$VRT
相关推文
按时间倒序
-
再排查 NVDA 架构变化下的台湾标的
@Leo66Gao 对，TWSE: 3324、TWSE: 3017、TWSE: 8996、TWSE: 2308、$VRT 这些都是我标出来的。 但我今天还想多做些尽调，以防 $NVDA 的架构变化会把其中一两个名字设计出去。
英文原文
@Leo66Gao Yeah TWSE: 3324, TWSE: 3017, TWSE: 8996, TWSE: 2308, $VRT, and others were the ones I flagged. But I wanted to do more DD today, just in case the $NVDA architecture changes might have designed or two out.
-
NVDA 推进 800V
$NVDA 正在积极推进 800V，这对今天来说是一个相当大的信号。 “2026 年 4 月 22 日，NVIDIA 已经与主要韩国电力设备公司展开讨论，探索基于约 800V 直流（DC）系统设计数据中心基础设施。” 之前提到的 Nvidia 合作伙伴包括：Infineon、$TXN、$STM、$NVTS、Delta、Flex、$ETN、Schneider 和 $VRT。 早就知道他们在往这个方向推了，但……看起来现在真的是 Nvidia 在主导。 我个人对 $NVTS 的敞口很小，因为它们在 800V DC 的 SiC / GaN IC 方面暴露很高。 我知道其他人更喜欢 $VICR 作为敞口。
英文原文
$NVDA actively pushing to 800V, and today was a pretty big signal. "On April 22, 2026, NVIDIA has initiated discussions with major South Korean power equipment companies to explore designing DC infra based on an approximately 800V direct current (DC) system" Previous named Nvidia partners include: Infineon, $TXN, $STM, $NVTS, Delta, Flex, $ETN, Schneider, and $VRT. It's been known they're been pushing for this but... looks like Nvidia is really driving it. Personally very small exposure to $NVTS since they're they have a high exposure for SiC/GaN ICs for 800V DCs. I know others really liked $VICR for exposure.
-
列出 AI Displacement 等权组合年初至今回报，认为自己无意中做出了一个很强的 ETF。
我看了一下书签数量，才意识到： 我低调地搞出了一个很能打的 ETF？ AI Displacement 等权组合年初至今： $AXTI：+191.53% $AAOI：+144.47% $SNDK：+140.38% $SOI：+114.3% $LITE：+61.22% $AEHR：+60.97% $BE：+56.56% $VRT：+47.42% Samsung：+42.8% $TER：+37.99% $MU：+35.1% SK 海力士：+34.42% $NBIS：+25.57% $COHR：+24.92% Mediatek：+17.01% $INTC：+16.23% $ASML：+15.63% Advantest：+11.78% $TSM：+5.85% $COPX：+4.56% $TSEM：+2.44% $MRVL：-1.71% $NVDA：-4.55% $AVGO：-7.32% 这只是我脑子里第一时间想到的名字。 我个人持有其中很多，也没持有另外一些，比如 $NVDA 或 $BE，但我还是把它们放进来了。 不过老实说，如果我现在不是在主动管理仓位集中度，我会很愿意把这些名字等权配置。 我预计 AI Displacement ETF 还会继续上涨： 因为这些公司都是人工智能扩展算力和推理最主要的受益者。
英文原文
I realized by the amount of bookmarks. I lowkey dropped a banger ETF? AI Displacement Equal Weighted YTD: $AXTI: +191.53% $AAOI: +144.47% $SNDK: +140.38% $SOI: +114.3% $LITE: +61.22% $AEHR: +60.97% $BE: +56.56% $VRT: +47.42% Samsung: +42.8% $TER: +37.99% $MU: +35.1% Sk Hynix: +34.42% $NBIS: +25.57% $COHR: +24.92% Mediatek: +17.01% $INTC: +16.23% $ASML: +15.63% Advantest: +11.78% $TSM: +5.85% $COPX: +4.56% $TSEM: +2.44% $MRVL: -1.71% $NVDA: -4.55% $AVGO: -7.32% These are just the first names that came to my head. I own a lot of these personally (and don't own others like $NVDA or $BE ) but included them anyway. But honestly, I'd be happy to equal weight all these names if I weren't actively managing my portfolio concentrations. I expect the AI Displacement ETF to keep rising: As they're the largest beneficiaries of scaling compute and inference for artificial intelligence.
-
认为 AI 已引发不可逆的岗位替代，投资相关基础设施和瓶颈是对冲失业冲击的主要方式。
这个消息真的很令人心碎： $META 裁员 20% $ORCL 裁员 $AMZN 长期将有 60 万名工人被机器人和 AI 取代。 这是一种反乌托邦未来。 企业在没有人工劳动成本的情况下，仍然能赚创纪录的利润。 唯一能从中受益的办法： 把 AI 当成对冲工具去投资。 未来几年，想逃离 AI 造成的永久底层阶级，感觉最主要的办法就是这样。 AI 产生的股东权益回报，将流向股东。 而那些没投资股票、只能靠工资过活的人之间的差距会继续拉大。 这不是未来，而是已经在发生。 - Opus 4.6 已经足够替代今天大多数软件工程师。 - Waymo 已经开始在旧金山这样的地方替代出租车司机。 - 我们知道 $TSLA 的人形机器人也快来了，因为它们在中国已经普及。 这件事现在就在发生。 伊朗的扰动只是 AI 建设加速过程中的短期事件。 AI 已经到达拐点，而且似乎不可避免。 你已经看到美国就业修正接近减少 100 万，这非常惊人。 而且我们正在看到，新一代 LLM 正在由它们前一代模型自己构建，AI 正在逼近奇点（AI 递归增长）。 如果你要投资那些运行 AI 所需的算力和硬件： 数据中心 / 电力 / 电网板块： $NBIS、$XLU、$VRT、$BE 光子板块，负责扩展 AI： $LITE、$COHR、$AAOI、$TSEM 半导体板块，负责芯片： $NVDA、$TSM、$ASML、$INTC 存储板块，负责芯片： $MU、$SNDK、SK 海力士、三星 ASIC，负责超大规模云 AI 推理： $AVGO、$MRVL、Mediatek 良率 / 测试板块，确保芯片能工作： $TER、$AEHR、Advantest 以及 AI 所需的原材料或基板： $AXTI、$COPX、$SOI 还有很多其他公司，都在成为对抗大规模 AI 失业的单一、最大对冲工具。 谁掌握了算力的生产资料（瓶颈、材料、数据中心）： 谁就掌握了 AI 的未来。
英文原文
The news is pretty heartbreaking: $META 20% layoffs $ORCL layoffs $AMZN 600,000 workers long term layoffs as they get replaced by robotics and AI. This is a dystopian future. Companies end up with record profits, without the cost of human labor. The only way to benefit: Investing in AI as a hedge. The next few years feels like the main way to escape the permanent underclass, caused by AI displacement. The return on equity derived from AI will go to the shareholders. While the gap between those who live paycheck to paycheck, not invested in stocks. Will continue to grow. This is not the future. - Opus 4.6 is good enough to replace most software engineers today. - Waymo has started to replace taxi drivers in places like SF today. - We know $TSLA Humanoids are coming next as they’re widespread in China, today. This is happening now. Disruptions in Iran are only temporary to the accelerating AI buildout. AI has hit the inflection point, and looks inevitable. You’re already seeing US job revisions down close to 1 Million, which is staggering. And we’re seeing the newest LLMs be built by their previous models, as AI approaches the singularity (AI led recursive growth). Investing in where the compute and hardware needed to run the AI: From the datacenter/power/grid sector: $NBIS, $XLU, $VRT, $BE Photonics sector needed to scale AI: $LITE, $COHR, $AAOI, $TSEM Semi sector needed for the chips: $NVDA, $TSM, $ASML, $INTC Memory sector for the chips: $MU, $SNDK, SK Hynix, Samsung ASICs for hyperscaler AI inference: $AVGO, $MRVL, Mediatek Yields sector to make sure the chips work: $TER, $AEHR, Advantest Along with the raw materials or substrates needed for AI: $AXTI, $COPX, $SOI And many others become the single, largest, hedge against widespread AI displacement. Whoever owns the means of compute (bottlenecks, materials, datacenters): Owns the future of AI.
-
复盘今日持仓表现，VRT符合预期，LITE与COHR意外。
@InFoTheLongTerm 鉴于预测市场赔率，我原本预期今天会是 $VRT。$LITE 和 $COHR 的表现则令人意外。
英文原文
@InFoTheLongTerm I expected $VRT today given prediction market odds. $LITE and $COHR together came in as a surprise.
-
OpenAI警告电力是AI竞争核心瓶颈，看好美国能源基建股。
@toptickcrypto 存在多个瓶颈。电力/电网是其中最大的瓶颈之一。 OpenAI 就 Deepseek 蒸馏问题向国会发送了一份备忘录： “维持美国在 AI 领域的优势，取决于我们能否可靠地大规模生成和输送电力。” 电力输送 - $VRT, $ETN, $PWR, $WMB, $KMI 一级能源供应商：$CEG, $VST, $TLN, $GEV, $NEE, $BEPC, $D 电网-能源/存储 - $TSLA, $FLNC, $NRGV, $BE 能源：$TE, $FSLR, $NRG 这对这些公司来说是一个被重申的顺风因素。 而对于那些已经锁定吉瓦(GW)级容量的公司，如 $IREN, $NBIS, $WULF 和 $CIFR，则存在二级顺风因素。 备忘录的核心问题围绕知识产权盗窃和国家安全问题。但关于维持优势的最大警告是能源。 OpenAI 警告国会，2024年中国新增了429吉瓦(GW)的电力容量，这超过了美国整个电网的三分之一，也超过了全球电力增长的一半。他们认为，如果没有对美国电网进行激进扩张，中国“蛮力”式的能源建设最终将使其超越西方的 AI 能力。 光子学(Photonics)、先进封装(Advanced Packaging)和存储(Memory)是目前增长最快的三个瓶颈。然而，OpenAI 明确警告美国政府，谁能产生最多的电力，谁就能赢得 AI 竞赛。 他们的信息是：投资美国能源。
英文原文
@toptickcrypto There's multiple bottlenecks. Power/grid is one of the largest ones. https://t.co/rIKoKRyQOu
-
OpenAI警告国会：电力供应是AI竞争核心瓶颈，需投资美国能源。
OpenAI 就 Deepseek 蒸馏问题向国会发送了一份备忘录： “维持美国在人工智能领域的优势，取决于我们能否可靠地大规模生成和输送电力。” 电力输送 - $VRT, $ETN, $PWR, $WMB, $KMI 一级能源供应商：$CEG, $VST, $TLN, $GEV, $NEE, $BEPC, $D 电网能源/储能 - $TSLA, $FLNC, $NRGV, $BE 能源：$TE, $FSLR, $NRG 这对这些公司来说是一个被重申的顺风因素。 而对于那些已经锁定吉瓦(GW)级容量的公司，如 $IREN, $NBIS, $WULF 和 $CIFR，则存在二级顺风效应。 备忘录的核心议题围绕知识产权盗窃和国家安全问题。但关于维持优势的最大警告在于能源。 OpenAI 警告国会，2024年中国新增了429吉瓦(GW)的电力容量，这超过了美国整个电网的三分之一，也超过了全球电力增长的一半。他们认为，如果没有对美国电网进行激进扩张，中国“蛮力”式的能源建设最终将使其超越西方的AI能力。 光子学(Photonics)、先进封装(Advanced Packaging)和存储(Memory)是目前增长最快的三个瓶颈。然而，OpenAI 明确警告美国政府，谁能产生最多的电力，谁就能赢得AI竞赛。 他们的信息是：投资美国能源。
英文原文
OpenAI sent a memo to congress regarding Deepseek distillation: "Sustaining the American advantage on AI depends on depends on whether we can reliably generate and deliver power at scale." Power Delivery - $VRT, $ETN, $PWR, $WMB, $KMI Tier-1 Energy Providers: $CEG, $VST, $TLN, $GEV, $NEE, $BEPC, $D Grid-Energy / Storage - $TSLA, $FLNC, $NRGV, $BE Energy: $TE, $FSLR, $NRG This is a tailwind reiterated for these companies. And there's a second-order tailwind for companies that already secured GW capacity like $IREN, $NBIS, $WULF, and $CIFR. The core issue of the memo was around IP theft and national security issues. But the largest warning about sustaining an advantage was Energy. OpenAI warning Congress that in 2024, China added 429 Gigawatts (GW) of new power capacity, which was more than a third of the entire US grid and more than half of global electricity growth. Without a radical expansion of the American power grid, they believe China’s "brute force" energy buildout will eventually allow them to surpass Western AI capabilities. Photonics, Advanced Packaging, and Memory are three fastest growing bottlenecks right now. However, OpenAI explicitly warned the U.S. government that whoever generates the most power wins the AI race. Their message: Invest in American Energy.
-
梳理AI半导体供应链瓶颈，看好存储、光子及台积电，指出中游为定价关键。
以下是我对 Semivision 瓶颈总结的 TLDR（太长不看版）及其映射到投资框架的内容： HBM（高带宽内存）： HBM4 (16Hi) - 三星，SK 海力士，$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.
-
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 $GOOGL are increasing their capex spend enormously. This just goes to show: AI buildout is accelerating and is funded by the most profitable companies in the world. Long AI supply chains.
-
博通财报后AI板块抛售是误解造成的买入机会，新云板块中OpenAI依赖股除外
博通[$AVGO]业绩及其对AI板块的影响，如$LIITE和$NBIS： 博通的业绩"双重超预期"，营收$180.2亿(+28% Y/Y)，EPS $1.95，超出共识预期。 但AVGO下跌-11.64%，并拖累了整个AI板块。 这是买入机会吗？ 是的。 博通被视为超大规模云厂商ASIC代理增长的代表，因为亚马逊$AMZN Trainium、微软$MSFT Maia、尤其是谷歌$GOOGL TPU V7 Ironwood都通过它进行规模化部署。 而像$ALAB(-13.2%)、$CRDO(-5.11%)、$LITE(-12.23%)、$TSM(-3.71%)、$COHR(-9.25%)等公司都是TPU/ASIC建设以及博通作为公司的直接受益者。 博通下跌有三个原因，市场下跌有一个原因： 就博通而言，有一些小问题，如税率变化影响EPS模型，或因更多定制AI芯片而非更高利润率软件导致的"利润率压缩"，但这只是会计处理框架问题。（类似于$META在一次性税收后最初的下跌） 对博通和整体市场而言，是积压订单预期问题。以上所有引用的内容与ASIC市场预期增长相比都是小问题。 博通披露未来18个月$730亿的AI积压订单。而有关Anthropic和META购买价值数十亿$GOOGL TPU的传言，人们隐含预期是$800亿+。 然而，这次抛售是由算法和短期AI泡沫情绪驱动的价格错位，而非基本面破裂。 这条积压订单引用是确认订单的最低合同底线。谷歌$GOOGL、亚马逊$AMZN等公司可能会继续增加ASIC订单，而市场未能辨别这一细微差别。 分析师预期营收转化会更加前置，Q4之后积压订单应该会减少，这给出了2026年更高的可能范围$550-600亿+，而非$730亿预期中的$500亿。 TLDR：关于超大规模云厂商ASIC加速以与$NVDA依赖竞争这一论点没有改变。$AVGO和其他如$COHR、Sk Hynix、$MU、$VRT和$LITE都将受益。 这不是关于营收积压订单的最佳消息，但由于交付周期/订单周期和最低底线而被误解。 如果非要说什么，较低的超级云厂商ASIC需求对$NVDA及其生态系统是有益的，但我们也看到$CRWV、$SMCI、$NBIS和$NVDA GPU/DC计算生态系统今天都从盲目抛售中下跌超过5%，尽管存在负相关性。 这又是典型的"AI泡沫"周期因误解而再次来袭。AI股票普遍下跌10-12%的恐慌是一个绝佳的买入机会。 （该推文引用了@aleabitoreddit的推文，引用内容仅供理解语境）: 甲骨文[$ORCL]业绩及其对新云板块如$NBIS和$IREN的影响： 甲骨文报告EPS超预期且积压订单创纪录，但盘后下跌12%。 甲骨文较9月11日高点下跌39.8%，并拖累整个板块。 原因如下： 这次抛售不仅仅是对边际营收miss的反应，而是算法做空和投资者对AI资本支出周期可持续性以及该板块主要租户信用资质的不看好： OpenAI。 甲骨文宣布2026年资本支出增加$150亿至近$500亿，这与报道的与OpenAI $3000亿合作伙伴关系密不可分。 最初，OpenAI是前沿LLM，对甲骨文、Coreweave等公司有前景良好的资本支出承诺，促成了该板块的初始重新定价。 然而，随着超过$1万亿的义务以及Anthropic、Gemini、XAI等竞争对手的增加，市场严重质疑甲骨文、Coreweave等公司是否为无法从运营现金流履行其义务的租户在建设。 我们看到市场有效地发出信号：甲骨文正在为OpenAI创造一种不可持续的债务驱动型"供应商融资"，而OpenAI无法履行其承诺。 因此下跌是理性的：这次抛售是由信用风险和资本密集度的理性重新定价所驱动。 OpenAI融资担忧是合理的：OpenAI缺乏资金履行合同的假设得到其营收($130亿)与义务($600亿/年)之间明显不匹配的支撑。 信用担忧是真实的：甲骨文CDS利差扩大显示"信用事件"降级或违约的概率上升。 此外，我们看到这在新云板块引发传染效应，$NBIS从$140跌至$90s，$IREN从$80跌至$40s，$CIFR从$24s跌至$17s。 但这对$WULF、$NBIS、$IREN等新云公司来说是买入机会吗？ 是的。 这对$ORCL来说是好的买入机会吗？ 不是。 前瞻展望： $ORCL（很大一部分）、$CRWV（25%积压订单）是两个主要依赖OpenAI的公司，这一叙事可能因OpenAI的融资活动而瞬间翻转（+30%+变化）。 如果OpenAI在2026年以高估值超额认购IPO，且其新GPT模型击败Gemini/Claude，我们可以看到这种改变。 然而，许多其他公司与OpenAI无关。新云板块的原始论点是Mag7资本支出从其现金牛业务（Azure、AWS、GCP）向下流向：$NBIS、$IREN、$CIFR、$WULF等。 但随着最大玩家($ORCL、$CRWV)下跌，这些公司算法性地拖累了整个板块。 如果你看各家公司，$CIFR和$WULF由$GOOGL兜底，$IREN/$NBIS由$MSFT资助。 这些是与超大规模云厂商/Mag7的锁定合同积压订单，而非OpenAI。 这种因误解风险而导致的不理性抛售为新云板块提供了绝佳的买入机会，但不是与OpenAI相关的公司如$ORCL和$CRWV。
英文原文
Broadcom [ $AVGO ] earnings results and its effect on the AI sector like $LITE and $NBIS: Broadcom's ER was "double beat" with $18.02B revenue (+28% Y/Y) and $1.95 EPS, beating consensus. But AVGO dropped -11.64% and brought down the AI sector. Is this a buying opportunity? Yes. Broadcom is seen as a hyperscaler ASIC proxy growth as companies like $AMZN Trainium, $MSFT Maia, and most importantly $GOOGL TPU V7 Ironwood are scaled through it. And by proxy companies like $ALAB (-13.2%), $CRDO (-5.11%), $LITE (-12.23%), $TSM (-3.71%), $COHR (-9.25%), and are direct beneficiaries of the TPU/Asic buildout and Broadcom as a company. There's three reasons why Broadcom fell and one why the market fell: For Broadcom, there's minor things such as tax rate changing EPS models or "margin compression" from accounting from just more custom AI chips than higher-margin software, but this is just accounting framing. (Similar to how $META dropped initially on one-time tax post-ER) For both Broadcom general market, it was backlog expectations. Everything cited above is all minor compared to expected growth of ASIC markets. Broadcom cited $73B in AI backlog for the next 18 months. And rumors of Antrophic and META buying billions of $GOOGL TPUs, people were implicitly expecting $80B+. However, the selloff represents a dislocation in price driven by algorithms and short-term AI Bubble sentiment rather than a fundamental breakage. This backlog quote was the MINIMUM CONTRACTUAL FLOOR of confirmed orders. Companies like $GOOGL, $AMZN, will likely continue ramping up ASIC orders and the market failed to discern this nuance. Analysts are expecting revenue conversion to be more front loaded, and that there should be less backlog beyond Q4 given the cycles, which gives a higher likely range of $55-60B+ for 2026 rather than $50B expected of the $73B. TLDR: The thesis regarding hyperscaler ASIC ramp to compete vs $NVDA dependency has not changed. $AVGO and other players like $COHR, Sk Hynix, $MU, $VRT, and $LITE all stand to benefit. It's not the best news regarding the revenue backlog, but it's misunderstood due to lead-time/order cycles and minimum floors. If anything, lower hyperscaler ASIC demand is beneficial to $NVDA and their ecosystem, but we've also seen $CRWV, $SMCI, $NBIS and $NVDA GPU/DC compute ecosystem drop over 5%+ today from an indiscriminate sell-off despite inverse correlation. This is just the typical "AI Bubble" cycle hitting again from misunderstanding. The widespread panic of AI stocks dropping 10-12% is a great buying opportunity.
-
分析Google TPU v7供应链，建仓Lumentum以博弈TPU生态扩张。
对 $GOOGL TPU v7 Ironwood 供应商的分析。 以下是受 Google TPU 建设影响最大的公司列表。 + 我正在建仓的 TPU 相关股票。 [关键] 设计/IP： - 博通 [ $AVGO ]：共同设计并实现 Google 的 TPU ASIC（专用集成电路） [关键/高] 半导体晶圆代工： - #1 台湾半导体 [ $TSM ]：TPUv7 在 TSMC 3nm 工艺制造 - #2 三星电子：次要存储及晶圆代工合作伙伴 [关键/高] 存储： - #1 SK 海力士：为 TPUv7 Ironwood 提供 HBM3E - #2 三星电子：~TPUv7 特定报告强调 SK 海力士 + 三星。 [高] 光网络： - Lumentum [ $LITE ]：Google 广泛使用光电路交换 (OCS) - Coherent [ $COHR ]：OCS 参与者但较弱 [高] 电源管理 IC： - Monolithic [ $MPWR ]：这是一个投机性观点，即 Vicor 将被 $MPWR 取代，源自财报中提及 TPU [中] 热管理： - Vertiv [ $VRT ]：Vertiv 供应作为液冷系统核心的 CDU（冷却分配单元），将冷却液泵送至 TPU 芯片的冷板 - Modine [ $MOD ]：更投机性地认为他们提供大型冷水机组和空气处理单元 (AHU) [中] 互连： - TTM Technologies [ $TTMI ]，$ANET，Unimicron，Ibiden ______ Google TPU v7 “Ironwood” 的建设代表了一个旨在打破 $NVDA GPU 垄断的平行硅生态系统的构建。 实质性影响最集中在博通（作为硅架构师和商业载体）、存储综合体（SK 海力士/三星）以及光网络/电源领域（Lumentum/Vertiv），这是基于公开证据创建的，但很大程度上取决于采用率、供应商份额和竞争反应的实际表现。 从这项供应链研究中，我正在 $LITE 建立新头寸，以防 TPU 成为推理领域的主导 ASIC。 Lumentum 是 Google 致力于 OCS 的主要受益者，并构成了 TPU 吊舱中使用的 “Apollo” OCS 交换机的核心。 TPU v7 集群的爬坡直接转化为 Lumentum 光开关模块的出货量。由于 OCS 是 Google 超大规模方法独有的定制架构，Lumentum 在此处面临的 commoditization（商品化）压力小于标准收发器市场。 然而，如果 Anthropic、Meta、Apple 和其他公司购买 $GOOGL ASIC 导致 TPU v7 规模扩大，该供应链中的所有公司都将受益。
英文原文
Analysis of the $GOOGL TPU v7 Ironwood Suppliers. Here's the list of what comapnies are the most materially impacted by the Google's TPU buildout. + the TPU stock I'm taking a position on. [Critical] Design/IP: - Broadcom [ $AVGO ]: co-designs and implements Google’s TPU ASICs [Critical/High] Semiconductor Fab: - #1 Taiwan Semi [ $TSM ]: TPUv7 is fabbed at TSMC 3nm - #2 Samsung Electronics: Secondary memory & foundry partner [Critical/High] Memory: - #1 SK Hynix: HBM3E for TPUv7 Ironwood - #2 Samsung Electronics: ~TPUv7-specific reporting emphasizes SK hynix + Samsung. [High] Optical Networking: - Lumentum [ $LITE ]: Google uses extensively uses Optical Circuit Switching (OCS) - Coherent [ $COHR ]: OCS player but weaker [High] Power Management ICs: - Monolithic [ $MPWR ]: This is speculative that Vicor will be replaced by $MPWR, from earnings mentioning TPU [Medium] Thermal Management: - Vertiv [ $VRT ]: Vertiv supplies the CDUs that act as the heart of the liquid cooling system, pumping coolant to the cold plates on the TPU chips - Modine [ $MOD ]: More speculative that they provide provides the massive chillers and air handling units (AHUs) [Medium] Interconnects: - TTM Technologies [ $TTMI ], $ANET, Unimicron, Ibiden ______ The buildout of the Google TPU v7 "Ironwood" represents the construction of a parallel silicon ecosystem designed to break the monopoly of $NVDA GPU. The material impact is most concentrated in Broadcom (as the silicon architect and commercial vehicle), the Memory Complex (SK Hynix/Samsung), and the Optical/Power sectors (Lumentum/Vertiv) and was created from public evidence but is largely dependent on adoption, vendor shares, and competitive responses actually play out. From this supply chain research, I'm initiating a new position in $LITE, in the event the TPU becomes the dominant ASIC for inference. Lumentum is primary beneficiary of Google’s commitment to OCS and form the core of the "Apollo" OCS switches used in TPU pods. The ramping of TPU v7 clusters translates directly to unit volume for Lumentum’s optical switch modules. And because OCS is a bespoke architecture unique to Google’s hyperscale approach, Lumentum faces less commoditization pressure here than in the standard transceiver market. However all companies in this supply chain are set to benefit if the TPU v7 scales up from Anthropic, Meta, Apple, and others buying the $GOOGL ASIC.
-
分析VRT受益于AI基建电力瓶颈，认为其将跑赢大盘，NVT风险回报更优。
哈哈，关于股票代码的评论太多了，无法一一回复。关于 $VRT，它是数据中心建设的高贝塔敞口。Q3 营收 26 亿美元，同比增长 29%，指引 FY 中值为 102 亿美元。整体利润率增长，例如 Q3 调整后营业利润增长 43% 至 5.96 亿美元。后者可能表明，由于电力基础设施组件严重短缺和电网容量限制，其拥有类似内存行业的定价权。 $NVT 在当前市值下（24.5 倍 EV/EBITDA，对比 33 倍）可能具有更好的风险调整后回报。但除此之外，能源并非我的专业领域，因此我无法像对 neoclouds 或金融科技那样提供详细细致的观点。 不过，我可以说 Vertiv 在 AI 基础设施建设和电力供应约束的背景下，凭借其角色（例如与 nvda 合作的冷却和 800vdc 电力），方向上应该会跑赢大盘。
英文原文
lol there's a lot of ticker comments, can't respond to every one. So for $VRT it's a high beta exposure to DC buildout, from Q3 earnings $2.6B revenue 29% y/y, guided FY $10.2B midpoint, overall margins grew eg. adjusted operating profit increased 43% to $596 million in Q3. The latter probably indicates better pricing power like memory sector from a severe shortage of power infrastructure components and grid capacity limitations. $NVT might be better risk-adjusted return at current marketcaps (24.5x EV/EBITDA), vs. 33x. But that aside, energy not my field of expertise so I can't give detailed nuanced opinions like neoclouds or fintech. However, I can probably say Vertiv directionally should be outperform due to backdrop of AI infra buildout + power supply constraints and it's role (eg. cooling, 800vdc power that they're doing with nvda)