全部推文

共 6749 条 · 第 89 / 225 页 · 滚到底自动加载更多

  1. 供应链分析 $AAOI

    说明 AAOI 做光器件的全链路业务,不会在转型中被替代。

    $AAOI 会先做激光 -> 设计 -> 组装 -> 再把收发器卖出去。所以它们把整条供应链都包了。 可插拔光器件不会消失,所以他们会填补那一块。 然后在另一边,他们还做 ELSFP(外部光源),所以在 CPO 侧就是激光 -> 设计 -> 封装。 也就是说,这两边他们都做,不会在行业转型里被替代掉。

    英文原文

    $AAOI does laser -> design -> assembly -> then sells the transceiver. So they have the whole supply chain. Pluggables won't go away so they'll fill that gap. Then on the side, they build ELSFPs (external light sources), so laser -> design -> packaging on the CPO side. So they do both. They don't get replaced in the transition.

  2. 供应链分析 $NVDA

    认为去年 NVDA 先把产能拿走了,时间大概是 2024 末或 2025 初。

    @LIama002 去年这事其实是在关门后发生的,当时 $NVDA 已经把所有产能都拿走了。 Trendforce 还发过文,说明这事在去年就更早发生了。 也许是 2024 年末,或者 2025 年初?

    英文原文

    @LIama002 It was behind closed doors last year with $NVDA securing all the capacity. Trendforce published that it happened earlier last year. Maybe late 2024, early 2025? https://t.co/WbwGaQ6Gwm

  3. 杂谈 $AAOI$AXTI

    说 AAOI 的波动很正常,经历过 AXTI 后更能适应。

    @MacadoMP 这对 $AAOI 来说就是正常波动。 你得慢慢习惯它,尤其是如果你以前做过 $AXTI,你对 +25% 到 -20% 这种波动就会比较麻木。

    英文原文

    @MacadoMP Just your normal volatility for $AAOI. It's something to get used to, especially if you've been in $AXTI before, you're conditioned to be fine with +25% -20% movements.

  4. 杂谈

    只有链接,没有正文。

    仅有链接。

    英文原文

    @2_dang https://t.co/hQsFrIrGJ9

  5. 认为市场低估了 Nvidia 投资下游后带来的上游供应冲击。

    市场漏掉了 Nvidia 投资的含义: 今天它向光学领域的 $COHR 投了 20 亿美元。 向光学领域的 $LITE 投了 20 亿美元。 今天又向光学领域的 $MRVL 投了 20 亿美元。 Nvidia 去年就用过同样的剧本。 他们意识到 800G/1.6T 可插拔器件的推进会耗尽全球 EML 供给。 所以他们接触了 $LITE、$COHR、Sumitomo,并提前预定了大部分产能。 我们已经在他们股价里看到了反映:自从那轮严重供给紧张以来,$LITE 已经涨了 955%。 现在我们看到同样的剧本在过去一个月里再次开始。 只是这次是一个新的架构。 而且这些交易都包含了几十亿美元的采购承诺和未来产能权利。 所以……下一步是什么?CW / EML 和 CPO 瓶颈。 Nvidia 只是更喜欢投资下游玩家,但真正的供应紧张发生在上游。 来自 $MTSI、$SIVE、$LITE、$COHR、Furukawa 和 Sumitomo 的激光供应商都在全速运转。 Win Semi、$TSEM、$GFS 这些晶圆代工也大概率在满负荷运转。 整条供应链都会受益(比如 $AEHR 的测试、$SOI 的衬底)。 但从晶圆代工到 ELS / CW 激光瓶颈的这两个环节,可能会是最大的受益者。 Nvidia 是下一轮发生什么的最大信号;现在只是等拐点到来的时间而已。

    英文原文

    The market are missing the implications from $NVDA investing: $2B into $COHR for optical. $2B into $LITE for optical. and $2B into $MRVL for optical today. Nvidia did this exact same playbook last year. They realized the push to 800G/1.6T pluggables would exhaust the global supply of EML. So they approached $LITE, $COHR, Sumitomo, and preallocated majority of production. And we've seen this reflected in their share price with $LITE rising 955% since the major supply squeeze. We're seeing the beginning of the same playbook happen now over the last month. Just for a new architecture, this time. As these deals included multibillion-dollar purchase commitments and future capacity rights. So... what's next? CW/EML and CPO bottlenecks. Nvidia just prefers to invest in downstream players. But the supply crunch happens upstream. Laser suppliers from $MTSI, $SIVE, $LITE, $COHR, Furukawa, and Sumitomo are on overdrive. Foundries from Win Semi, $TSEM, $GFS are likely on overdrive. The entire supply chain benefits (eg. testing from $AEHR, substrates with $SOI). But these two segments from foundries to ELS/CW laser chokepoints are likely to be the biggest beneficiaries. Nvidia is the biggest signal of what's coming next; it's just a waiting period for the inflection point to hit.

  6. 杂谈

    表示这是自己最自豪的作品之一。

    @semivision_tw 这是我最自豪的作品之一。

    英文原文

    @semivision_tw This is one of my proudest creations

  7. 个股论点 $NVDA$RPI

    分析RPI估值偏低,Forward P/E从20.6x降至9.5x,增长58%,AI代理硬件新机遇或带来结构性需求。

    $RPI 就 GAAP 前向预测来看确实看起来有点便宜。 - 2026 前向 P/E: $5.29 / $0.256 = 20.6倍 - 2027 前向 P/E: $5.29 / $0.400 = 13.2倍 - 2028 前向 P/E: $5.29 / $0.553 = 9.5倍 基于个人预测。尤其是当你在同比增长 58%(预期会再次上调,因 OpenClaw 的购买): 当以 2 倍前向市销率交易时,实际上并没有多少溢价。尤其是在 AI 代理硬件(agentic hardware)领域还有新的 TAM 扩张机会。 下行风险是,OpenClaw 只是短期趋势。但 Jensen 在 GTC 上将其验证为下一个重大趋势。 如果这是结构性需求,而 Raspberry Pi 是消费代理硬件领域的 $NVDA……以约 $1B 市值来看,可能是一个绝佳的机会。

    英文原文

    $RPI does look sort of cheap for GAAP forward projections. - 2026 Forward P/E: $5.29 / $0.256 = 20.6x - 2027 Forward P/E: $5.29 / $0.400 = 13.2x - 2028 Forward P/E: $5.29 / $0.553 = 9.5x Off personal projections. Especially when you’re growing 58% Y/Y (expect this to beat again due to OpenClaw buying): There’s not really much of a premium when it trades at 2x fwd/sales. Especially when theres new TAM expansion opportunities in the AI agentic hardware space. The downside risk is, OpenClaw is just a short term trend. But Jensen did validate it as the next major trend at GTC. If it is structural demand and Raspberry Pi is the $NVDA for consumer agentic hardware… could be a great opportunity at ~$1B MC.

  8. 杂谈 $NBIS

    说自己仍对 NBIS 被看多这件事有 PTSD。

    @0xjawor 我到现在对他当时看多 $NBIS 这件事还有 PTSD。

    英文原文

    @0xjawor I still have PTSD from when he was bullish on $NBIS

  9. 杂谈 $MRVL

    抱歉 MRVL 持有人。

    ****。抱歉 $MRVL 的持有人们。

    英文原文

    ****. Sorry $MRVL holders. https://t.co/JMZ16QJRsL

  10. 供应链分析 $AAOI$AEHR

    AEHR获光收发器厂商认证订单,处于类似AAOI早期阶段,面临量产拐点。

    $AEHR 从一家领先的光收发器(optical transceiver)公司获得了资格认证订单。 它正处于早期阶段,类似于 $AAOI 正在被主要超大规模云服务商(hyperscaler)供应链测试用于光收发器/硅光子学(silicon photonics)阶段。 在可能随时到来的大批量量产拐点(mass volume inflection point)之前。 $AEHR 盘前上涨 12.73%。

    英文原文

    $AEHR new qualification orders from a leading optical transicever company. It’s in the early $AAOI stage where it’s getting tested by major hyperscaler supply chains for optical transceivers/silicon photonics. Before the mass volume inflection point that may be at any time. $AEHR up 12.73% premarket.

  11. 个股论点 $RPI

    等财报验证 thesis 后,RPI 现在估值看起来还不错。

    @Test126863 是啊……我已经厌倦了所有噪音,所以就等财报来验证我的 thesis。 现在 $RPI 作为一家 fabless AI 硬件公司,大概只交易在 2 倍前瞻销售。

    英文原文

    @Test126863 Yeah… got tired of all the noise so just waited for earnings to validate my thesis. Now $RPI is trading at roughly 2 fwd p/s as a fabless ai hardware company.

  12. 方法论 $RPI

    回顾自己对 RPI 的收入预测比分析师更乐观,结果验证了。

    @AscendingLife3 是啊,我当时预计营收增长 55%,而分析师全都在说 14%。 结果 $RPI 最后做到了 58%。

    英文原文

    @AscendingLife3 Yeah I was projecting 55% revenue growth while analysts were all saying 14%. $RPI ended up hitting 58%.

  13. 个股论点 $SIVE

    认为 SIVE 在低市值下值得给更高溢价。

    @glfikk 也许应该给 100-200% 的溢价,因为 $SIVE 的市值实在太低了。

    英文原文

    @glfikk Maybe 100-200% premium since $SIVE MC is incredibly low.

  14. 业绩复盘 $RPI

    复盘 RPI 财报后上涨,并指出 2027 年估值仍有空间。

    thesis 验证正在发生。 $RPI 财报后上涨 44.76%。 Raspberry Pi 在 2027 年大概还有很长的路要走,市盈率可能在 11-13 倍左右。 2026 年大概在 19-22 倍。 在需求大幅跳升、前瞻增长从预期 14% 提高到 58% 的情况下…… Raspberry Pi 已经成了 OpenClaw 首选的、快速加速的 AI 编排硬件。

    英文原文

    Thesis validation in effect. $RPI up 44.76% after earnings. Raspberry Pi likely has a long way to go with some ~11-13x p/e projections for 2027. And ~19-22x for 2026. With a massive surge in demand to 58% fwd growth from 14% expected… Raspberry Pi has become the rapidly accelerating AI orchestration hardware of choice for OpenClaw.

  15. 供应链分析 $POET$SIVE$NVDA$MRVL

    认为 POET 受益,但 SIVE 对 Nvidia 和 Marvell 的 Celestial 更关键。

    @marioyep5 是的,$POET 确实是受益者,但 $SIVE 作为光源,对 $NVDA 和 $MRVL 的 Celestial 现在来说非常关键。

    英文原文

    @marioyep5 Yes, $POET is beneficiary but $SIVE is incredibly important to $NVDA and $MRVL Celestial now as the light source. https://t.co/mn3oESTKlw

  16. 供应链分析 $NVDA

    认为 Nvidia 正在强推 CPO 路线图并锁定主要参与者。

    @TheValueist 是啊,$NVDA 真的在强力推动他们的 CPO 架构路线图,而且正在把最大的几家玩家锁进自己的标准里。 我们正在实时看到光子学里下一步会发生什么。

    英文原文

    @TheValueist Yeah $NVDA is really pushing hard on their CPO architectural roadmap and locking in all the biggest players into their standards We’re seeing what’s coming next in photonics real time.

  17. 供应链分析 $SIVE$MRVL$AVGO

    认为 Nvidia 可能更希望 SIVE 保持独立,也可能出现股权收购或被 Broadcom 并购。

    @yuntungshieh 我觉得 Nvidia 可能更希望 $SIVE 保持独立。 所以我们可能会看到 $MRVL、AlChip 对它做股权收购……或者也可能被 $AVGO 收购,用来卡死竞争对手的 CPO 路线图。现在还不确定最后会怎么演。

    英文原文

    @yuntungshieh I think Nvidia would rather have $SIVE stay independent. So we might see some share acquisition from $MRVL, AlChip… or maybe an acquisition by $AVGO to chokepoint their competitors CPO roadmap. Not sure how this plays out.

  18. 供应链分析 $SIVE$AVGO$NVDA$MRVL

    调侃如果 Broadcom 买下 SIVE,就能控制后续多家公司的路线图。

    如果 $AVGO 真的花 3 亿美元买下 $SIVE,那就太可惜了…… 这样一来,它就能从上游光源层面控制 $NVDA 和 $MRVL 的近端路线图。 我相信博通肯定不会在想这个……对吧?

    英文原文

    It would be a shame if $AVGO bought $SIVE for $300m… Then controlled the near term roadmap for $NVDA and $MRVL from the upstream light source level. I’m sure Broadcom isn’t thinking about this… right? https://t.co/wrS431DG8n

  19. 供应链分析 $POET$SIVE

    认为 POET 受益巨大,但很多分析师忽略了其光源来自 SIVE。

    @InvestWithFelix 对,$POET 是巨大的受益者。它在 Celestial 体系里更广为人知。 但连分析师都会忽略,Poet 的 ELS 光源其实来自 $SIVE。 不过 Nvidia 对这个生态的验证,对整个链条都很重要。

    英文原文

    @InvestWithFelix Yep $POET is a massive beneficary. It’s more well known with Celestial. Even analysts miss though Poet’s light source for ELS comes from $SIVE. But huge for their whole ecosystem that Nvidia is validating it

  20. 供应链分析 $NVDA$SIVE

    把 Nvidia 备货 EML 产能的历史剧本,类比到下一代光子瓶颈。

    这就是 $NVDA 当年提前锁定全部 EML 产能时的那套剧本,在那之前超大规模客户那边就已经出现了巨大的供给冲击。 这也是为什么我一直在为下一处瓶颈里的 $SIVE 拉响警报。 因为 $NVDA 现在也在下一代架构上做同样的事。

    英文原文

    This is the exact playbook $NVDA did securing all the EML capacity before there was a massive supply shock for the rest of the hyperscalers. That’s why I’ve been sounding the alarm with $SIVE in the next chokepoint. Since $NVDA is doing the same with the next gen architectures.

  21. 供应链分析 $SIVE$NVDA$MRVL$POET

    认为 SIVE 已经进入 Nvidia 路线图,且市值仍很小。

    @degenlurker_ 有意思的是,$SIVE 现在已经被写进 $NVDA 的路线图里了,而市值还只有大约 3 亿美元。 只是现在还没人把 Sivers -> Poet -> Celestial -> Marvell 这种多级关系串起来,虽然 Nvidia 已经砸了 20 亿美元去加速他们的 CPO 路线图了。

    英文原文

    @degenlurker_ It’s funny $SIVE is now baked into $NVDA roadmap now at ~$300M MC. Just no one associates multi hop relations between Sivers -> Poet -> Celestial -> Marvell yet when Nvidia invested $2B into accelerating their CPO roadmap yet.

  22. 供应链分析 $NVDA$MRVL$SIVE

    Nvidia 对 Marvell 的 20 亿美元投资,也验证了 SIVE 的光源位置。

    刚刚消息出来:$NVDA 向 $MRVL 投了 20 亿美元。 这笔交易主要是为了“联合硅光工作”,也验证了 Marvell 和 Celestial AI 的光互连 CPO 路线图。 但……谁在给 $MRVL 的光源供电,这点市场可能漏了?(提示:$SIVE)

    英文原文

    Just in: $NVDA invests $2B in $MRVL. This deal is largely for “joint silicon photonics work” and validates Marvell and Celestial AI's optical fabric CPO roadmap. But… guess who powers the light source for $MRVL that markets might have missed? (Hint: $SIVE) https://t.co/9lqJLIx7nG

  23. 供应链分析 $AXTI$COHR$LITE$MRVL$RPI$SIVE

    以Marvell收购Celestial为例,分析CPO供应链中Sivers作为CW激光器关键瓶颈供应商的巨大TAM扩张潜力。

    看看$MRVL收购Celestial($SIVE是其~2.9亿美元的光源供应商)这个例子就知道了。Marvell以55亿美元收购Celestial(32.5亿美元+22.5亿美元里程碑付款)。2026/2027年预期:零收入,亏损5000万美元。2028年预期:5亿美元收入。2029年预期:10亿美元收入。这是巨大的TAM(市场规模)扩张和收入加速增长。Sivers只是Celestial之上的客户之一,此外还有O-Net、Ayar、其他未公开的CPO(共封装光学)厂商,Jabil负责桥接量产爬坡并赢得良率。那些对光子学毫无理解的散户……正在做的事相当于:因为Celestial在2025、2026年没有CPO收入,就把它建模为5000万美元。因为它目前处于净亏损状态,而不是着眼于2027年之后的大规模爬坡。$SIVE是我抢先布局下一个架构超级周期(位于最早期阶段)的首选标的,涉及ELS TAM扩张+CPO+连续波(CW)激光器瓶颈。如果你想找后期介入的标的,可以看$LITE,它已经涨了3800%。

    英文原文

    Just look at $MRVL Celestial ( $SIVE is the light source at ~$290m ) as an example. Marvell bought it for $5.5B. ($3.25B + 2.25B milestones). 2026/2027 expectations: 0 revenue, -$50m loss 2028 expectations: $500m revenue 2029 expectations: $1B revenue. This is massive TAM expansion and revenue acceleration. This is just one of Sivers customer on top of O-Net, Ayar, other undisclosed CPO players, with Jabil bridging volume ramp and Win with yields. Random retail users with no understanding of photonics… are doing the equivalent of modeling Celestial at $50m because of no CPO revenue in 2025 2026. And because it’s operating at a net loss instead of looking at the massive ramp 2027 onward. $SIVE is my #1 pick for frontrunning the next architectural supercycles at the very beginning for ELS TAM expansion + CPO + CW laser chokepoints. If you want a later stage pick, you can look at $LITE that already ran 3800%.

  24. 方法论 $AXTI$LITE$COHR$SIVE$MRVL$RPI

    批评外行和媒体拿财务指标误判光子学、CPO 与 RPI 这类高增长标的。

    请别再发消息问我那些随机散户或者记者对 $AXTI 或 $SIVE 的分析了。 对于 $AXTI: 大多数人根本不知道自己在说什么。 三星代工和 $LITE 的 InP fab 完全是供应链的不同环节,这就像把 NAND 和 DRAM 混成一个存储类别一样。 就算说 Sumitomo 会取代 AXT,也说明这些人根本不懂上游的开采 / 加工 / 坩埚瓶颈,而 InP 衬底级公司像 $COHR 或 Sumitomo 本来就不是干这个的。 对于 $SIVE: 这就是 2027 年以后 CPO 规模化放量的质变点。 我看到很多对光子学完全不懂的人,拿毛利率、烧钱速度,以及前几年的营收甚至今年的营收来分析。 可 2027 年以后才会迎来最大的一次架构拐点。 如果你看 $MRVL Celestial 的时间点,它们 2027 年开始放量,2028、2029 年还会指数级增长。 所以我才说 $SIVE 有一天能成为 100 亿美元以上的公司,因为我看的是前瞻增长,而且这是最早能切入的位置。 我做的事情就是找 alpha: 市场可能像那些把 $RPI 前瞻增速说成 14-17% 的分析师一样,把它错定价了。 然后我建模成 55%,结果事实是 58%。 一个是光子学瓶颈,自有其原因。 另一个是被设计进 Jabil 和 Celestial 的,自有其原因。

    英文原文

    Pls stop tagging me in stuff where random retail investors or journalists tries doing an analysis on $AXTI or $SIVE. For $AXTI: Majority of folks have 0 clue what they’re talking about. Samsung Foundry or $LITE InP fab are completely different parts of the supply chain and this is like conflating NAND and DRAM as memory. Even talking about Sumitomo displacing AXT means they don’t understand the upstream extraction/processing/crucible bottlenecks that InP substrate level players like $COHR or Sumitomo don’t do. For $SIVE: This is mass inflection point of CPO scale up 2027 onwards. I have idiots out there with no understanding of photonics looking at gross margin, burn rate, and former revenue numbers for previous years or even the current year. For the biggest architectural inflection 2027 onwards. If you look at $MRVL Celestial point its starting volume ramp 2027 and exponentially increasing 2028, 2029. That’s why I said $SIVE can be a $10B+ company one day since it’s looking at forward growth and this is the earliest out on. What I do is find alpha: This is what markets might have mispriced like all the analysts saying 14-17% fwd growth with $RPI. Then I modeled 55% and it turned out to be 58%. One is a photonics bottleneck for a reason. The other is designed into Jabil and Celestial for a reason.

  25. 方法论 $SIVE

    说自己被质疑,因为 CPO 放量是 2027 年而他们在提前布局。

    @JustinasRoland0 我之前把那人拉黑了,因为他根本不知道自己在说什么。 CPO 的放量是 2027 年,这就是我在提前布局;但还有一堆人还在试图拿新前沿技术(比如 CW 激光)以前的盈利和烧钱速度来建模。

    英文原文

    @JustinasRoland0 Blocked the guy earlier because they have no clue what they're talking about. CPO ramp is 2027, this is frontrunning it, and there's idiots out there still trying to model previous earnings and burn rate of new frontier technologies like CW laser tech.

  26. 杂谈 $RPI

    感谢支持,并表示对外界质疑的情绪只是发泄。

    @futureproof_inv 感谢支持! 越是你比别人更早看到一件事,收到的质疑和嘲讽就越多。 我最近因为 $RPI 这事从自己圈子外的人那里吃了太多烂话,所以这条更像是在发泄。

    英文原文

    @futureproof_inv Thanks for the support! It's always when you're earliest to something that you get the most criticism / doubt. I just got a ton of endless sht over $RPI from everyone outside my follower base, so this is just coming out as a rant.

  27. 业绩复盘 $RPI

    认为自己关于 RPI 的 thesis 被财报验证,媒体和分析师都看错了。

    甚至连普通分析师和记者都把 $RPI 叫成了“meme stock”。 我觉得这就是典型的事后偏差:我身边以外的大多数人都否定了这个 thesis。 不管怎样,财报终于验证了它。58% 的同比增长非常出色,我只是一直憋着这口气等财报出来。

    英文原文

    Even regular analysts called $RPI a "Meme Stock" on top of reporters. Think this is hindsight bias, vast majority of people outside my following dismissed the thesis. Regardless, finally vindicated from earnings. 58% Y/Y growth is phenomenal, just pent up frustration waiting for this report to come out.

  28. 个股论点 $GME$RPI

    批评媒体将$RPI误标为散户股,强调其58%增长修正后的基本面价值。

    @GeneralMarketx FT和其他新闻媒体那些该死的家伙在用虚假标签制造恐慌。说这是$GME的“散户股领地”而非实际基本面。$RPI在58%的增长预估修正后可能会被大幅重新定价上涨。

    英文原文

    @GeneralMarketx Actual ****s over at FT, and other news outlets sparking panic with the false labels. Saying it was $GME "Meme stock territory" over actual fundamentals. $RPI is likely going to be repriced upward a lot after the 58% growth estimate revisions.

  29. 个股论点 $AXTI$SIVE

    认为被叫 meme stock 的标的最后往往成了大公司,SIVE 会是下一个。

    @innovavest 我不知道。所有被散户和媒体贴上“meme stock”标签的我的票,最后都变成了像 $AXTI 这样的十亿美元级公司。 $SIVE 会是下一个,我很有信心。

    英文原文

    @innovavest Idk. All the retail and news "labelled meme stocks" for my picks ended up becoming billion dollar companies like $AXTI. $SIVE is next, have high confidence.

  30. 业绩复盘 $RPI

    说自己正确看到了 RPI 的基本面变化,而外界曾嘲笑它是 meme。

    @QGrowthCap 我就是很烦,X 和社交媒体上所有人都把它叫成“meme stock”。 可我说 $RPI 的基本面在变化,而且我判断对了。 现在他们都闭嘴了,然后才去更新自己对这只 AI fabless 股票的模型,而它当时的前瞻市盈率只有 19 倍。

    英文原文

    @QGrowthCap Just frustrating everyone on X and social media called it a "meme stock" When I called fundamental changes on $RPI and got it right. Now they're all silent, and then just updated their models on an AI fabless stock trading at fwd 19 P/E.