成钻石（AI芯片散热器） - J-TEC Corporation (3446.T) - $COHR - 将钻石衬底引入光学领域（可能是合作伙伴）。他们于2026年1月推出了可键合钻石。Element Six可能扩大钻石晶圆生产 -> $COHR将其键合到AI处理器上。 - EDP (7794.T) - Element Six 是该领域可能的赢家。文中提到了“合成钻石生产厂”，但未点名这些公司。从技术逻辑上看，他们是赢家。 2. 数据中心燃气发电 - SoftBank（明确点名） - GE Vernova $GEV - Quanta Services $PWR - Hitachi (6501.T) 3. 深水石油终端 - EPD $EDP - ENEOS (5020.T) - Idemitsu Kosan (5019.T) 抱歉还没找到文章的英文版本（且翻译工具失效）。 第一阶段为400-450亿美元（约6-7万亿日元）。
英文原文
A new release shows Japan is investing: $550B USD (86 Trillion Yen) into several sectors. The first round of investments go into these three: 1. Synthetic Diamonds for Semis (heat sink for AI chips) - J-TEC Corporation (3446.T) - $COHR - diamond substrates into optical (likely partner). They launched bondable diamonds in Jan 2026. Likely Element Six grows Diamondond Wafer -> $COHR bonds it to AI processors. - EDP (7794.T) - Element Six Are the likely winners for this sector. The text mentions a "synthetic diamond production plant," but it does not name these companies. They are the logical technical winners. 2. Gas Power for Data Centers - SoftBank (Explicitly named) - GE Vernova $GEV - Quanta Services $PWR - Hitachi (6501.T) 3. Deepwater Oil Terminals - EPD $EDP - ENEOS (5020.T) - Idemitsu Kosan (5019.T) Sorry couldn't find English version of the article yet (and translation tools were off). Phase 1 is $40–45 Billion (approx. 6–7 Trillion Yen).
-
超大规模云厂商资本支出激增导致GPU供应紧张，利好新云服务商。
黄仁勋在CNBC表示：“六年前销售的GPU价格正在上涨。” GPU折旧曾是$NBIS、$IREN和$CRWV最大的担忧。 但现在，它们拥有： - 超大规模云厂商资本支出增加带来的顺风 - 全面的需求增长 - GPU折旧担忧缓解。 所有顺风同时利好新云服务商(Neoclouds)。 （引用内容：根据3Fourteen Research：2月份主要云服务的GPU可用性大幅下降。随着$META、$AMZN、$GOOGL大幅增加基础设施资本支出，主要赢家是新云服务商：$IREN、$NBIS和$CRWV是三家提供AI云服务的企业。随着Opus 4.6等新AI模型的推出，可能出现新一轮需求冲击。从B200等新一代到旧型号，需求全面增长。这是产能限制累积效应的可视化体现。）
英文原文
Jensen Huang on CNBC: "GPUs sold six years ago are increasing in price." GPU depreciation was the largest concern for $NBIS, $IREN, and $CRWV. But now, they have: - Tailwinds from increased capex spend across hyperscalers - Increased demand across the board - GPU depreciation fears eased. Every tailwind just hit at once for Neoclouds.
-
云厂商GPU供应紧缺，Neoclouds因需求冲击受益。
根据3Fourteen Research的数据：2月份主要云服务提供商的GPU可用性大幅下降。 与此同时，$META、$AMZN和$GOOGL大幅增加了基础设施资本支出。 主要的受益者是Neoclouds（新型云服务商）： $IREN、$NBIS和$CRWV是拥有AI云服务业务的三家。 随着Opus 4.6等新AI模型的发布，很可能出现了一波新的需求冲击。 这种需求增长是全方位的，涵盖了B200等新一代芯片以及旧型号。 这是产能限制累积效应的直观体现。
英文原文
According to 3Fourteen Research: GPU availability across major cloud services have dropped immensely in Feb. This comes as $META, $AMZN, $GOOGL have increased infrastructure capex spend immensely. The major victors, Neoclouds: $IREN, $NBIS, and $CRWV are the three with AI cloud offerings. There's likely been a new wave of demand shock with new AI models coming out from Opus 4.6 and others. And the demand increase is across the board with newer gens like B200, and older models. This is the cumulative effect of capacity constraints, now visualized.
-
超大规模云厂商AI资本支出激增，利好上游半导体供应链，建议坚守基本面做多。
市场正遭受AI资本支出(AI capex)冲击的洗礼。 这实质上是超大规模云服务商(hyperscalers)与外国之间的一场AI军备竞赛。 2025年至2026年间，资本支出(capex)增长如下： $GOOGL：从910亿-930亿美元增至1750亿-1850亿美元 $AMZN：从1250亿美元增至约2000亿美元 $META：从720亿美元增至1150亿-1350亿美元 $MSFT：从1100亿美元增至1500亿美元 $TSLA：从85亿美元增至200亿美元以上 当谷歌表示其年度支出（用于数据中心、TPU、基础设施）从去年的920亿美元激增至1850亿美元时，这一规模是巨大的。 $AMZN等公司的资本支出预测涵盖机器人到太空领域，情况不一，但同一家内存公司和半导体公司实际上能从中双重受益。 市场可能不欣赏$GOOGL如此巨大的支出，但上游AI供应链 certainly 乐见其成。 尤其是那些最明显受益的公司，例如： 专用集成电路(ASICs)：$AVGO（谷歌TPU、Meta MTIA等） 内存：SK海力士、三星、美光 $TSM（没有台积电你造不出“上帝”） $NVDA（训练） 每个建设项目中都有隐藏的受益者（例如$MSFT带动的$AAOI，$MRVL带动的$POET，$GOOGL带动的$LITE）。但有时，像内存或$TSM这样显而易见的“傻瓜式”标的反而就是最好的选择。 判断AI交易是否持续的最大信号是超大规模云服务商的支出： 而他们刚刚彻底超出了所有预期。 在AI建设浪潮中保持做多，基本面比宏观环境和短期流动性冲击更重要。
英文原文
Markets have been hit with an AI capex shock. This is effectively an AI arms race between hyperscalers and foreign nations. Between 2025 to 2026, capex has increased from: $GOOGL: $91B - $93B to $175B – $185B $AMZN: $125B to ~$200B $META: $72B to $115B-$135B $MSFT: $110B to $150B $TSLA: $8.5B to $20B+ When Google says they are spending $185 Billion in 1 year (on DCs, TPUs, infrastructure), from $92B the previous year, that is enormous. Capex projections like $AMZN are mixed from robotics to space, but the same memory companies and semiconductors actually benefit twice. Markets might not appreciate $GOOGL spending that much but AI supply chains upstream certainly do. Especially the most obvious ones that benefit from most such as: ASICs: $AVGO (Google TPUs, Meta MTIA, etc.) Memory: SK Hynix, Samsung, Micron $TSM (You cannot build God without TSM) $NVDA (training) There's hidden beneficiaries in each buildout (eg. $AAOI with $MSFT to $POET with $MRVL, $LITE with $GOOGL). But sometimes, the braindead obvious plays like memory or $TSM are are just the best. The biggest signal of whether the AI trade continues is hyperscaler spending: And they just blew all expectations out of the room. Stay long in the AI buildout, fundamentals matter more than macro and short term liquidity shock.
-
特斯拉Optimus或分中美双供应链，高端用VPG，量产用中国。
情况比这更微妙。根据我的研究，为了实现 $20k Optimus 的廉价大规模生产，特斯拉可能正在设计两种不同的型号。大规模生产的型号可能仅采用中国供应链。然而，面向大众市场的中国传感器尚未完全匹配美国高成本、超专业化的精密传感技术。$VPG 可能被用于高端或高精度的 Optimus 版本，例如特斯拉超级工厂级别/工业级版本。因此，首批数千台特斯拉人形机器人可能会使用 $VPG。这仅仅是关于可能的大规模生产/消费级版本，这些版本很可能为了追求最低成本而交由中国制造商生产。这也发出了中国对美国本土机器人供应链构成威胁的信号。
英文原文
It’s a lot more nuanced than that. From my research, for cheap mass production $20k Optimus, it’s possible Tesla is designing two different models. The mass production models may just be Chinese supply chains. However, the mass-market Chinese sensors haven't quite matched US high cost ultra specialized precision sensing. $VPG may be used for high-end or high precision Optimus versions, like the Tesla giga factory grade/industrial versions. So the first thousands of Tesla humanoids may use $VPG. This was just talking about possible mass production/consumer grade versions that likely went to China manufacturers for the lowest cost possible. This also signals the threat China has on Made in America robotics supply chains.
-
特斯拉Optimus量产供应链或在中国，成本大幅降低威胁美国制造。
看起来 $TSLA Optimus 的大规模量产供应链将在中国制造。 - 浙江华正（热管理） - 宁波拓普（执行器） - 埃文（传感器） 本周的一份新报告将这些公司列为潜在供应商。 据报道，由于中国的补贴和低成本制造，Optimus 的成本从超过 10 万美元降至 4.5 万美元以下，“美国制造”的机器人供应链可能受到威胁。
英文原文
It looks like $TSLA Optimus supply chains for mass production will be manufactured in China. - Zhejiang Huazheng (thermal control) - Ningbo TuoFu (actuators) - Everwin (sensors). Were named as likely suppliers from a new report this week. “Made in America” robotics supply chains are likely threatened as China subsidies and low cost manufacturing reportedly brings Optimus costs from over $100K+ down to <$45k.
-
市场因对美联储政策感知的恐慌而抛售，但这可能是自4月以来的最大机会。
白银暴跌。比特币暴跌。从 $CRDO 到 $HOOD 的股票都在暴跌。 这恰好发生在与 2021 年相同的时间点，图表看起来基本一样。 价值百万的问题是，这次是一样的吗？ 从结构上讲，是的。 它看起来令人毛骨悚然地相似，市场正在抢跑 2021 年的流动性枯竭。 上一次的触发因素是量化紧缩 (Quantitative Tightening, QT)，在鲍威尔表示美联储将加速“缩减购债 (Tapering)”和 QT 以抑制通胀后。随后他们在 2022 年 1 月以超出任何人想象的速度开始了 QT。 今年，有中期选举。特朗普希望降低利率并增加流动性，以确保 11 月获胜。而且，最重要的是，他非常关心股市。 就在 2025 年 12 月，也就是两个月前，鲍威尔悄悄宣布恢复购买资产，即隐性量化宽松 (stealth QE)。 这一次，这看起来像是表面文章和自证预言。 然而，市场运作基于越南战争的闪回和创伤后应激障碍 (PTSD)，主要鉴于新任美联储主席 (Fed chair) 的样子。 如果新任美联储主席按照特朗普的政策行事（人们可能会预期如此），这可能是自 4 月以来最大的机会之一。 这一次，市场抛售是因为“感知”而非任何正式的政策变化。 历史不会重演，但会押韵。
英文原文
Silver is crashing. Bitcoin is crashing. Stocks from $CRDO to $HOOD are crashing. Happened to be around the same time as 2021 and chart looks basically the same. The million dollar question, is it the same this time? Structurally, yes. It looks eerily similar with markets frontrunning a liquidity drain back in 2021. The trigger last time was Quantitative Tightening after Powell stated the Fed would accelerate "Tapering" and QT for inflation to come down. And then they began QT faster than anyone imagined in Jan 2022. This year, there's Midterms. Trump wants lower rates and higher liquidity to ensure a win in November. And, most of all, he cares deeply about the stock market. And just 2 months ago in December 2025, Powell gave a stealth QE announcement to resume purchasing assets. This time, it looks like optics and a self-fulfilling prophecy. However, the market operates off Vietnam flashbacks and PTSD, mainly given the look of the new Fed chair. If the new Fed chair operates in line with Trump's policies (which one might expect), this could be one of the biggest opportunities since April. This time, markets are selling off because of "perception" rather than any formal policy change. History doesn't repeat itself but it does rhyme.
-
比特币作为风险管理工具，及特斯拉垂直整合对VPG的供应链风险。
@Bonek2801 出于风险管理目的。我宁愿在约 7 万美元价位持有 50% 的比特币，也不愿持有现金以降低波动性。此外，$VPG 的下行风险在于 $TSLA 自行制造传感器并实现垂直整合，从而将其排除在供应链之外。$GOOGL 与 $LITE 在光学共封装(CPO)方面的看空逻辑也类似。
英文原文
@Bonek2801 Risk management purposes. I’d rather hold 50% in Bitcoin around $70K than sit on cash to lower volatility. Also the downside risk of $VPG is $TSLA creating their own sensors and vertically integrating it away. Similar bear case with $GOOGL and OCS with $LITE though.
-
联发科AI收入预测远低于谷歌资本支出，谷歌大幅上调支出聚焦推理成本。
Counterpoint Research 指出，联发科(Mediatek)对2026年云端AI/ASIC(专用集成电路)收入约10亿美元的预测，与 $GOOGL 的资本支出(capex)以及 $AVGO 目前的营收相比，简直是沧海一粟。 但谷歌确实带来了震撼，将其资本支出预测上调了50%以上，并表示将专注于推理(inference)成本。 所以谁知道呢
英文原文
Counterpoint is Mediatek's ~$1B projections for Cloud AI/ASIC revenue for 2026 was a drop in the bucket compared to $GOOGL capex spend and what $AVGO currently brings in. But Google did deliver a shock upping its capex projections by 50%+ and said they would focus on inference costs. So who knows
-
谷歌巨额资本支出验证联发科在AI供应链中的地位。
联发科是 $GOOGL TPU 生态系统（TPU v7e 和 v8e）中有趣的一部分。 以及 $NVDA Blackwell（例如 GB10 Grace 设计）。 他们曾在 2025 年做出雄心勃勃的预测，认为其将占据数据中心 ASIC 市场 10-15% 的份额，与 $AVGO 相比。 但鉴于谷歌目前 1750-1850 亿美元的资本支出，看来他们的说法正在得到验证。 披露：无持仓。
英文原文
Mediatek is an interesting part of $GOOGL TPU ecosystem (TPU v7e and v8e). And $NVDA Blackwell (eg. GB10 Grace design). They did make ambitious projections of 10–15% of the DC ASIC market back in 2025 vs. $AVGO. But given Google's $175-185B capex spend today, looks like their claims are getting validated. Disclosure: Have no positions.
-
TPU/DC供应链庞大，难点在于识别BOM及相对市值受益最大的公司。
@PhotonCap 是的，TPU/DC（数据中心）的供应链规模巨大，$GLW 以及其他受益于 $META 和 $GOOGL 资本支出的公司都是赢家。主要问题（也是最难的部分）在于识别BOM（物料清单），并找出相对于当前市值而言受益最大的公司。
英文原文
@PhotonCap Yes, the supply chain for TPU/DC is massive with $GLW and other beneficiaries from $META and $GOOGL capex. Main thing (hardest part) is identifying BOM and who benefits the most from this relative to current MC.
-
谷歌转向TPU利好相关托管商，NBIS非其受益者。
@lawbrax 这对通过 Fluidstack 与 $GOOGL 相关的托管设施(Colo)公司，如新云(NeoCloud)领域的 $CIFR 和 $WULF 是利好消息。 看起来谷歌主要转向张量处理单元(TPU)了。我认为 $NBIS 并非谷歌的受益者，但他们曾是 $META 的受益者。
英文原文
@lawbrax This is positive for $GOOGL related colo companies through Fluidstack like $CIFR and $WULF in neocloud sector. Looks like Google is primarily shifting to TPUs now. Don't think $NBIS is a beneficiary of Google but they were with $META.
-
AI供应链企业欢迎谷歌巨额资本支出增加
@beauty_oe 市场对 $GOOGL 的巨额支出感到厌恶。然而，受益于5000亿美元资本支出(CAPEX)增加的AI供应链企业却对此表示热烈欢迎！
英文原文
@beauty_oe 市場は $GOOGL の巨額支出を嫌気しています。しかし、500億ドルの設備投資（CAPEX）増額の恩恵を受けるAIサプライチェーン企業たちは、これを大歓迎しています！
-
谷歌巨额资本支出加速AI基建，利好相关供应链。
$GOOGL 公布了财报，其资本支出(CapEX)规模巨大。 1750亿-1850亿美元 vs. 1200亿美元。 这对AI基础设施建设是利好。 跟随资金流向下游： - $AVGO、联发科、$TSM（设计/晶圆代工） - SK海力士、三星、$MU、$SNDK（存储） - $ANET、$LITE、$COHR、$VRT（网络、光子学、能源） 等等，因为谷歌表示支出主要在于：“包括服务器、数据中心和网络在内的AI基础设施”。 从 $META 到 $GOOGL 的超大规模云服务商正在大幅增加资本支出。 这仅仅表明： AI基础设施建设正在加速，并由全球最盈利的公司资助。 做多AI供应链。
英文原文
$GOOGL reported earnings and their CapEX spend was enormous. $175-$185 Billion vs. $120 Billion. This is bullish for AI buildout. Follow the money flow down to: - $AVGO, Mediatek, $TSM (design / foundry) - Sk hynix, Samsung, $MU, $SNDK (memory) - $ANET, $LITE, $COHR, $VRT (network, photonics, energy) etc. as Google said spend was: "primarily on AI infrastructure including servers, data centers and networking" Hyperscalers from $META to $GOOGL are increasing their capex spend enormously. This just goes to show: AI buildout is accelerating and is funded by the most profitable companies in the world. Long AI supply chains.
-
美国能源供应链落后致电网容量不足，制约AI建设。
我持有一个小仓位，仅仅是为了支持这一事业。 我只是不喜欢看到美国能源供应链远远落后于其他国家，因为美国太阳能等产业被做空至崩溃长达太久。许多有前景的公司在美国本土重建过程中破产。如今，美国因此缺乏用于AI建设的电网容量。
英文原文
I own a small position, just to support the cause. I just dislike it when American energy supply chains are so far behind other nations because US solar and others were shorted to the ground for so long. Many promising companies went bankrupt during the buildout + Made in America restoring process. And now America lacks grid capacity for AI buildout because of it
-
三大巨头转向AI致旧DRAM市场真空，南亚科面临中国供应链竞争威胁。
南亚科(Nanya)受益，因为“三大巨头”向产业链上游转移，并切断了用于HBM3e/HBM4/DDR5的旧式DRAM（如DDR4）供应。 南亚科一直独自享受旧市场。因此，针对中国特定供应链的南亚科现在面临可信的中期威胁。 SK海力士(Sk Hynix)和美光(Micron)并不在意，因为他们已有意放弃该市场，转向利润率更高的AI领域。
英文原文
Nayna benefited because big 3 moved up the food chain and cut off legacy DRAM like DDR4 for HBM3e/HBM4/DDR5. Nayna's just been enjoying the older market by itself. So now there's a credible mid-term threat for Nanya in specific for China specific supply chains. Sk Hynix and Micron don't care since they knowingly abandoned that market for higher margin AI segments already
-
驳斥媒体误导，指出中美存储禁令不影响美光等巨头AI业务。
在《日经亚洲》新文章发布后，围绕 $MU、SK海力士和存储芯片领域出现了大量虚假信息。我预料到会发生这种情况，因为这篇文章写得极差且具有误导性，而存储行业其实非常微妙。这对美光和美光以及SK海力士几乎没有实质性影响。长鑫存储(CXMT)和中微半导体(YMTC)已被美国政府列入黑名单，无法进入中国以外的供应链。这纯属国内事务，文章却写成耸人听闻的抨击之作。例如：“中国两大主要存储制造商将全球供应短缺视为机遇……” YMTC被全面禁止（美国实体清单）——国防部1260H条款名单（中国军事公司）。CXMT也在国防部1260H名单中，属于军事关联名单。美国商务部很可能对该企业实施更多出口管制。最大的影响其实是对南亚+Nanya和旺达+Winbond这类“填补空白”的传统存储公司，这才是文章应聚焦的重点。但文章和那些想寻找逢低买入机会的人为了点击量，通过引入美光等公司散布虚假信息。细微差别在于：美光、SK海力士和三星已基本清除了来自中国的市场需求（例如美光约10%，SK海力士约15-18%），并将DRAM产能转向HBM3e和HBM4。只有三星因其在移动/消费设备领域的份额受到轻微影响。这对超大规模AI存储交易没有任何影响。
英文原文
There's a lot of disinformation around $MU, SK Hynix, and Memory after the new Nikkei Asia piece. Expected this to happen, since this is so badly written and misleading when the memory industry is so nuanced. This has close to 0 material impact on Micron and Sk Hynix. CXMT and YMTC are blacklisted by US government and supply chains outside of China. This is domestic only and is written as a sensational hit piece. eg. "China’s two major memory manufacturers view the global supply shortage as an opportunity..." YMTC is fully banned (US Entity List) - DoD Section 1260H list (Chinese Military Companies) CXMT - DoD Section 1260H list. Military-linked list. The U.S. Commerce Department is likely to slap more export controls on this firm. The biggest impact it would have is on "legacy fill in the void" type companies like Nanya + Winbond, which should have been the focus. But the article and people who want a dip-buying opportunity decided to clickbait/spread disinformation by throwing in Micron and others for views. Naunce is Micron, Sk Hynix, and Samsung have largely hollowed out their demand from China (eg. ~10% micron, ~15-18% Sk hynix), and shifted dram capacity to hbm3e and hbm4. Only Samsung is slightly affected given their share with mobile/consumer devices. There is no effect on the hyperscaler AI memory trade.