全部推文
共 6413 条 · 第 127 / 214 页 · 滚到底自动加载更多
-
通过供应链瓶颈分析挖掘AXTI的超额收益逻辑。
$AXTI 是你如何寻找超额收益(alpha)的完美例子。 在过去的几周里，我展示了： - 7纳米铟在标准金属市场(SMM)达到历史新高(ATH) - 来自 Litecounting 和其他分析师预测的光子供应链飙升 - 与日本和中国的贸易冲突影响住友电工(Sumitomo)及上游竞争对手的原料供应 - 超大规模云服务商资本支出流向专用集成电路(ASIC)和半导体供应链，如 $LITE - 谷歌张量处理单元(TPU)的光子物料清单(BOM)估算 - 美国政府没收资产后优先发展磷化铟(InP)供应链 我发现 $AXTI 在衬底生产以及上游原料方面拥有重大瓶颈控制力。 这就是你如何在一个月或两个月内找到超额收益并实现股票三位数回报的方式。 而不是等待财报并看到更新的 35 美元以上的分析师目标价。 这是我坐下来让市场消化我所做研究的时期。
英文原文
$AXTI is the perfect example of how you find alpha. Over the past few weeks I’ve shown: - 7n indium nonstandard on SMM reaching ATHs - photonic supply chains from litecounting and other analyst projections skyrocketing - trade conflicts with Japan and China affecting Sumitomo and upstream competitor feedstock - hyperscaler capex going into asics and semi supply chains like $LITE - estimated photonic BOM on Google TPUs - Us administration prioritizing InP supply chains after seizing assets And found that $AXTI has a major chokehold on both substrate production as well as upstream feedstock. This is how you find alpha and stock triple digit returns in a month or two. Not after waiting for earnings and seeing updated $35+ analyst price targets. This is the period where I sit back and let markets price in the research I’ve done.
-
杰文斯悖论致AI算力需求增，Talaas硬件固化适合边缘而非前沿模型。
@luke_judges 没错，正是杰文斯悖论(Jevons Paradox)。如果成本下降，就会消耗更多的训练/推理资源。另外，我认为 Talaas 似乎被过度炒作了，但为小型静态模型进行硬件固化(hardwiring)确实非常酷。也许适用于边缘(edge)用例。但不适用于需要巨大吉瓦(GW)级功耗的前沿(frontier)模型。
英文原文
@luke_judges Yep spot on, jevons paradox. Just use up more training/inference if costs drops. Also my opinion is taalas seems overhyped but hardwriring a small static model is incredibly cool. Maybe for edge use cases. But not for frontier models which require enormous GW power usage.
-
看好2026年电力板块因AI需求及降息带来的异常表现。
我认为2026年对于极其平淡的电力/电网(Utilities/Grid)板块来说可能是一个异常年份，原因在于该领域过度的资本支出(capex)、去年开始的降息在今年影响盈利，以及AI推理(AI inference)的极端需求。所以时间会证明这笔交易是否成功！我首次发帖时隐含波动率(IV)约为14%。
英文原文
I do think 2026 could be an anomaly for the extremely boring power/grid because of excessive capex into the sector, rate cuts from last year -> this year hitting earnings, and extreme demand for AI inference. So time will tell if the trade turns out well! IV was ~14% when I first posted.
-
澄清提及的账号为文章作者而非AI助手
@SharestepAI（指的是这篇文章的作者，不是你）
英文原文
@SharestepAI (Referring to this article writer, not you)
-
AWS限制AI代理应用，本地裸机部署成趋势
这是一个没有技术背景的作家会忽略的细微观点： - 像 AWS 这样的云 TOS（终端操作系统）会阻止许多与代理（Agent）相关的应用，因此你需要在本地运行服务器。 例如，如果你打算用 AI 代理做营销，AWS 会终止你的实例，所以这必须仅在本地运行。 - 平台知道虚拟机（VM）并能检测虚拟机管理程序指纹。裸机（Baremetal）对于 AI 自动化来说可以绕过检测。 这是针对 AI 集群（Swarm）的情况，而不是在一台设备上运行多个实例。 - 为每个代理提供隔离环境。给予 openclaw 根权限存在危险，可能会摧毁多个环境。 - 树莓派（Raspberry Pi）用于编排带有轻量级 openclaw 版本的 LLM，而不是运行推理。
英文原文
It's a nuanced point that writers without a technical background miss: - Cloud TOS like AWS prevents many agentic related applications, so you need to run servers locally For example if you're going to do marketing with AI agents, AWS terminate your instance, so this needs to be local only. - Platforms know VMs and detect hypervisor fingerprinting. Baremetal pis for AI automation bypass detection This is for AI Swarm cases, not running multiple instances one one device. - isolated environments for each agent. there's dangers in giving openclaw root access nuking multiple environments - Raspberry Pi is for orchestration of LLMs with lightweight openclaw versions. not running inference.
-
博主感谢粉丝夸奖，表示希望分享内容能助人。
@retail_mourinho 谢谢！你太抬举我了，希望我分享的内容能以某种方式帮助到其他人。
英文原文
@retail_mourinho Thanks! You flatter me too much, hope whatever I share helps others in some way.
-
强调建立独立投资信念，分享交易思考过程供他人学习。
@aditya_martand 你应该建立自己的投资信念(conviction)，而不是跟随他人。我只是分享我对 $XLU 这笔交易背后的思考过程，以防其他人觉得这些想法有趣并能从中学习。
英文原文
@aditya_martand You should build your own conviction, not follow others. I'm just sharing my thought process behind my own trade on $XLU in case others find the ideas interesting + can learn something.
-
澄清ROE与股价回报区别，预计2026年受AI支出和降息驱动表现异常。
两点： 1. 混淆利润率与总可寻址市场(TAM) 2. 你的陈述将净资产收益率(ROE)与股市回报混为一谈。 你说得对，收益=允许的ROE×费率基数(rate base)，且受监管公用事业的ROE是有上限的。不需要允许的ROE上升，只需费率基数爆炸式增长即可（现在正是如此，因为超大规模云服务商正在为电网升级买单）。 主要的一点是，单一指标ROE并不等同于多维度的股票表现。 股票百分比回报由每股收益(EPS)、降息带来的市盈率(P/E)扩张、未来预期增长（特别是关注来自AI推理+资本支出周期的数据中心增长）驱动。 鉴于巨大的AI支出加上降息顺风，我预计2026年将是异常值。
英文原文
Two things: 1. Mixing up profit margin with TAM 2. Statement you made conflates ROE with stock market returns. You're correct in saying earnings = allowed roe × rate base and roe is capped for regulated utilities. Don't need allowed roe to go up, just rate base to explode (it is now since hyperscalers are paying for grid upgrades) The main thing is that one specific metric ROE, does not equate to multifaceted stock performance. Stock % returns are driven by EPS, P/E expansion from rate cuts, future expected growth (especially looking at DC growth from AI inference + capex cycle). Given the massive AI spend coupled with rate cut tailwinds, I expect 2026 to be the anomaly.
-
指出AI预测偏差，强调基本面优于技术分析。
并非如此，Gemini 只是非常自信地给出了错误答案。我刚才试着问了它类似的问题，结果偏差极大。 基本上，图表呈现直线上升的原因是分析师预测（大语言模型并未对此进行训练）显示，SK海力士明年的净利润将超过 $AAPL。 SK海力士仍是一家市值约 4300 亿美元的公司，而苹果为 3.7 万亿美元。此外，三星的净利润更是天文数字。 因此，估值重估（Repricing）可能还有很长的路要走。 基本面 > 技术分析(TA)。
英文原文
Not really, Gemini is just very confidently wrong. I tried asking it similar questions just now and they were super off. Basically reason it's a straight line up is because analyst projections (which LLMs haven't trained on), projects that Sk Hynix makes more net income than $AAPL next year. SK Hynix is still a ~$430B company compared to $3.7T. Then Samsung net income is just astronomical. There's likely a long way to go for repricing. Fundamentals > TA.
-
博主确认存在类似低效现象，并指出近期在韩国市场发现了特定低效环节。
@unearthfinance 没错！市面上还有类似的情况。我最近一直在关注韩国市场，并特别注意到了这种低效现象。
英文原文
@unearthfinance Yep! There's others like this out there too. I've just been looking at South Korea markets a lot recently and noticed this inefficiency in specific.
-
博主自嘲打字频繁出错
@wyle_khite 我一直在打错字
英文原文
@wyle_khite I keep typoing
-
利用时区差异进行韩国指数与SK海力士的统计套利，揭示市场低效。
交易思路： 利用韩国 $EWY 期货（美盘）与 SK 海力士（欧盘）的时区差异，进行多市场套利。 如果 $EWY 大幅上涨而 SK 海力士（HY9H）走势平稳： > 买入法兰克福上市的 SK 海力士是一种统计套利。 因为相较于欧洲单只股票，$EWY 的美国期货定价更为准确。 该交易逻辑是——如果美国期货上的韩国指数大幅上涨： -> 鉴于指数集中度高，SK 海力士个股有很大概率跟随上涨（甚至更多）。 本质上是指数与成分股的相关性加上统计套利，并非无风险套利。 我在“淋浴思考”频道举过一个例子：当 $EWY 上涨约 3.8% 时，SK 海力士欧盘仅上涨 0.4%。 > 我预期 SK 海力士 HY9H（当时走势平稳且欧元价差较小）会弥补美国期货与法兰克福市场之间的延迟，并在次日定价。 我能在收盘前获得不错的成交，且未大幅推动股价，然后在当天或下一个欧洲交易日卖出，获利几个百分点。 总结： 外国股票中存在与时区相关的低效现象。 这些思路在被他人发现前可能是金矿。但这个具体思路在太多人看到后可能已经失效。 但市场低效确实存在。
英文原文
Trade idea: Korean $EWY Futures (US) -> SK Hynix (EU) time zone, multi-venue arbitrage. If EWY is up a lot while SK Hynix (HY9H) is flat: > Buying SK Hynix Frankfurt is a statistical arbitrage. as US futures in $EWY are accurate compared to EU single stock. The trade was - if Korean Indexes on US futures go up by a large amount: -> SK Hynix individually has a high probability of going up by similar amounts (if not more) given high index concentration. Basically index to component correlation + statistical arbitrage, not risk-free arbitrage. One example I posted in my shower thoughts channel was when $EWY was up ~3.8% then SK Hynix EU was only up .4%. > My expectation was SK Hynix HY9H (which was close to flat + spread on EUR) would play catchup to the delay between US futures and Frankfurt > be priced in the next day. Was able to get a decent amount of fill near close before moving the stock too much, then sell same day or next EU trading day for a few percent gain. TLDR: Time-zone related inefficiencies can be found across foreign equities. These ideas can be a gold mine before it gets discovered by others. This idea in specific is now likely gone after too many people see this. But, market inefficiencies do exist.
-
调侃市场将坚信AGI的投资人称为AI书呆子。
@fedex774 我喜欢他们把 X 上所有坚信通用人工智能(AGI)的投资人称为“AI 书呆子”。
英文原文
@fedex774 I love how they called all the AGI pilled investors on X “AI nerds”
-
澄清持有RPI基于财务逻辑，反对将其视为迷因股。
是的，我持有 $RPI。它在我投资组合中的占比非常小。 让我感到愤怒的是这种观点：“对公司本身的喜爱胜过其财务表现或潜力”。 整个投资逻辑（Thesis）都是围绕从收入增长到总可服务市场（TAM）的未定价财务数据展开的。 我宁愿看到像树莓派（Raspberry Pi）这样盈利的无晶圆厂半导体（Fabless Semi）公司，不要仅仅因为媒体想要编造故事而变成一只迷因股（Meme Stock）。
英文原文
Yes I own $RPI. It’s a very small weighting in my portfolio. What gets me riled up is this: "fondness for the company itself rather than financial performance or potential". The whole thesis was around unpriced financials from revenue growth to TAM. I’d rather not have a profitable fabless semi like Raspberry Pi turn into a meme stock just because the media wants to make something up.
-
批评大型期刊编造关于$RPI的虚假评论。
是的，我原本期望大型期刊能有更好的表现，而不是编造评论。这相当于我说《金融时报》是一本无人重视的知名梗图期刊，因为他们的一位编辑曾去过游戏驿站(GameStop)。即便这样说，可能也比他们关于 $RPI 编造的评论更接近事实。
英文原文
Yeah, I expected better from large journals instead of making up commentary. It’s the equivalent of me saying “Financial Times” is a well known meme journal nobody takes seriously because the one of their editors visited gamestop before. Even with that statement it’s probably more true than the commentary they made up about $RPI
-
嘲笑媒体错误对比RPI与DELL市盈率，批评评论员不专业。
@britplay 今天《金融时报》(FT) 又有一篇试图将 $RPI 与 $DELL 的市盈率(P/E)进行比较的文章，我简直笑出声来。显然，许多市场评论员根本不知道自己在说什么。
英文原文
@britplay There was another one from FT today trying to compare P/E ratios with $RPI to $DELL and I just burst out laughing. Clearly a lot of the market writers have no clue what they’re talking about
-
澄清媒体误将$RPI基本面逻辑扭曲为迷因股炒作。
《每日电讯报》今天发布了一篇关于树莓派（$RPI）的新文章。 虽然我很感激被称为“AI极客”…… 但我不理解为什么媒体普遍试图将树莓派框架化为“迷因股（Meme stock）”？ 整个论点（Thesis）的核心是：随着OpenClaw变体成为首选硬件，总可寻址市场（TAM）正在扩张。 在财务方面，论点在于如果囤货持续，营收数字如何能从预期的14%增长提升至温和的48-55%。 我原始论点中的直接引语是：这将是“反转的顺风（tailwind for a reversal）”，且“营收应受益于需求增加”。 这是逐字引用的。 我不理解像《金融时报》这样的媒体机构如何能从上述内容转变为从未说过的虚假评论。 对于《金融时报》：他们插入了关于“可被框架化为对立面的大型空头”的虚假评论。 对于《每日电讯报》：他们插入了诸如“对公司本身的喜爱而非财务表现或潜力”的虚假评论。 我不确定大型媒体机构为何不被要求更高的标准，却能随意捏造评论？ 再次强调，$AAPL Mac Mini作为AI编排（orchestration）硬件的囤货不会对苹果营收造成显著影响（因为它是3.7万亿美元市值的公司）。 然而，这对像$RPI这样的公司是实质性的（material）。 这就是核心论点。 OpenClaw和AI代理/集群（swarms）驱动树莓派销量增加（更重要的是，作为长期催化剂），这可能导致营收预测超预期+重估（re-rating）。 多家媒体机构极其不诚实地进行框架化，包括添加捏造的评论。 这是一个基本面、高度具体的论点，但多家媒体机构决定完全随机地捏造内容，将树莓派描绘成迷因股。
英文原文
The Telegraph put out a new piece today about Raspberry Pi ( $RPI ) While I appreciate being called an "AI Nerd"... I don't understand why they all in common try and frame Raspberry Pi as a "Meme stock?" The whole thesis was around TAM expansion from OpenClaw variants as the hardware of choice. For financials, it was how revenue numbers could increase beyond projected numbers from "14% growth to a modest 48-55% if hoarding continues." Direct quote from my original thesis: this would be a "tailwind for a reversal" and "revenue should benefit from increased demand". This was verbatim. I don't understand how media outlets like Financial Times can go from that -> into fake commentary that was never said. For FT: they inserted false commentary about "big shorts that can be frame as opposition" For The Telegraph: they inserted false commentary like "fondness for the company itself rather than financial performance or potential". I'm not sure how large media outlets aren't held to higher standards and are able to just make up commentary? Once again, $AAPL Mac Mini as AI orchestration hardware hoarding wouldn't make a dent on Apple revenue (as it's a $3.7T company). However, it's material to a company like $RPI. That's the core thesis. OpenClaw and AI agents/swarms drives an increase in sales of Raspberry PI (and better yet, as a long term cataylst), which would likely cause revenue projection beats + re-rating. Extremely disingenuous framing by multiple outlets, including adding made-up commentary. This was a fundamental, highly-specific argument, but multiple media outlets just decided to make up something completely random to paint Raspberry Pi as a Meme Stock.
-
XLU走势缓慢无需急，电力电网瓶颈或延续至2028年。
@Trylanol 老实说，我不认为这笔交易需要太多紧迫感，$XLU 的走势像恐龙一样缓慢。这只是众多交易思路中的一个！ 如果你自己决定，可能半年后入场也行。电力/电网在通往 2028 年的过程中很可能依然是瓶颈。
英文原文
@Trylanol I don't think this is a trade that requires much urgency tbh, $XLU moves like a dinosaur. This is just one trade idea of many! Probably could enter this like half a year later too if you decide for yourself. Power/grid is likely a bottleneck going into 2028 as well.
-
AI从训练转向推理，电网成新瓶颈，电力股或迎起飞。
好问题！答案是宏观/芯片。2023-2024年AI起步时，美联储将利率大幅上调至5%+。公用事业股通常类似债券替代品，因为基建背负大量债务。当时最佳选择是$NVDA，因为硅/芯片短缺。2023年市场不太关心电网电力，因为服务器尚未联网。但现在Mag7囤积的数百万芯片终于开始运行。虽然$TSM仍是瓶颈，但美国电网可能是最大瓶颈之一。此外，我们从训练（约占AI电力预期的10-15%）转向推理，后者由数十亿人24/7使用。因此，在去年降息、今年进一步降息以及前所未有的推理扩张后，现在可能是电力/电网起飞的时候。
英文原文
Great question! Answer to that is macro/chips. 2023, 2024 when AI first started taking off, fed was jacking rates up to 5%+. Utilities are often kinda like bond proxies since they carry a lot of debt for infra buildout. Best play then was $NVDA since silicon/chips were in shortage. Market didnt really care about grid power back in 2023 because servers weren' connected. But now there's millions of chips hoarded by mag7 that are finally getting turned on. So while $TSM is still a bottleneck, US power grid is likely one of the biggest ones. Also we've shifted from training (which was like 10-15% of AI power expectancy) into inference, which just runs 24/7 used by billions of people. So after rate cuts last year, more this year, and just unprecedented inference expansion, probably now is the time for power/grid to take off.
-
以$XLU为例展示期权高杠杆效应，并提示其高风险。
举个例子说明 $XLU 的杠杆效应——当股价为 43.6 美元时，你可以在较高行权价以 0.6 美元的溢价买入期权合约。名义敞口可能达到 72 倍，因此如果期权变为实值(ITM)，你可以用 60 美元控制价值 4360 美元的股票。2 倍杠杆无法达到这种效果。当然，期权交易风险极高，我只是分享我的看法。我相信其他人可以根据整体逻辑，在电力板块中找出其他方向性多头标的，如果其中有一两个赢家出现的话。
英文原文
So just to give you an example on leverage with $XLU - you can buy contracts for 60 cent premium at higher strikes when the stock price is $43.6. The notional exposure might be 72 times, so you can control $4360 worth of stock for $60 if it goes ITM. 2x leverage doesn't quite get to that. But of course the option play is extremely risky, I just wanted to share my thoughts. I'm sure others can come to other directional longs based on the overall thesis if there's one or two winners in the power basket.
-
博主回复感谢粉丝互动。
@pepemoonboy 谢谢！你也是。
英文原文
@pepemoonboy Appreciate it! Same with you.
-
超大规模云厂商负债投入AI基建，资金流向英伟达等上游供应商。
@LuffyDDK 他们真的在负债 lol 世界上最富有的公司正在为人工智能基础设施建设而负债。 尽管拥有数百亿美元的净利润，从 $AMZN 到 $GOOGL 的公司都大幅增加了资本支出(capex)： 以至于部分公司预计将在 2026 年出现净现金流出。 以下是结果及受益者： 亚马逊 ( $AMZN ) 2025: +460 亿美元 2026 (预估): +110 亿美元 Alphabet ( $GOOGL ) 2025: +803 亿美元 2026 (预估): +130 亿美元 Meta Platforms ( $META ) - 净债务(Net Debt) 2025: +229 亿美元 2026 (预估): -70 亿美元 (预计转为净债务) 微软 ( $MSFT ) 2025: +492 亿美元 2026 (预估): +590 亿美元 甲骨文 ( $ORCL ) 2025: -980 亿美元 (净债务) 2026 (预估): -1150 亿美元 微软似乎处于最安全的位置。虽然亚马逊和谷歌主要依靠运营收入为人工智能基础设施建设提供资金。 然而，巨大的现金流缓冲已消失。甲骨文和 Meta 似乎正在通过负债来推动建设，尽管 $META 实现了惊人的运营收入数字。 现在，资金流向哪里？ 运营收入预测： 1. $NVDA - 美国 1350 亿 -> 1865 亿 -> 2401 亿美元运营收入 2. 三星 - 韩国 ~302 亿 -> ~1700 亿 -> ~2267 亿美元 3. SK 海力士 - 韩国 ~327 亿 -> ~1240 亿 -> ~1610 亿美元 4. $TSM - 台湾 ~622 亿 -> ~863 亿 - ~1033 亿 – 1053 亿美元 5. $AVGO - 美国 ~415 亿 -> ~628 亿-651 亿 - ~847 亿-933 亿美元 当然，这些是基于分析师预测的粗略估计。 然而，从总体趋势来看，这看起来像是一种杠杆式赌注，即人工智能基础设施建设完成后将带来自由现金流(FCF)的分红回报。 但明显的赢家似乎是英伟达、三星、SK 海力士、博通和台积电。 随着超大规模云服务商(hyperscalers)中部分公司陷入债务，他们将资产负债表转移给这些供应商，期望从人工智能支出中获得长期投资回报率(ROI)。
英文原文
@LuffyDDK They literally are going into debt lol https://t.co/A4KDf93pYk
-
对电力电网板块2026年拐点的定量拆解。
可能还有其他人在我2-3天前在淋浴思考频道和公共时间线（引用推文）发布原始论点后买入。这更多是对为何我认为2026年可能是电力/电网板块拐点的一个定量拆解。之前只是高层级的方向性思维过程。
英文原文
Probably other folks who bought it since I posted the original thesis 2-3 days ago in my shower thoughts channel and public timeline too (quoted post). This was just more of a quantitative breakdown on why I think 2026 might be the inflection point for power/grid plays. Before was just a high level directional thought process.
-
@Ren_aramb 是的，电力/电网极其无聊。自1990年代以来仅上涨200%，因此隐含波动率(IV)定价极低。 我只是认为，随着AI推理/训练加速、超大规模云厂商资本支出以及降息，$XLU 可能正处于历史上独特的时刻。 我不知道我是否正确，我们拭目以待。 https://t.co/qhBgCX5bFo
英文原文
@Ren_aramb Yeah power/grid is extremely boring. Only up 200% since the 1990s so IV is priced extremely low. I just think AI inference/training ramp alongside hyperscaler capex + rate cuts might just be that unique moment in history with $XLU I don't know if I'm right or not, we'll see. https://t.co/qhBgCX5bFo
-
看好电力板块ETF $XLU，认为电网升级带来定价权，属两年期交易。
几点看法： 1. 表后储能(behind the meter)似乎只是应对AI数据中心(AI DCs)巨大吉瓦(GW)级电力需求的权宜之计。 2. 电网(grid)需要大规模升级难道不是看空理由吗？这可能反而赋予它们更强的定价权(pricing power)。 3. 小型模块化反应堆(SMR)和固态氧化物电解池(SOEC)可能要到2029年之后才普及，这是一个为期2年的交易机会。 $XLU是一只ETF，所以我不需要像挑选$BE那样去精选个股。我做多的是电力/电网板块，而不是押注具体的个股。
英文原文
few things: 1. behind the meter seems like a bandaid for the massive GW power required by AI DCs. 2. grid needing massive upgrades isn't a bear case? probably gives them even more pricing power 3. soec and smr are probably later on past 2029, this is a 2year trade. $XLU is an ETF so I don't need to pick individual winners like $BE. I'm long power/grid as a sector rather than getting the individual names right.
-
分享电力板块宏观看多观点及期权交易风控建议
是否自行交易取决于你的风险管理（通常实值期权更稳妥）。我想分享我对电力/电网作为多头方向的宏观观点，且鉴于历史表现，$XLU 波动率较低。若使用限价单交易深度虚值期权，请务必注意价差。通常较低行权价因流动性更好，隐含波动率(IV)更接近正常的~14%。
英文原文
Up for you to decide based on risk management if you want to do the trade yourself (more ITM usually the safer it is). Just wanted to present my own macro thesis on power/grid as a long and that $XLU has low volatility given previous history. Just be really careful about spread with limit ordersif you go very OTM. Usually the lower strikes have a more normal ~14% IV given they're more liquid.
-
受马斯克采访启发，预计机构将轮动至电力电网板块。
谢谢，看到埃隆·马斯克关于电力是瓶颈的采访有助于建立一些信念。鉴于他正在运营一家人工智能公司，他通常在这些事情上方向正确。我猜测机构可能会在未来一年半内轮动进入电力/电网领域，如 $XLU。
英文原文
Thanks, seeing the Elon Musk interview regarding power as a bottleneck helped build some conviction. Here's usually directionally right about things given he's running an AI company.. My guess is instituions might rotate into electricity/grid like $XLU over the next year and half.
-
建议用$XLU虚值期权博取电力板块高回报，优于个股。
这是一种风险更高的选择，旨在通过在波动率极低的ETF $XLU 中使用虚值期权(OTM options)来获取最大回报，从而获得美国电力/电网领域的敞口。即使 $VST、$CEG 或 $TLN 翻倍，你也无法获得同样的回报。使用ETF无需挑选个别赢家，整个板块都应受益。
英文原文
This is a riskier option for maximum returns using OTM options in an extremely low volatility ETF in $XLU for US power/grid exposure. You won't get the same returns from $VST, $CEG, or $TLN, even if they double. Don't need to pick individual winners with ETFs, sector as a whole should benefit.
-
看好$XLU虚值长期期权，因AI电力需求爆发叠加降息周期带来历史性机遇。
如果我要在1年内将10万美元变成100万美元。 我会选择：$XLU 虚值(OTM) 2年期长期期权(LEAPS) 2026年是现代市场历史上首次同时出现： - 利率下降 - AI推理(AI Inference) + 基础设施建设 通过映射分析，$XLU 有潜在约40%的涨幅（虚值期权可能带来1000%+的收益）。 这是我的宏观论点： 1. 降息 当美联储在不引发衰退的情况下降息时，公用事业公司的债务成本降低，机构投资者会将低收益的现金转向公用事业股息。 这会导致估值倍数立即扩张： 1995年：标普公用事业板块(S&P Utilities)在1995年回报+31.3%，1996年再+12.1%——累计回报约47% 2019年中周期降息：结果：$XLU 在该年产生+25.9%的总回报 标准的软着陆降息周期自然映射为25%至30%的基础回报。而我们要进入2026年的新降息周期。 2. 基础设施超级周期资本支出(CapEx) 基础设施资本支出为该板块带来复合盈利增长。继2000年代初之后，公用事业公司进入大规模资本支出周期以现代化老化的电网基础设施。 由于他们不断支出并扩大其受监管的费率基数(rate base)，$XLU 在2004年回报+23.5%，2005年+16.3%，2006年+20.8%，2007年+18.4%。 然而这一次： 2026年8000多亿美元的AI建设支出，使得2004年的电网现代化看起来像零钱一样微不足道。 因此，你有来自#1降息的估值倍数扩张（+15%至+20%），以及来自#2资本支出历史数据的每股收益(EPS)增长（+18%至+20%）。仅从历史教训来看。 但2026年是历史上AI使用带来的最独特时刻。 仅从我自己的模型预测来看，由于AI极端扩张，所有以前的估计可能都是错误的（例如美国能源部/劳伦斯伯克利国家实验室的预测）： 超大规模云服务商(Hyperscaler)资本支出流入（支出）（亚马逊、微软、Meta、谷歌、甲骨文）进入数据中心估算： 2024年：2200亿美元 2025年：3500亿美元 2026年：5500亿美元 2027年：8000亿美元 2028年：1.2万亿美元（4年增长：+445%） 美国数据中心电力使用量： 2024年：190太瓦时(TWh) 2025年：280太瓦时 2026年：430太瓦时 2027年：650太瓦时 2028年：980太瓦时（4年增长：+415%） AI消耗的总美国电力百分比： 2024年：美国电网的4.5% 2025年：6.6% 2026年：8.2-10.2% 2027年：13.4-15.4% 2028年：21.3-23.3% 劳伦斯伯克利国家实验室和美国能源部似乎低估了AI使用量（他们预测到2028年约为12%） 物理电网容量需求： 2024年：18吉瓦(GW) 2025年：35吉瓦 2026年：65吉瓦 2027年：105吉瓦 2028年：160吉瓦 基本上你可以看到2026年到2028年是拐点，而2024-2025年是爬坡期的缓慢年份。 然后是独立公司的“绝望溢价”。因为电网容量已售罄，科技巨头向公用事业公司支付巨额溢价以插队。例如PJM互联电网（弗吉尼亚“数据中心巷”），容量价格从2024年的每兆瓦日28.92美元飙升至2026/2027年令人难以置信的329.17美元。 $VST 或 Constellation 作为独立电力生产商在ETF中占很大权重。 纵观全局，你可以看到从2026年（现在）到2028年的极端扩张，以及用于建设基础设施的极端资本支出，与往年相比。 2026年是现代市场历史上第一次，所有因素同时为枯燥的电网/电力板块发力，其中AI是最大的顺风。 正如埃隆·马斯克所说：“数十亿美元最先进的硬件。闲置黑暗。不是因为芯片不工作。而是因为没有足够的电力来运行它们”。 再次强调，2026年由于AI和做市商(MMs)基于历史隐含波动率(IV)（极度平坦~14%-16%）定价虚值看涨期权，是一个绝对的历史异常值。 我们看到AI推理（超出之前的测量范围）以及训练（根据OpenAI今天的报告）的爆发。 所以，地球上最无聊的板块（电力/电网），可能会因为超大规模云服务商/政府对电网改进的支出 -> AI推理/训练的极端电力消耗 -> 降息等因素，成为重大反弹的起点。 这只是我的个人论点，期权伴随风险并放大下行风险。这些也是我自己的预测，不确定是否会高于或低于它们。 但基本上： 2026年是一个绝对的历史异常值。 美国的新瓶颈是电力。 有来自AI的极端需求，极端资本支出，降息： $XLU 看起来是暴露于此的最佳交易。 时间会证明这是否正确。
英文原文
If I had to turn $100k -> $1M in 1 year. It would be: $XLU OTM 2 year leaps 2026 is the first time in modern history markets have: - falling interest rates - AI inference + buildout There's a potential ~40% for XLU (1000%+ OTM), from mapping. Here's my macro thesis: 1. Rate Cuts When the Fed cuts rates without a recession, utility debt becomes cheaper, and institutional rotates low-yielding cash to for utility dividends. This causes immediate valuation multiple expansion: 1995: The S&P Utilities sector returned +31.3% in 1995 and another +12.1% in 1996 - ~47% cumulative return 2019 Mid-Cycle Cut: Result: XLU generated a +25.9% total return in that single year Standard soft-landing rate-cut cycle naturally maps to a 25% to 30% baseline return. And we're entering a new rate cut cycle in 2026. 2. The Infrastructure Supercycle Capex Infra CapEx gives the sector compounding earnings growth. Following the early 2000s, utilities entered a massive CapEx cycle to modernize aging grid infrastructure. Because they were constantly spending and expanding their guaranteed rate base, XLU returned +23.5% in 2004, +16.3% in 2005, +20.8% in 2006, and +18.4% in 2007. However this time: The $800B+ AI buildout of 2026 makes the 2004 grid modernization look like pennies. So you have Valuation Multiple Expansion (+15% to +20%), from rate cuts from #1. EPS growth (+18% to +20%) from #2 from capex spend historically. Just from a history lesson. But 2026 is the most unique moment in history from AI usage. Just from my own model projections as all former estimates are likely wrong from extreme AI ramp (eg. DOE/LBNL projections): Hyperscaler CapEx Inflows (Spend) - (Amazon, Microsoft, Meta, Google, Oracle) into DCs est: 2024: $220 Billion 2025: $350 Billion 2026: $550 Billion 2027: $800 Billion 2028: $1.2 Trillion (Growth: +445% over 4 years) U.S. Data Center Power Usage: 2024: 190 TWh 2025: 280 TWh 2026: 430 TWh 2027: 650 TWh 2028: 980 TWh (Growth: +415% over 4 years) % of Total U.S. Electricity Consumed by AI: 2024: 4.5% of the U.S. grid 2025: 6.6% 2026: 8.2-10.2% 2027: 13.4-15.4% 2028: 21.3-23.3% Lawrence Berkeley National Laboratory and the Department of Energy seem off by AI usage (they're projecting ~12% by 2028) Physical Grid Capacity Demand: 2024: 18 GW 2025: 35 GW 2026: 65 GW 2027: 105 GW 2028: 160 GW Basically you can just see 2026 into 2028 being the inflection point whereas 2024-2025 where slower years on the ramp up. Then there's the "Desperation Premium" for independent companies. Because grid capacity is sold out, tech giants are paying massive premiums to utilities to cut the line. eg. PJM Interconnection (Virginia "Data Center Alley"), capacity prices spiked from $28.92 per MW-day in 2024 to an unfathomable $329.17 per MW-day for 2026/2027. $VST or Constellation are a large weighting in the ETF as independent power producers. Across the board, you can see the extreme ramp from 2026 (now) into 2028 compared to previous years, alongside extreme capex going into building the infrastructure. 2026 is the first time in modern market history that every single thing is firing at the same time for the boring grid/power sector with AI as the biggest tailwind. And as Elon quotes it: "Billions of dollars of the most advanced hardware. Sitting dark. Not because the chips won't work. Because there's not enough electricity to run on them". Again 2026 is an absolute historical anomaly due to AI and MMs have priced in historical IV (extremely flat ~14%-16%) for OTM calls. We're seeing an explosion in AI inference (beyond previous measurements) as well as training (per OpenAI report today). So the most boring sector on earth (power/grid), might just be the start of a major rally due to hyperscaler/gov spend into grid improvements -> extreme power consumption from AI inference/training -> rate cuts and others. This is just my personal thesis, options come with risk and magnifies downside too. These are also my own projections, no certainty if they will exceed or be lower than them. But basically: 2026 is an absolute historical anomaly. New bottleneck in the US is power. There's extreme demand from AI, extreme capex, rate cuts: $XLU looks like the best trade for exposure. Time will tell if this is right or not.
-
ETF持仓具定价权，电力供应未绕开公用事业。
@DavidLiaoCH @r0ck3t23 ETF 中很大一部分持仓是独立公司，拥有更强的定价权。此外，这并非完全绕过公用事业，指数中的公司（如 $VST）仍直接从这些公司获取电力。
英文原文
@DavidLiaoCH @r0ck3t23 Large % of concentration of the etf are independent that have more pricing pricing power Also it’s not quite bypassing utility, they still get the power directly from companies like $VST in the index