全部推文
共 6775 条 · 第 196 / 226 页 · 滚到底自动加载更多
-
对比CIFR与IREN，认为前者估值合理，后者合同条款差且利润率低。
对于 $CIFR，当前盈利无关紧要，市场定价基于远期盈利。乍看之下，除了通过 Fluidstack 现有的 $GOOGL 合作外，AWS 的交易看起来棒极了。因此其 20% 的涨幅实至名归。 另一方面，$IREN 的合同细看之下越来越糟。 营收不重要，重要的是该营收带来的自由现金流(FCF)/利润率，尤其是考虑到他们正在花费 58 亿美元购买 GPU。我看过一些保守情景测算，显示其年化营业利润仅约 9800 万美元/年，这真的太低了。 即便乐观估计，毛利率也在 30% 以下甚至 20% 出头，而非其他人引用的 90%+ 的高性能计算(HPC)板块。这才是规模化后的利润率现实。 $IREN 是在购买资产，而非人们一直声称的提供托管服务(colocation service)。 $NBIS 与微软(MSFT)的交易在标准化至 300MW 后，其盈利能力和毛利率远优于 $IREN。当然，我还需要花更多时间进行建模。
英文原文
For $CIFR, doesn't matter about current earnings, market prices in forward earnings. And at first glance AWS deal looks amazing on top of their existing $GOOGL deals through Fluidstack. So well deserving of their 20% bump. For $IREN on the other hand, the contract worse and worse when you examine it more. Revenue does not matter. FCF/margins from that revenue is what matters, especially since they're spending $5.8bn GPUs. I've seen some conservative-case calculations where they generate ~Annualized operating profit 98m+/year, which is really low. Even so optimistically, gross margins are in the <30s or even 20s, and not the 90%+ HPC segments that others are quoting. This is the reality of margins at scale. IREN is buying the asset, rather than the colocation service people kept claiming $IREN would do. $NBIS deal with MSFT was extremely better than IREN’s in terms of profitability and margins when normalized to 300 MW. But of course I need to spend more time on modeling.
-
对比NBIS与IREN微软合约，前者利润率更高且能产生自由现金流。
经过更深入的研究，$NBIS 与 $MSFT 的交易在标准化至 300 MW 后，其盈利能力和利润率似乎远优于 $IREN。 这不仅仅关乎营收，因为如果不能产生自由现金流(FCF)，营收就没那么有用。 对于 $IREN： 总合同价值 97 亿美元 GPU 资本支出(Capex) 58 亿美元，包括辅助设备 ~38% 的毛利率。 $NBIS 51.76%+ 的毛利率。 我需要做更多计算来验证，这只是在交易差异方面的一次快速深度研究计算。
英文原文
After looking into it more, looks like $NBIS's deal with $MSFT was largely better than $IREN's in terms of profitability and margins when normalized to 300 MW. It's not purely about revenue, since if you're not generating FCF from it, it's not that useful. For $IREN: $9.7bn total contract value $5.8bn GPU capex, including ancillaries ~38% gross margin. $NBIS 51.76%+ gross margin. I need to do more calculations to verify, this was just a quick deep-research calculation in terms of deal differences.
-
IREN/CIFR涨幅回吐，NBIS回调提供买入良机。
哇，$IREN 和 $CIFR 刚刚抹去了其涨幅。至少对于 $IREN 而言，我想市场已经提前计入了超大规模云服务商(hyperscaler)的合同预期。 不过 $CIFR 的 AWS 公告确实让所有人都大吃一惊。 $NBIS 随大盘下跌，但鉴于整个 Neocloud 行业因超大规模云服务商合同接连公布以及财报季(ERs)期间前景极为看涨，这 -5% 的回调买入机会极好。
英文原文
Wow $IREN and $CIFR just wiped out its gains. For $IREN at least, I guess the market was pricing in a hyperscaler contract already. $CIFR AWS announcement was a really big surprise to everyone though. $NBIS just dropped with the rest of the market but an extremely good -5% dip buy since the whole Neocloud industry is really bullish with hyperscaler contracts announced left and right + during ERs.
-
IREN完成重大去风险事件，值得重估，需时间建模。
@moninvestor 就像人们常说的手越大越好（~190亿美元 $NBIS 对比 ~90亿美元 $IREN）。 半开玩笑地说，这对 $IREN 来说是一个实质性的去风险事件，值得重新评级。我需要更多时间来进行估值。 （这是制作层级列表的人，如果其他金融X用户想做一个的话） https://t.co/oefhQy7Trn
英文原文
@moninvestor Like what they say with hands, the bigger the better (~$19B $NBIS vs ~9B $IREN). Half-joking aside, this was a material de-risking event for $IREN and deserves re-rating. Will need more time to do valuations (This was the tierlist maker if any other finx wanted to make one) https://t.co/oefhQy7Trn
-
Neocloud获微软AWS大额订单，算力需求旺盛，板块极度看涨。
Neocloud 名单已更新： $IREN - IREN 与微软达成 97 亿美元 GPU 云合同，包含 20% 预付款，将在五年内部署 NVIDIA GB300。 $CIFR - CIFR 与亚马逊云科技 (AWS) 达成 55 亿美元、15 年的租赁协议。 随着超大规模云服务商 (Hyperscalers) 的新合同涌入，Neocloud 板块变得极度看涨，这显示了对算力 (Compute) 的无限需求。 我在原始论点中预计 $ORCL、$META 和 $AMZN 会达成交易，看来 AWS 已与 $CIFR 入场，预计与其他 Neocloud 的交易也会接踵而至。$META 最近也与 $CRWV 签署了 140 亿美元的交易，从其不断扩大的 AI 资本支出 (Capex) 来看，我们可能会看到更多与 Neocloud 的交易。 这对整个板块来说都是极度看涨的。 我附上了我的个人梯队排名 (Tierlist) 供娱乐，但随着有更多时间分析交易和利润率，数据中心/矿工板块可能会迎来全面重评。
英文原文
The Neocloud List got updated: $IREN - IREN secured a $9.7B GPU cloud contract with Microsoft, including 20% upfront prepayment, to deploy NVIDIA GB300s over five years. $CIFR - CIFR secured a $5.5 billion, 15-year lease agreement with Amazon Web Services (AWS). The Neocloud sector just got insanely bullish with new contracts coming in from hyperscalers and this just shows insatiable demand for compute. I expected $ORCL, $META, and $AMZN to make deals in the original thesis and looks like AWS entered the fray with $CIFR, with likely more deals incoming with other Neoclouds. $META also signed a $14B deal with $CRWV recently and from their ER scaling AI Capex, we'll likely see other deals with Neoclouds. This is insanely bullish across the whole sector. I attached my personal tierlist for fun, but we'll likely see a re-ratings in data centers/miners across the board when there's more time to analyze deals and margins.
-
维持NBIS看涨目标，看好新云服务商整体利好及NBIS高非对称性。
我仍维持 $NBIS 1年目标价 $400 的看涨观点，但逢低平均成本是个好主意。 $IREN 今天刚获得 $MSFT 的 90 亿美元订单，$GOOGL 也刚刚宣布融资，这对所有新云服务商(Neoclouds)来说总体是利好。 几家挖矿公司因该交易大幅降低了风险，但我仍认为 Nebius 具有最高的非对称性(Asymmetry)。
英文原文
I still maintain $NBIS $400 bull case 1Y PT but it’s good to cost average. $IREN just got a $9B deal from $MSFT today and $GOOGL just announced financing so it’s generally bullish for all Neoclouds. Several miners got de-risked a lot today from the deal, but I’d still say Nebius has highest asymmetry.
-
质疑Meta用户数据，建议资本支出聚焦基础模型竞争。
@doctarzz @zackslab 是的，我同意他们可能稍微夸大了10亿月活跃用户(MAU)的数据。 作为投资者，如果资本支出(capex)能用于在基础模型层面与ChatGPT/Gemini竞争，我会更高兴，因为FAIR/Meta超级智能正在做一些相当酷的事情。 我们很可能会很快看到他们推广独立的AI应用。
英文原文
@doctarzz @zackslab Yeah I'd agree they probably inflated 1B MAU a bit. I'd be happier as an investor if capex went to compete vs ChatGPT/Gemini on the foundational model side since FAIR/Meta Superintelligence is doing some pretty cool stuff. We'll likely see them promote the standalone AI soon
-
Meta 资本支出聚焦前沿模型训练，以应对 AI 竞争。
撇开这点不谈，我同意我认识的人都不会使用 Messenger/WhatsApp AI，虽然不清楚他们如何获得 10 亿月活用户(MAU)，但这只是轶事。Facebook 正在消亡的感知(perception)与现实(reality)略有不同。 资本支出(Capex)很可能流向其前沿模型(frontier model)（例如 $14B crwv 合同和其他数据中心），以便训练更好的模型并与 Gemini/ChatGPT 竞争。
英文原文
That aside, I'd agree nobody I know would use messenger/whatsapp AI, no clue how they got 1B MAU but it's just anecdotal. Perception like Facebook dying is slightly different than reality. Capex is likely going toward their frontier model (eg. $14B crwv contract and other datacenters) so they can train a better model and compete vs. Gemini/ChatGPT.
-
转发推文，无具体信息。
@doctarzz @zackslab https://t.co/UV78P38WrP
英文原文
@doctarzz @zackslab https://t.co/UV78P38WrP
-
META财报后算法抛售属非理性，基本面强劲，正逢低买入。
就 $META 而言，我的观点是：每股收益(EPS) 1.05美元对比预估 6.67美元，是财报发布后算法抛售的主要触发因素，因为他们未计入 7.25美元的一次性税收细节。做市商(MM)的期权相关资金流和机械对冲触发 -> 做空股票以保持对冲，并引发次日抛售+散户清洗。每个人都需要一个“叙事”来解释这次下跌，所以这次是“天哪 $META 在AI资本支出(capex)上烧钱！”，尽管这很愚蠢且业绩其实很棒。下跌是算法性的，基本面不仅完好，而且好于预期。当然人们阅读叙事导致更多恐慌性抛售，但一旦消化信息，他们会发现抛售很愚蠢。所以我正在利用这次下跌。最好忽略媒体叙事，关注实际存在的情况，就像 $GOOGL 在 145美元时那样。我们可以争论 $META 的AI是否糟糕以及资本支出是否在全行业陷入黑洞，但基本面完好且在增长。
英文原文
So with $META at least, my take is that EPS $1.05 vs est. $6.67 was the primary trigger algorithmic selloff immediately after earning report since they didn't get the $7.25 EPS one-time tax nuance. Option related flows and mechanical hedging by MMs triggered -> short stock to stay hedged and there's day 2 selling + retail flush. Everyone needs some "narrative" to explain that drop so in this case, it was "omg $META spending money on capex on AI!", even though it's just stupid and earnings were amazing. Drop was algorithmic and fundamentals are not just in-tact, they're better than expected. Of course people read the narratives and this leads to more panic-selling, but once people digest the information, they'll see the selloff is stupid. So I'm taking advantage of the drop. Better to ignore the media narrative and go with what's actually there like with $GOOGL at $145. We can have another argument whether or not $META AI is crap or not and capex is going into a black hole across the board but fundamentals are intact and growing.
-
地缘风险被高估，美国必护台以保AI供应链核心。
关于 $TSM - 说得好。从美国投资者的角度来看，台积电（$TSM）在“台湾可能被入侵”这件事上，总是一个令人头疼的逆风因素。不过，这同时也是一个过于关键的国家安全议题。我相信，美国若有任何国家威胁到台湾，一定会出兵干预。原因很简单——台湾是整个供应链和人工智能基础设施的核心。而且，我也相信，没有任何国家愿意与美国直接开战。当涉及关键矿产贸易争端时，台积电和台湾方面也都会发表声明，表示他们的供应链并未受到影响。这就是为什么地缘政治风险并不像市场定价那样高的原因。
英文原文
For $TSM - 说得好。 从美国投资者的角度来看，台积电（$TSM）在“台湾可能被入侵”这件事上，总是一个令人头疼的逆风因素。 不过，这同时也是一个过于关键的国家安全议题。我相信，美国若有任何国家威胁到台湾，一定会出兵干预。原因很简单——台湾是整个供应链和人工智能基础设施的核心。 而且，我也相信，没有任何国家愿意与美国直接开战。 当涉及关键矿产贸易争端时，台积电和台湾方面也都会发表声明，表示他们的供应链并未受到影响。 这就是为什么地缘政治风险并不像市场定价那样高的原因。
-
看好ALAB/CRDO连接技术爆发，建议持有至明年，提及MINT非对称收益。
@soulbiri1 $ALAB 和 $CRDO 上涨 15%+ 是意料之中的，因为随着数据中心建设，连接技术（Connectivity）正像内存（Memory）$MU/三星海力士那样疯狂爆发。 我个人会持有它们到明年，也许稍微减仓一点点。 $MINT 只是非对称上行（Asymmetrical Upside）。致 @Neat_Lama，这是专辑的第二首歌。https://t.co/yvd68BBX0R
英文原文
@soulbiri1 $ALAB + $CRDO 15%+ increase was expected due to connectivity going insane like memory $MU/Sk Hynix on data center buildout. I'd personally hold them into next year, maybe trim a tiny bit. $MINT was just asymmetrical upside. To @Neat_Lama, here's another song #2 of the album. https://t.co/yvd68BBX0R
-
推荐年末满仓MINT组合（META/IBIT/NBIS/TSM），利用错杀与高非对称回报。
只有四只股票： 🍬 MINT 拥有极其高、非对称的回报。 意味着低资本风险，高上行回报。 这仅适用于2025年11月和12月。 就是它... MINT 关于为何在2025年底满仓MINT的解释： M = $META I = $IBIT N = $NBIS T = $TSM 将这些神圣的成长型公司纳入你的投资组合，你可以让呼吸更清新，从而吸引Sydney Sweeney： Mag7。加密货币。AI数据中心。半导体。 $META - Meta ___________ 愚蠢的抛售。 营收超预期。每股收益超预期。强劲指引。 当你看到营收512.4亿美元（同比+26.25%），净利润27.1亿美元（同比-82.73%）时，纸面上看起来极其糟糕。 🟢 营收：512.4亿美元 vs. 预期494.1亿美元 剔除一次性税收影响 → 🟢 每股收益：7.25美元 vs. 预期6.67美元 包含一次性税收影响 → 🔴 每股收益：1.05美元 vs. 预期6.67美元 我猜测最初的抛售是由算法驱动的，随后由机械性资金流引起。许多高频交易(HFT)和新闻解析系统会对直接从数据源（彭博、路透等）拉取的头条每股收益数字做出反应。 如果报告的GAAP每股收益（1.05美元）与共识预期（6.67美元）并列显示，系统会将其标记为大幅不及预期。显然这是推测性的，但许多模型并未编程以归一化处理一次性税收费用（这是新的），也不会知道去查看实际超预期的调整后每股收益（7.25美元）。 现在新闻声称抛售是由于AI资本支出惊喜所致。如果接受这个说法，因“AI资本支出增加”而下跌15%+纯粹是叙事，毫无意义，因为$MSFT、$GOOGL、$AMZN、$TSLA以及每一个Mag7都在增加资本支出。 $META做同样的事却遭到抛售说不通，除非人们认为它只是进入一个没有投资回报的黑洞（市场不同意，因为AI基本上在推动$SPY）。 一旦周末有时间让人们消化实际发生的情况，恢复将是极强的买入机会。误读+愚蠢的叙事可能导致短期过度反应（1-2个交易日），通常会被逆转。 $IBIT - 比特币 ___________ 它是比特币。目前位于11万美元。 随着美国政府的支持，以及持续的货币通胀+资金流向新一代人，没有比BTC更好的通胀对冲工具了。 Polymarket仍然定价比特币在2个月内上涨40%至13万美元，即使它上涨几个百分点，你也可以在加密货币和底层资产上获得上行空间。 $NBIS - Nebius ___________ 核心业务估值：310-365亿美元 分部加总（现金、资产、投资组合）：~106亿美元（削减40% = ~63亿美元） 基准情景估值：~390亿美元 → 即使零增长也有高上行空间 核心业务有望在FY26-27实现~50亿美元年度经常性收入(ARR)，60-70%毛利率，30%息税折旧摊销前利润(EBITDA)目标。 通过再达成一个超大规模云厂商协议+SMB扩张，实现80-120亿美元ARR可能证明~1000亿美元估值合理。 无有毒债务，高GPU利用率，全栈架构 → 更高的经营杠杆。 你可以看我引用的要点比较，以便你对$NBIS vs $IREN做出自己的判断。 然而，我将始终维持$NBIS拥有包括$IREN到$CIFR在内的最高非对称回报，因为现有的超大规模云厂商协议、企业客户（Shopify, Accenture）、政府客户、1GW容量、来自全栈的高毛利率等。 这并不是说来自$WULF的Fluidstack + Google协议以及其他容量方面没有高上行空间。 Nebius的下行风险最低。而且它真的、真的有高上行空间。 $TSM ______________ 它是AI交易的中心。 他们已经指引利润率增加——这是天文数字般的。 营收：331亿美元 vs. 315亿美元预期。 每股收益：2.92美元 vs. 2.59美元预期。 指引：322-334亿美元 vs. 320亿美元预期。 他们超收了营收指引，利润率增加，他们印钱的速度真是天文数字。 财报前交易在305-310美元，现在回到300美元。 这是一家万亿+美元的公司，同比增长30%+且利润率提升——到底怎么回事。 我之前在$GOOGL 145美元和$HOOD 20美元时说过同样的话。他们印钱。 没必要辩论$ALAB vs $CRDO，$IREN vs $NBIS，$AMD vs $NVDA。 $TSM是整个AI建设周期的中心，所以这只是等待游戏。 _ 就是它$MINT。 享受歌曲的预告，有点炸。🎶
英文原文
There’s only Four stocks: 🍬 MINT That have ABSURDLY HIGH, asymmetrical return. Meaning Low Capital Risk. High Upside Return. This is only for Nov + Dec 2025. Here it is... MINT An explanation of why to full port MINT for end of year 2025: M = $META I = $IBIT N = $NBIS T = $TSM With these four holy growth companies in your portfolio, you can make your breath smell better so you can pull Sydney Sweeney: Mag7. Crypto. AI Data Center. Semiconductor. $META - Meta ___________ Stupid sell-off. Revenue Beat. EPS Beat. Strong guidance. It looks extremely bad on paper when you look at Revenue $51.24B (+26.25% Y/Y), Net income $2.71B (-82.73% Y/Y). 🟢 Revenue: $51.24B vs. est. $49.41B Without one-off tax → 🟢 EPS: $7.25 vs. est. $6.67 With one-off tax → 🔴 EPS: $1.05 vs. est. $6.67 My guess was the initial sell-off was algorithmically driven, then caused by mechanical flows. Many high-frequency trading (HFT) and news-parsing systems react to headline EPS numbers pulled directly from data feeds (Bloomberg, Reuters, etc.). If the reported GAAP EPS ($1.05) is printed alongside a consensus of $6.67, the system flags it as a massive miss. Obviously this is speculative, but many models aren't programmed to normalize for one-time tax charges (which is new) and won’t know to look at adjusted EPS ($7.25), which actually beat. Now news are claiming the sell-off was due to AI capex surprise. If we go with this, -15%+ on “AI capex spending increasing” is just pure narratives and means absolutely jack since $MSFT, $GOOGL, $AMZN, $TSLA, and every single Mag7 is increasing on capex. $META doing the same and selling off doesn’t make sense, unless people believe it’s just going into a black hole with no ROI (which markets disagree since AI is basically running $SPY). Extremely strong buy on recovery once there’s time over the weekend for people to digest what actually happened. Misreads + stupid narratives can cause short-term overreactions (1–2 sessions), often reversed. $IBIT - Bitcoin ___________ It’s Bitcoin.And sitting at $110k. Bitcoin is here to stay with US government support, and with all the ongoing currency inflation + flow of money going to newer generation, there’s no better hedge against inflation than BTC. Polymarket still prices Bitcoin 40% to $130k in 2 months’ time, even if it moves up a few percent you can have upside on CCs and upside on the underlying asset. $NBIS - Nebius ___________ Core business valuation: $31–36.5B Sum-of-parts (cash, assets, portfolio): ~$10.6B (slashed 40% = ~$6.3B) Base case valuation: ~$39B → high upside even with zero growth Core business on track for ~$5B ARR (FY26–27), 60–70% gross margins, and 30% EBITDA target. Hitting $8–12B ARR (via one more hyperscaler deal + SMB expansion) could justify ~$100B valuation. No toxic debt, high GPU utilization, full-stack architecture → higher operating leverage. You can look at the bullet point comparison that I quoted so you can make your own judgement on $NBIS vs $IREN for example. However, I’ll always maintain $NBIS has the highest asymmetrical return over anything including $IREN to $CIFR due to existing hyperscaler deals, enterprise clients (Shopify, Accenture), Government Clients, 1GW capacity, high gross margins from full-stack, and many others. That’s not saying there’s not high upside from $WULF from Fluidstack + Google deals, and others from capacity. The downside risk on Nebius is the lowest. And it has really, really high upside. $TSM ______________ It’s the CENTER. OF. THE. AI. TRADE. They’ve already guided INCREASING MARGINS - which is astronomical. Revenue: $33.1B vs. $31.5B est. Earnings per Share: $2.92 vs. $2.59 est. Guidance: $32.2B – $33.4B vs. $32.0B est. They’ve beat revenue guidance, increasing margins, and it’s just astronomical how much money they’re printing. It was trading at $305–310 pre-earnings, now back at $300. It’s a $1T+ company that grew 30%+ Y/Y with their margins — what the actual. I said this about the money printer $GOOGL back at $145 and $HOOD back at $20. They print money. There's no need to debate $ALAB vs $CRDO, $IREN vs $NBIS, $AMD vs $NVDA. $TSM the center of the whole AI buildout, so it’s just a waiting game. _ There is it $MINT. Enjoy a teaser of the song, kinda a banger. 🎶
-
澄清对IREN的立场，指出其实际利润率低于宣传值，强调权衡取舍。
另外，当我发布关于 $IREN 和 $NBIS 的对比，或谈论 $IREN 拥有与 $RIOT 和 $MARA 相当的能力，并指出规模化后的利润率比算力更重要时，我并非看空 IREN，只是陈述现实情况。 同样，当人们声称他们拥有 92% 的毛利率时，如果你计入折旧和其他费用，事实并非如此。他们只是进入了另一个类别，如果你按照其他高性能计算(HPC)公司核算毛利率的方式进行标准化处理，利润率会低得多。 但这并不意味着情况糟糕，总是存在权衡取舍。
英文原文
Also when I post about comparisons between $IREN and $NBIS or talking about $IREN having comparable capacity to $RIOT and $MARA, and saying margins at scale matter more than compute, I'm not bearish on IREN, just giving the reality of the situation. Same when people try claiming they have 92% gross margins, they don't when yo factor in deprecation and other expenses. They just went into another category and if you normalize it with how other HPC companies do accounting for their gross margins, margins are a lot less. But that doesn't mean it's bad, there's always tradeoffs.
-
分析超大规模云服务商自建ASIC对NBIS等公司的长期下行风险及合约保护机制。
是的，长期来看，超大规模云服务商自建ASIC和产能，且在合约容量结束后不再使用Neoclouds，这是$NBIS等公司的下行风险。我在最初的投资逻辑中提到，$CRWV、$NBIS几乎充当$NVDA的“云业务部门”，有助于抵御来自超大规模云服务商的利润率压缩并实现业务多元化。这就是为什么超大规模云服务商将收入导向这些小型公司。但总体而言，如果你的合约条款是10年或15年，你会拥有更多保护，但收入增长幅度较小（例如$NBIS与$MSFT的交易，收入在1-2年内从约11亿美元ARR增长至38亿美元以上ARR）。话虽如此，那是4-5年后的事，目前推测为时过早。通常你会根据新信息的出现，按月进行重估或审视风险与回报。
英文原文
Yeah longer term that's the downside risk for $NBIS and others in terms of hyperscalers building out ASICs and capacity and not using Neoclouds after contracted capacity ends. I said in my original thesis that $CRWV, $NBIS and almost serve as $NVDA's cloud arm, and helps defend against margin compression from hyperscalers and diversify their business. That's why hyperscalers funnel revenue down to these small guys. But generally, you have more protection if your contract terms are 10Y or 15Y but less revenue ramp. (eg. with $NBIS $MSFT deal, revenue ramps to $3.8B+ ARR in 1-2 years on top of their ~1.1B ARR) That being said, that's 4-5 years from now, it's too far forward to speculate as of now. Usually you do things like re-rate or examine risk vs. reward month-by-month whenever new information comes out.
-
吐槽 X 平台 Bug 导致评论回复干扰主页信息流。
@DeepValueBagger 这太让人分心了，X 平台有 Bug。通常情况下，除非你单独回复自己的原帖，否则对评论的回复不会出现在你的主页信息流中。https://t.co/4X6DPgHXOH
英文原文
@DeepValueBagger This was too distracting, X is buggy. Normally replies to comments don't show up on your main feed unless you reply to your original post by itself. https://t.co/4X6DPgHXOH
-
对比不同数据中心租赁期限对收入爬坡及长期竞争力的影响。
我删除了之前的评论，因为 X 平台有 Bug 导致时间线混乱： 简而言之： $APLD 建设具有显著电力容量的数据中心，然后将基础设施租赁给客户。目前有 $110 亿的 $APLD 和 $CRWV 针对 Ellendale 园区的 15 年租赁协议。 Polaris Forge 2 约为 200+ MW，合同价值 $50 亿。从交易者视角来看，我不太看好较长的合同期限（例如 $CIFR | $WULF 的 10 年），相比 $NBIS | $MSFT 的 5 年合同，因为前者收入爬坡较慢且成本存在变数。但这并非真正的利空，因为如果你投资的是 1 年周期，你会希望收入爬坡尽可能快。 但如果看 FluidStack $WULF 的 15 年合同协议，它为公司在长期内提供了更多保障；反之，如果像 $MSFT 这样的超大规模云厂商最终自建容量和定制 ASIC，那么一些 Neocloud 公司在 4 年后可能会陷入困境。
英文原文
I'm deleting my earlier comment to this since X is bugged and it's messing up my timeline: But TLDR from before: $APLD builds data-centers with significant power capacity and then lease that infrastructure to customers. So there's the $11B $APLD, $CRWV lease right now over 15 years for Ellendale campus. Then there's Polaris Forge 2 is ~200+ MW with a $5B contract. Not much of a fan of longer contract terms (eg. 10 Years $CIFR | $WULF), vs. 5 Year $NBIS | $MSFT, since slower revenue ramp and changes of costs, but this is coming from a trader's perspecitve. Not as an actual negative when I talk about longer contract terms since if you’re investing for 1Y timeframe, you’d want the fastest revenue ramp possible. But if you look at fluidstack $WULF contract agreements eg. 15 years it provides a lot more guarantees for a company over time, vs. if hyperscalers like $MSFT end up building out their own capacity and custom ASICs then some Neoclouds might be in trouble after 4Y
-
投资无绝对优劣，需基于当前市值权衡风险与回报。
@Antonzeehow 其实并没有绝对的从优到劣排名。更多是基于当前信息，在风险与回报之间的权衡取舍。 别太当真，这只是我基于当前市值(Market Cap)的个人直觉。 https://t.co/RiCqPEOANm
英文原文
@Antonzeehow There's not really a best to worst. It's more of trade-offs from risk vs. reward given current information. Don't take this too seriously, but this is just my personal feeling right now given current MC's. https://t.co/RiCqPEOANm
-
博主调侃全仓虚值期权若成功能否接任温迪董事长。
@StudentToStreet @moninvestor @Wendys @DeepValueBagger 你们能问问，如果我在财报发布时全仓押注温迪(Wendy's)的周权，行权价虚值125%且真的打中了……我能接任董事长职位吗？ 我一个月只吃一次快餐，所以我对这个领域有着深刻的专业知识。
英文原文
@StudentToStreet @moninvestor @Wendys @DeepValueBagger Can you ask if I full port Wendy's weeklies on earnings, 125% OTM and it hits.. Can I take over the Chariman's position? I eat fast food like once a month, so I have great domain knowledge.
-
看好WEN，认为其垃圾桶后有巨大未变现收入潜力。
@moninvestor 那肯定是 @Wendys $WEN。垃圾桶后面有着巨大的、未被开发的、未变现的收入潜力。我认为 @DeepValueBagger 在他最近的帖子中也意识到了这一点。
英文原文
@moninvestor It’s gotta be @Wendys $WEN. Huge untapped, unmonetized revenue potential behind the dumpsters. I think @DeepValueBagger realized this too on his most recent post.
-
对比ASML、TSM和AMD，认为TSM在非对称回报上优势巨大。
@Docofgothm @pepemoonboy $ASML 追踪晶圆厂周期。$TSM 是对 AI 整体建设情况的衡量。在这种情况下，没有理由选择 ASML。就纯上涨空间而言，此时 $AMD 可能胜出，但在非对称回报（同等上涨空间下下行风险最低）方面，$TSM 以巨大优势胜出。
英文原文
@Docofgothm @pepemoonboy $ASML tracks fab cycles. $TSM is a measurement of AI buildout as a whole. No reason to go with ASML in that case. In terms of raw upside, maybe $AMD wins at this point, but asymmetrical returns (lowest downside for upside) $TSM wins by an enormous margin.
-
TSM凭借AI基建垄断地位可随意提价，无视下游竞争。
$TSM 可以随时提价以获得更高利润率（他们预测增长 3%），而 $NVDA 无论如何都将被迫使用其服务。此外，盈利预测如预期般大幅超预期，整个 AI 建设都建立在 $TSM 之上，它并不关心 $AMD 与 $NVDA 谁胜谁负。纯粹的垄断。
英文原文
$TSM can raise their prices anytime for higher margins (which they projected 3% increase) and $NVDA will be forced to use them anyway. Also earning projections were a blowout as expected, whole AI buildout is on top of TSM, it doesn’t care about who wins like $AMD vs $NVDA. Pure monopoly.
-
分析AI基建公司融资方式、股权稀释风险及现金流依赖。
你说得对，低市值（MC）公司的风险比人们预期的要大，我会在下一篇帖子中指出其中的权衡取舍。 即使是市值较大的公司也会稀释股权，但稀释结构极其重要，就像 $CIFR 和 $NBIS 发行可转换债券（convertibles）那样。 有些公司主要通过资本支出（capex）进行融资，比如 $CORZ，但大多数公司（如 $NBIS）的目标是通过 $MSFT 的交易产生的自由现金流（fcf）来为资本支出融资。只是在初始建设阶段，像 $SLNH 这样的公司需要弄清楚如何为项目管道或 GPU 收购获取资金。
英文原文
You’re right there’s a lot more risk involved than people expect for lower MC ones and I’ll point off the tradeoffs in the next post. Even bigger ones dilute but dilution structure is extremely important like how $CIFR, $NBIS did convertibles. Some fund through primarily through capex like $CORZ, but the goal is for most like $NBIS to fund capex through fcf with the $MSFT deal. Just in initial buildout like $SLNH they need to figure out how to get funding for pipelines or GPU acquisitions.
-
博主分享持仓相似性并询问对方看好NBIS的理由
@pepemoonboy 哈哈，自从我前几天加仓 $META 后，我们的多头持仓现在很相似。我只是在 $TSM 上押注更重，而不是 $NVDA。是什么让你确信 $NBIS？
英文原文
@pepemoonboy lol we have similar longs now that I added $META the other day. I just went harder on $TSM instead of $NVDA. What convinced you on $NBIS?
-
补充 $CORZ 分析：Coreweave 长期合同改善现金流并资助资本支出。
你说得对，我漏掉了 $CORZ。感谢指出，也感谢 @cazenove_uk 等人的提醒。 鉴于 $CRWV 的合并前几天已终止，现在应将其视为独立实体看待，不知怎么竟疏忽了。我会在下一篇 2/2 帖子中补充。 $CORZ 与 $CRWV 签署的 12 年期 87 亿美元 数据中心托管(Colo) 合同带来了更可预测的现金流，且约 80% 的资本支出(Capex)（约 1.96 亿美元/第三季度 2.445 亿美元）由 Coreweave 资助，这种模式对股东远比可转换债券或侵蚀利润的有毒债务利息友好。 此外，拥有约 13-15 GW 容量，Coreweave 仅占用 500 MW，且收入能抵消资本支出，这相当有趣。
英文原文
You’re right, I missed $CORZ. Thanks for pointing it out, and appreciate others like @cazenove_uk for that too. Should just treat it as standalone now given $CRWV merger got terminated the other day, don’t know how it slipped my mind. I’ll add that in the next 2/2 post. $8.7 billion colo with $CRWV over 12 years gives $CORZ more predictable cashflows, and majority of capex ~80% ($~196m/ $244.5 million Q3) or so was funded by Coreweave, which is a lot nicer of a model for shareholders than convertible debt or toxic debt interest cutting into margins. Also, with ~1.3-1.5 GW capacity, Coreweave only taking up 500 MW, and revenue offsetting capex, it’s quite interesting.
-
确认CRWV行使期权锁定110亿，将更新文章
@squarecapital3 @DeepValueBagger 是的，你说得对，$CRWV 行使了 150 兆瓦的期权，因此 110 亿美元已锁定，加上 160 亿美元的积压订单收入。我会更新我的下一篇文章，感谢指正。
英文原文
@squarecapital3 @DeepValueBagger Yeah you’re right, $CRWV exercised the 150 mw option so $11b is locked in with $16B backlog revenue. I’ll update my next post, thanks for the correction
-
SLNH依赖融资推进管线，成败取决于能否吸引资金。
谢谢！是的，对于 $SLNH 来说，其逻辑基本是：“我们有一个计划，能否通过足够的看多宣传(bullpost)来吸引更多投资者给我们资金？”如果这招奏效，上行空间巨大。如果人们不给钱，他们的整个计划就会崩溃，因为他们是一家市值很小的公司，没有其他东西来支撑其研发管线(pipeline)。是的，此时做研究很好，很多小市值公司都是绝对的宝石(gems)。
英文原文
Thanks! Yeah for $SLNH it's basically: "We have a plan, can we bullpost enough for more investors to give us money?" If that works, upside is big. If people don't give them money, their whole plan falls apart since they're a small marketcap company with nothing else to back it up their pipeline. Yeah it's good to do research during this time, a lot of small marketcaps were absolute gems.
-
SLNH产能规划难变现，下行风险致市值仅2.8亿
@_stockResearch 再次强调，$SLNH 是产能管道不等于自由现金流(FCF)/利润率(利润率)的完美例证。你可以拥有 2.8 GW 的规划产能，但这未必能实现规模化转化，且存在巨大的下行风险。这正是其市值仅为 2.8 亿美元的原因。
英文原文
@_stockResearch Also once again $SLNH is the PERFECT EXAMPLE of why capacity pipelines does not mean FCF/Margins. You can have 2.8 GW capacity lined up but that might not translate at scale and there's huge, huge downside risk. That's exactly why they're a $280m market cap.
-
看多WYFI，警示SLNH微盘股融资风险及产能变现不确定性。
我个人对 $WYFI 持看涨态度。至于 $SLNH，人们一直声称其拥有 2.8 GW 的订单管道（与 $IREN 合作拥有 3 GW 产能等），但显然存在不利因素。 关于 SLNH：据我了解，他们从事弃电能源货币化（curtailed energy monetization），即建设模块化数据中心或风能/太阳能农场，其产生的能源超过电网可接受的限度。 他们更像是一个开发商/运营商，试图建设这些站点并谈判获取廉价能源。但我们尚未看到这种模式实现规模化或获得全额融资。 这只是潜在产能。而且这通常无法转化为高利润率和高自由现金流（FCF）。 这也是为什么它的市值低于 3 亿美元，且起步时低于 7000 万美元。如果他们能说服足够的散户投资者通过稀释股权提供资金，然后用这些资金为其所有产能管道和 GPU 融资，那么风险/回报比是值得的。 否则，如果该板块发生什么事，或者人们对 $SLNH 转看跌，那么整个计划就会消失，公司就会倒闭。 这只是我个人的看法，人们不能指望一只从 8000 万美元涨到 2.8 亿美元的微盘股没有下行风险，然后只是盲目吹嘘 2.8 GW 的产能。
英文原文
I'm personally bullish on $WYFI. As for $SLNH, People keep claiming 2.8 GW pipeline (3 GW capacity with $IREN, etc) but obviously there's downsides. So for SLNH: from my understanding, they do curtailed energy monetization, where they build modular data-centers wind or solar farms that produce more energy than the grid can accept. They're more of a developer/operator trying to build these sites and negotiate cheap energy access. But we haven't seen this type of model scaled or fully financed yet. It's potential capacity. And that doesn't translate into high margins and FCF lot of time. There's a reason it's a sub <$300m marketcap and started off sub $70M. If they can convince enough retail to give them money for dilution, and then use that to finance all their capacity pipelines and GPUs, then the risk/reward is worth it. Otherwise if something happens with the sector or people grow bearish on $SLNH, then the whole plan just goes disappears and the company goes under. That's just my personal take, people can't really expect some $80m -> $280m microcap stock to not have downside risk and then just shill 2.8 GW capacity.
-
AI工具在分析Neoclouds时存在缺陷，作者更倾向手动整合信息并采用估值建模。
我发现像 @AskPerplexity、Grok 等 AI 工具在研究 Neoclouds（新型云服务商）和每日新闻频发的板块时，结果往往极不准确。它们会遗漏大量细节，例如 $NBIS 的白皮书中有关于与 $CRWV 等对比的利用率信息，这些数据对 AI 不可见（暗示了 GPU 利用率带来的毛利率增长）。当 Meta 与 $CRWV 达成 140 亿美元交易，且上周 $AMZN、$MSFT 等发布财报显示 AI 资本支出增加时，AI 可能会忽略那些未直接提及 $NBIS 但暗示超大规模云厂商交易增加的顺风因素。或者几天前的额外降息提振了涉及债务的 Neoclouds 的前瞻性盈利。亦或是关于 Clickhouse 等子公司或投资组合公司的新信息，AI 可能无法将其串联起来。FinX 散户在将这些信息拼凑成前瞻性增长方面做得更好，鉴于 AI/Neocloud 的总可寻址市场(TAM)具有高度投机性和快速增长，很难建模。这也是我采用估值方法的原因（在我之前关于 $NBIS 如何达到 1000 亿美元市值的帖子中），但如果看 $UPWK 等更标准的业务，我的估值建模方式则不同。
英文原文
I've actually found AI tools like @AskPerplexity, Grok and others to be extremely off when looking at Neoclouds and sectors with new news every day. They miss a lot of details, eg. with $NBIS you have utilization information from Whitepapers comparing Nebius to $CRWV and others that aren't publicly viewable to AI (which insinuated growing gross margins from GPU utilization). When Meta struck a $14B deal with $CRWV and you have earnings report last week from $AMZN, $MSFT, and others you have increased capex spend into AI, and they might miss tailwinds that don't mention $NBIS directly with increased changes of more hyperscaler deals. Or when there's an additional rate cut a few days ago that boosts forward earnings especially with Neoclouds that involve debt. Or when there's new information about subsidiaries or portfolio companies like Clickhouse that AI might put 1+1 together with. FinX retail does a lot better job with piecing all this information together into forward growth and AI/Neocloud TAM is really hard to model given how speculative and rapidly growing it is. That's kinda why I go with valuation approaches (in my previous post on how you get to $100B MC for $NBIS), but if you look at $UPWK and more standard businesses. I do valuation modelling differently