全部推文

共 6399 条 · 第 72 / 214 页 · 滚到底自动加载更多

  1. 杂谈

    引用特朗普关于文明可能消失的表态,并调侃这并不算正面。

    特朗普: “一个完整的文明今晚就会消亡,再也不会回来。 我不想那样发生,但它很可能会发生。” 嗯……这种话听起来不太正面吧?

    英文原文

    Trump: “A whole civilization will die tonight, never to be brought back again. I don’t want that to happen, but it probably will.” Yeah… statements like that’s aren’t too positive? https://t.co/J3GhvahLVg

  2. 杂谈 $IBKR

    推荐 IBKR 用于国际股票交易。

    @japhy_zeng 我推荐用 $IBKR 来买国际股票。反正你大概也该开个账户,因为他们的执行最好。

    英文原文

    @japhy_zeng I recommend $IBKR for international stock access. Probably want to get an account anyway since they have the best execution

  3. 杂谈

    用抓到稀有宝可梦却舍不得买高级球来打比方。

    @rGeQM1SvLq29316 这就像是在荒野里抓到一只闪光拉帝欧斯,然后因为高级球太贵而不用高级球一样。

    英文原文

    @rGeQM1SvLq29316 This is the equivalent of finding a shiny Latios in the wilderness. Then not using an ultra ball because the ball was too expensive

  4. 供应链分析 $3105$LITE$SIVE$AAPL$AVGO

    认为 Win Semi 被严重低估,在机器人、太空和 AI 供应链里都很关键。

    一个被严重低估的冷门 OP 标的。 还是几乎没人谈论 Win Semi($3105.TWO)…… 从 $LITE 到 $SIVE,以及从 $AAPL 到 $AVGO 再到 SpaceX。 感觉大家其实都在用它? 我觉得 54 亿美元的估值低估了它在机器人、太空和 AI 超大规模客户供应链中的重要性。

    英文原文

    What a sleeper OP pick. Still almost nobody talks about Win Semi ($3105.TWO)… From $LITE to $SIVE. And $AAPL to $AVGO to SpaceX. Feels like everyone uses them? $5.4B price tag is underselling its importance to robotics, space, and AI hyperscaler supply chains imo. https://t.co/lMg92KzufM

  5. 方法论

    把股价下跌更多归因于稀释压力,而不是别的因素。

    @neverlongqcom 这很复杂。我个人更倾向于把它归因于稀释压力,因为没人想在他们把流通股翻倍的时候还继续持有。

    英文原文

    @neverlongqcom It’s multifaceted. I personally attribute it more to dilution overhang because nobody wants to be holding the stock if they double the float

  6. 个股论点 $AXTI

    认为大规模增发会很掠夺性,若通过就会清仓 AXTI。

    @LeaderInvests 那种行为就是掠夺性操作,尤其是增发规模这么大。 如果它通过,我会把自己手里的每一份 $AXTI 都卖掉。

    英文原文

    @LeaderInvests That’s called being predatory, especially with the share issuance size. I’m dumping every $AXTI position I have if it passes.

  7. 供应链分析 $AAOI

    认为 AAOI 的 5000 万美元相对未来扩产资本开支并不夸张。

    @nasdaq90 5000 万美元,相比为了把美国做成最大的 1.6T 产能、用于 InP fab 和封装扩产所需要的 70 亿美元,其实差别很大。 而现在却是 5200 万股流通股里要再加 5000 万股,至于到底是为了什么,没人知道。

    英文原文

    @nasdaq90 $500m relative to $7B for InP fab and assembly expansion for becoming the biggest 1.6T capacity in America for $AAOI. Is materially different than 50m shares of 52m share float for nobody knows what.

  8. 方法论 $AXTI$IREN

    区分正常融资和会直接摊薄股东权益的疯狂增发。

    @BitcoinAIGuy 通过像 Northland 那样融个 1 亿美元和把流通股直接翻倍、像 $AXTI 或 $IREN 那样把股东权益稀释掉,这两者是完全不同的。 至少 AXT 这次还得经过股东投票。IREN 的 ATM 还在继续。

    英文原文

    @BitcoinAIGuy There’s a difference between raising capital like $100m from Northland. Then getting your equity wiped out by doubling the float with $AXTI or $IREN. At least with AXT it’s a shareholder vote. IREN ATM is ongoing.

  9. 方法论 $GME$AMC$IREN

    拿 GME/AMC 作比,认为 IREN 的大幅增发对普通股东很伤。

    如果你是 $GME、$AMC 的 CEO,一个逻辑上可行的选择可能就是把 ATM 测试抛给散户股东,让公司受益,因为他们会接住。 但对 $IREN 的套牢者或者 meme 股来说,我个人不会就这样眼睁睁看着自己被稀释到原来 5200 万流通股的两倍左右。 我希望它别通过,因为这太蠢了。他们之前已经从 Northland 融到足够多了,这次就是掠夺性操作。

    英文原文

    Yes if you were $GME, $AMC CEO’s a logical choice might be to dump ATMs on a retail shareholders to benefit the company because they would take it. However $IREN bagholders or meme stocks, I would not personally just sit there and get diluted from around double the 52m share float. I hope it doesn’t pass since it’s stupid. They already raised enough from Northland and this is predatory

  10. 供应链分析 $LITE

    认为市场已经在给 LITE 的后续下跌定价,且其下跌会拖累其他标的。

    @LepoulpePoulpo 是的,已经到 31 号了,而且你也看到了从那之后股价的下跌。在我看来,市场只是在提前定价。 而且 $LITE 跌了 7%,算法上也会把其他玩家一起带下来。

    英文原文

    @LepoulpePoulpo Yeah it’s 31st, and you’ve seen the stock drop since then. Imo markets are just pricing in. Also $LITE is down 7% and algorithmically it brings down other players too.

  11. 个股论点 $AXTI$IREN

    解释 AXTI 增发提案的稀释幅度,认为若通过会对股价不利。

    如果你在关心 $AXTI: 它在这条新的潜在稀释消息出来后跌了 21%。 董事会想再增加 5000 万股(最高可稀释到 20 亿美元级别),要在 14 号的股东投票里决定。 也就是从 7000 万股增加到 1.2 亿股。 我对 $IREN 那种 60 亿美元过度稀释也是同样看法,而我持有的 AXT 这份价格提案也是一样。 我不会“相信管理层会明智使用”这些股份,而且这份文件本身就是个危险信号。 不过话说回来,我们就看 14 号会发生什么。如果它通过了这 5000 万股的新增授权稀释,我个人不会继续持有。 但如果它没通过,我预计会反弹。

    英文原文

    If you’re curious about $AXTI: It’s down 21% on the new potential dilution news. Board wanted to add 50M more shares (up to $2B worth to dilute) for shareholders vote in the 14th. 70m -> 120m shares. I say this about $IREN excessive $6B dilution and it’s the same with AXT price proposal that I hold. I would not “trust in management” to use it wisely and the fact it’s filed is a red flag. That being said: we’ll see what happens on the 14th. If it passes for 50m more share authorized dilution, I personally would not hold. But it fails, I expect a recovery.

  12. 杂谈

    表示看到大家在 Hammond 讨论里的行业经验很酷。

    @BrokeDadCapital 在评论里看到大家分享 Hammond 相关的行业经历,感觉很酷。

    英文原文

    @BrokeDadCapital Super cool to hear industry experiences around Hammond in the comments

  13. 杂谈

    说 Hammond 口碑不错,但讨论的人并不多。

    @zephyr_z9 @rwang07 Hammond 看起来口碑不错,不过真正在谈它的人并不多。

    英文原文

    @zephyr_z9 @rwang07 Hammond seems well liked. Not rlly many people talking about it though

  14. 供应链分析

    指出变压器是当前很大的瓶颈,海外也在抢货。

    @clarity_99X 现在变压器的瓶颈真的很大,大约一种型号要等 2 年,其他类型要等 3 到 5 年。 超大规模客户现在都在海外四处抢货,而且刚好有一家相对较小、加拿大市值 20 亿多的大公司掌握了很大的市场份额。

    英文原文

    @clarity_99X Yeah there’s a pretty big transformers bottleneck right now ~2Y for one type and 3-5Y for the others. Hyperscalers are kinda scrambling abroad for them, and it just so happens there’s one relatively small 2B+ CAD company with a very large market share. https://t.co/qUZ4TsWOys

  15. 供应链分析 $AMZN$MSFT$META$GOOGL$ORCL

    用订单和 backlog 说明变压器 / 开关设备是数据中心的关键瓶颈。

    当前瓶颈:变压器 / 开关设备。 交易思路:在 184 加元买入 Hammond(约 22 亿加元 / 15 亿美元)的多头。 它们主导的市场包括: - 变压器(干式,多年瓶颈,约占市场 23%) - 开关设备(2-3 年瓶颈) - 还生产液体型产品(5 年瓶颈,规模更大) 我个人预计其价格会像 NAND 一样上调,因为 $AMZN、$MSFT 等公司都在争夺配额。 你可能见过这类说法: “美国一半的数据中心建设被推迟或取消,增长受限于电力基础设施短缺”…… 再往下看: “为了解决短缺……加拿大、墨西哥……成了 AI 数据中心高压变压器的最大供应国”…… 猜猜谁在加拿大的 Guelph、墨西哥的 Monterrey 3 和 4,以及美国都有布局? Hammond。 再看文章引用的,为什么超大规模客户的数据库扩张正在崩掉: “这些挫折背后的主要原因,是关键电气组件的可获得性——比如变压器、开关设备。” 机构大概都在看 Powell、Eaton 之类的公司……但他们可能不知道? 这些公司其实会买 Hammond 的变压器再装进自己的开关设备里(“大量卖给数据中心、开关设备制造商”)。 它们在变压器市场的份额其实很大(例如干式约 23%)。 最有说服力的信号: -> 2025 年 backlog 同比增长 122%。我们可以据此推测超过 10 亿加元。 例如,公司 2025 年销售额为 8.98 亿加元,产能已经顶到上限。管理层说,2025 年第三季度末的订单价值相当于整个第三季度 backlog 的 53%。 考虑到 2025 年第四季度营收是 2.54 亿,而 backlog “翻了一倍多”,我们可以推算总 backlog 超过 10 亿加元。 另外: “去年毛利率压缩是因为墨西哥工厂建设,但预计毛利率都会上升,工厂扩建会在 2026 年第二季度转化为加速收入增长”,也就是现在。 下行风险在于原材料成本(铜、电工钢)如果再次上升,不过既然处在这种瓶颈里,他们是可以提价的。 我个人的 2026 年前瞻市盈率估计大概是 18-21 倍,2027 年不到 15 倍,因为有产能爬坡。 但如果他们能一边提价,一边接到超大规模客户的紧急订单,前瞻市盈率甚至有可能到个位数。不过那可能会和新并购混在一起。 总之它还是很便宜。 简单总结一下: $AMZN、$MSFT、$META、$GOOGL、$ORCL 的数据中心正在被变压器 / 开关设备短缺卡住。 市场似乎错过了这家市场份额很大的小公司,尽管它有 backlog 可见度,而且随着产能扩张上线,营收还在增长。 我个人觉得很有吸引力,所以建了多头仓位。 当然,这只是我的个人想法,做决定前请自己 DYOR。

    英文原文

    The current bottleneck: Transformers/Switchgear. 

Trade Idea: Long Hammond (~2.2B CAD / ~$1.5B USD) at 184 CAD. They dominate the market for: -Transformers (dry, multi year bottleneck ~23% of market), -serve to switchgear (2-3Y bottleneck) -and manufacture liquid too (5Y, larger bottleneck) 
I personally anticipate components price hikes like NAND, as $AMZN, $MSFT and others compete for allocation. 

You might have seen: “Half of US data center builds have been delayed or canceled, growth limited by shortages of power infrastructure”… Then you go further:

“To address shortages… Canada, Mexico… became the biggest suppliers of high-power transformers for AI data centers to AI data centers”

Guess who is in Canada (Guelph).. Mexico (Monterrey 3 and 4)… and the US?

Hammond

Then here’s the reason the articles cite why hyperscaler DB buildouts are falling apart: 
 “Major reason behind these setbacks is the availability of key electrical components — such as transformers, switchgear”.  Institutions are probably looking at Powell, Eaton, and others… but little do they know? Companies like these actually buy Hammond’s transformers to put inside their own switchgear (“strong sales into data centres, switchgear manufacturers")

Their market share over the transformers market is actually pretty large (eg. ~23% dry).  
The most compelling signal:

-> 122% Y/Y 2025 backlog increase. And we can infer this to be 1B+ CAD.  Eg. company achieved 898m CAD in sales in 2025, capacity ceiling. Management said close of Q3 2025 orders were valued at 53% of the entire closing third-quarter backlog. Given that Q4 2025 revenue was 254 million and the backlog is "more than doubled," we can infer a total backlog value exceeding 1 billion CAD. Also: 
“Gross margin compression last year was due to the buildout of their Mexico facility, but both gross margins are expected to increase and the facility expansions are expectied to turn into accelerated revenue Q2 2026)” which is now.

Downside is if raw material costs (copper, electrical steel) spike again, but given this bottleneck, they can price hike. 

Personal FWD P/E estimates would be ~18-21 for 2026, <15 for 2027 from volume ramp. But I think it’s possible to hit single digit fwd P/E if they do price hikes mixed with hyperscaler emergency orders. But that might get a little mixed with the new acquisition. Regardless still looks cheap. 
 Just a TLDR:  
$AMZN, $MSFT, $META, $GOOGL, $ORCL datacenter are being bottlenecked because of a lack of transformers/switchgear.

Seems like markets missed this little player with large market share, despite backlog visibility and increasing revenue from capacity expansion coming online. I personally found it pretty compelling, so I went long. Just sharing my personal thoughts, of course DYOR before making any decisions yourself.

  16. 个股论点 $AEHR

    认为 AEHR 仅一天就涨幅可观,可能被低估。

    $AEHR 这波真是很快。 涨了 14.28%……我原本确实觉得它是下一个硅光瓶颈中心,市值大概 11 亿美元,还有很多超大规模客户在做认证。 可能之前确实低估它了。

    英文原文

    Well that was fast with $AEHR. Up +14.28%… I did think the center of the next silicon photonics bottleneck at $1.1B, with tons of hyperscalers qualifying it. Was kind of undervaluing it. https://t.co/gDVc6xzF85

  17. 杂谈

    抱怨自己发言被 X 审查。

    @SpaceSector001 不是啊,我只是因为拿 Mariachi 乐队和 capex 开了个玩笑,X 的负责人就把我评论审查掉了。 而且我连问一句费用或溢价到底去哪了都不行。 我就只是个普通人在这儿。

    英文原文

    @SpaceSector001 Nah, head of X censored my comments for joking about Mariachi bands and capex. And I can't even ask a question about where any of the fees or premiums go. Just a random on here

  18. 杂谈

    要求对费用去向保持透明。

    @funniestlines 他们至少应该透明说明费用都去哪了。 最糟糕的就是不回答,然后直接关掉工单。 我只是想问一下费用拆分,应该不难吧?

    英文原文

    @funniestlines The least that they can do is provide a transparent breakdown on where the fees go. The worst they can do is not answer then close the ticket. I’m just asking for transparency on fee breakdown, it shouldn’t be difficult?

  19. 杂谈

    吐槽自己被收取远超实际金额的费用。

    @KiakotosC 很抱歉听到你支付的费用竟然是实际应收金额的 240%。 我只是想知道那多出来的溢价到底去哪了。 但他们不回答,反而说自己不能答复,然后把工单关了。

    英文原文

    @KiakotosC Sorry to hear that you’re paying 240% of what the actual charge is supposed to be. I’m just wondering where all that extra premium charge goes to. But instead of replying they say they can’t answer then close the ticket.

  20. 杂谈

    说自己是最支持 Elon 的,但仍然被莫名屏蔽。

    我其实是最支持 Elon 的 humanoid / space 计划的人了,而且我和 DOGE、Neuralink、SpaceX 那些正在和 Elon 面试的人都很熟。 所以我真不知道 Nikita 为什么要把我那条拿 Mariachi 乐队和 capex 开玩笑的评论屏蔽掉…… 我明明是支持 X 的,但现在完全不知道到底发生了什么。

    英文原文

    I’m literally the most pro-Elon humanoid/space program there is and I’m close friends with ppl from DOGE, Neuralink, SpaceX and others doing interviews with Elon. Not sure why Nikita blocked and hid my comment when I joked about Mariachi bands and capex… Like I’m supportive of X but I have no clue what’s going on anymore.

  21. 杂谈

    质疑 X 对海外用户额外收费且拒绝解释。

    X 到底怎么了? 它居然对我的海外用户在订阅费基础上又加收了 46%。 我只是想问,多出来的这些溢价到底去哪了…… 我可是金融类里付费订阅人数最多的人之一。 -> 他们拒绝给一个简单答案,然后直接把工单关掉。 我一直努力劝别人继续留在 X 上,而不是去金融竞品平台……但这到底算什么? 我本来以为这个平台追求的是最大透明度?

    英文原文

    What's even going on at X? X is charging my foreign users an additional 46% on top of my subscription. I'm only asking where the additional premiums go to... From someone who has the most paid subscribers in the finance category. -> They refuse to give a simple answer and closes the ticket. I've been trying hard to advocate other users staying on on X compared to financial competitors... but wtf. I thought this platform was one seeking maximum transparency?

  22. 方法论 $AEHR$SIVE$SOI

    把内存仓位更集中到 SiPh/CPO 相关标的,并期待 SK 海力士和三星财报带来反弹。

    @B38B37 我个人把内存相关仓位更多转向了 SiPh/CPO 相关的名字,比如 $AEHR、$SIVE 和 $SOI。 我现在还是有不少内存敞口,只是比今年早些时候少一些。不过我确实预计接下来 SK 海力士和三星的财报会再带来一轮上涨。

    英文原文

    @B38B37 I personally pivoted concentration in memory more toward siph/cpo names like $AEHR $SIVE and $SOI. I still have a good amount of memory exposure but less than earlier this year. However I do expect the upcoming sk Hynix and Samsung earnings to trigger another rally

  23. 杂谈

    说很多公司找自己做 SPV 或基金,但觉得手续麻烦,暂时只想玩 X。

    @HeLiuLeo 现在很多公司都来找我,希望我去发起一个 SPV 或者让我来管一个基金。 我个人觉得这些只是多了点手续和额外麻烦而已。我也不觉得自己还需要那笔额外的钱。 所以我现在还是先继续在 X 上发帖,图个开心。

    英文原文

    @HeLiuLeo lot of firms reached out for me to launch a SPV or fund for me to manage. Imo just a bit of a hassle with paperwork and extra stuff. I don’t think I need the extra money either. So just having fun posting on X for now

  24. 杂谈

    感谢对方的一美元打赏,并表示更在意 thesis 被市场验证。

    @HeLiuLeo 谢谢这一美元!这对我意义很大。我其实更在意的是 thesis 能在市场里被验证,以及能用想法帮助别人,而不是这笔钱本身。 不过它确实够我早上买一杯拿铁了。

    英文原文

    @HeLiuLeo Thank you for the $1! Means a lot, I do care more about thesis validation in the market + helping others with ideas over the monetary amount. But it does help pay for a latte in the morning.

  25. 杂谈 $NVDA

    回忆自己做深度学习时准确率超过 DeepMind,还曾被英伟达招募。

    我以前在为一本自然类出版物搭建深度学习模型时,准确率其实还超过了 Google DeepMind。 后来 $NVDA 还试图招我去带他们的 AI 团队,哈哈。 最后我还是在做自己的事情,不过有一点编码背景确实很有帮助,能做出一些很酷的东西!

    英文原文

    I actually beat Google deepmind’s in accuracy back when I was building out deep learning models for one of my nature publications. $NVDA tried recruiting me to lead their AI teams after that lol. Ended up just doing my own things now but it does help to have some background in coding to do cool stuff!

  26. 方法论

    说自己为了研究,专门调过 LLM 和数据管线,不用 Bloomberg。

    @0xInitialAce 我其实自己把用于研究的 LLM + 数据管线都细调过了,因为我受够了 Gemini / ChatGPT 老是把事情搞错。 自定义工具确实帮了很多……我个人也不用 Bloomberg terminal 之类的东西。

    英文原文

    @0xInitialAce I actually fine turned my own LLM + data pipeline for my research cause I got tired of Gemini/chatgpt getting everything wrong. Custom tools helps quite a bit… I personally don’t use Bloomberg terminal and ithers

  27. 杂谈 $LITE$TSEM

    调侃自己被 Bloomberg 等称作 meme trader。

    @jamesxtrades 不知道啊,照 Bloomberg、FT、Reuters 这些说法,我算是个“meme trader”。 真不知道机构为什么会想去复制交易像 $LITE、$TSEM 这种 meme 股。

    英文原文

    @jamesxtrades Idk according to Bloomberg, FT, Reuters, and others I’m a “meme trader”. Not sure why institutions would want to copy trade meme stocks like $LITE, $TSEM and others.

  28. 业绩复盘 $TSEM$NBIS

    回顾 TSEM 从横盘到三周翻倍,怀疑机构开始跟单。

    我还是觉得 $TSEM 从年初到现在整年都横着走这件事挺好笑的。 我一做多,它现在三周里就快翻倍了。 这可是个 220 亿美元+ 的股票(体量接近 $NBIS)。 也许机构开始跟单了?

    英文原文

    I still find it funny $TSEM was flat the entire year to date. I go long and now it’s close to doubling in 3 weeks. This is a $22B+ stock BTW (close to $NBIS size). Maybe institutions started copy trading? https://t.co/kUqw9b6Ksk

  29. 供应链分析 $NBIS$CRWV$META

    认为超大规模客户延迟扩建带来的外溢,会利好 Neocloud

    我最近对上个月那份报告想得更多了…… 我觉得它对 Neocloud 板块,比如 $NBIS,甚至 $CRWV,的利好可能比市场想象得更大。 原因在于超大规模客户扩建延迟所带来的外溢效应。 这也部分解释了为什么 $META 可能签了 270 亿美元的交易。 那些已经为扩建锁定了产能 / 组件订单的公司,大概率会成为赢家,因为其他竞争对手会被卡住。

    英文原文

    I’ve been thinking about this report more from last month… It’s probably extremely bullish for the Neocloud segment like $NBIS (and even $CRWV) more than markets think. Because of the hyperscaler buildout delay spillover. And partly why $META might have signed a $27B deal. Companies that already have secured capacity/component orders for their buildout are likely to be major winners as other competition stalls.

  30. 方法论 $SIVE$AEHR$AAOI$SOI$MRVL

    认为现在还很早,自己更偏好少数纯标的和更稳健的名字。

    还不算晚进场,实际上还非常早,正处在下一个拐点的起点。 如果让我现在选两只更小、更纯的标的,我个人会喜欢 $SIVE 和 $AEHR,再加一只 $AAOI 也许不错。 更稳一点的话,Win、$SOI、$MRVL 看起来都很扎实? 可选名字其实很多……

    英文原文

    Not really late too the party, it’s extremely early and at the beginning of the next inflection point. I personally like $SIVE, $AEHR if I had to pick two more smaller purer play companies around now, maybe throw in $AAOI on the next drop. For safer, Win, $SOI, $MRVL seem solid? Quite a lot of names…