全部推文
共 6436 条 · 第 169 / 215 页 · 滚到底自动加载更多
-
重申NBIS长期看涨至250美元，对期权亏损致歉。
是的，我一个月前说 $NBIS 即使在 130 美元时也是极强的买入标的，这一观点我至今仍坚持，如果它交易在 125 美元我会买入。我认为跌至 75 美元就像 $HOOD 在关税恐慌中跌至 28 美元然后随着散户投降涨至 100 美元以上那样。如果人们跟随我的 Nebius 论点使用本月到期的期权，我感到非常内疚。我自己也因下跌在 3 月/5 月到期的期权上损失惨重。但中长期我相信一旦他们在 2026 年实现管理层预测，$NBIS 轻松突破 250 美元。
英文原文
Yep, you’re correct when I said that $NBIS is an extremely strong buy even at $130 a month ago. I still maintain that today, and would be buying Nebius if it were trading at $125. I believe the drop to $75 is like the moment when $HOOD dropped to $28 during tariff fears then went to $100+ when retail capitulated. I do feel extremely guilty if people followed my thesis on Nebius with options that expired this month. I lost a lot of money myself so far on March/May expiry options from the drop. But medium-long term I believe $NBIS easily blows past $250 once they achieve management projections in 2026
-
作者认为AAOI和POET虽具十倍潜力但因执行风险未去风险化，不属于高确信度标的。
对我个人而言：当一家公司在根本上已“去风险化(de-risked)”并正从该点迈向规模化时，我才视其为“高确信度(high conviction)”标的。 例如，$NBIS 在与微软达成交易后。在光子学(photonics)领域，我个人更看好 $AAOI 而非 $POET。但它们都尚未“去风险化”。 正如你提到的，Poet 受益于 Celestial（已被 $MRVL 收购）。Marvell 基本上通过 Celestial 向超大规模云厂商(hyperscalers)出售 Poet 的中介层(interposer)，用于连接 TPU/Trainium。 但这中间隔着多层环节。 $AAOI 已经拥有与超大规模云厂商的直接合同，例如与 $AMZN 合作 Trainium，与 $MSFT 合作 Maia。此外还有类似 $ALAB 与亚马逊那样的股权协议。 然而，它们实现了完全垂直整合且工厂建在德克萨斯州。如果能让整个供应链在美国实现规模化并完美执行，它们有巨大的10倍增长潜力。 但由于相比 $COHR 规模较小，$AAOI 存在巨大的执行风险，因此不属于高确信度标的。$POET 同理，目前尚无让我完全确信的去风险化因素。 不过，Poet 和 AAOI 都是具备10倍潜力的光子学标的，这也是我持仓部分 $AAOI 的原因，但它们都不属于高确信度投资。
英文原文
For me personally: A stock is "high conviction" when they are fundamentally de-risked and are going from that point to scale. For example, $NBIS post-MSFT deal. I like $AAOI much better than $POET in the photonics space personally. But neither of them are "de-risked yet". So Poet as you mentioned, benefits from Celestial, which got acquired $MRVL. Marvell basically sells Poet's interposer via Celestial to hyperscalers for connecting tpus/trainium. But this is via multiple hops away. $AAOI already had direct hyperscaler contracts like $AMZN for Trainium + $MSFT for Maia. And equity agreements like what $ALAB has for Amazon. However, they're fully vertically integrated + built in Texas. They have a pretty big potential to 10X if they can get the whole supply chain to scale in America + perfect execution. But because they're at lower scale compared to $COHR, $AAOI has huge execution risk + it's not high-conviction. Same with $POET, there's no de-risking factor yet that would make me fully convinced. However, Poet and AAOI are both photonics plays that can 10x, which is why I have some $AAOI in my portfolio, but neither are high conviction.
-
通过对比 IREN 缺乏核心 AI 资产，论证 NBIS 的竞争优势。
我会用一个关于 $NBIS 与 $IREN 的问题来回答你的问题。 IREN 是否拥有： - 一家由 Uber 规模化运营的 FSD（完全自动驾驶）L4 级 Robotaxi（自动驾驶出租车）公司 - 一家被从 Tesla 到 Meta 等所有顶级公司使用的 DB（数据库）公司 - 一家被 Amazon、Anthropic 等使用的 AI 标注公司？ 这些业务是否都随着 Nebius 的主营业务实现了 100%+ 的同比增长？
英文原文
I will answer your question with a question regarding $NBIS vs $IREN. Does IREN have a: - FSD lvl4 robotaxi company scaled by Uber - a DB company used by every single top company from Tesla to Meta - an AI labeling company used by Amazon, Anthropic and others? Each growing 100%+ Y/Y alongside Nebius's main business?
-
感谢 LITE 提醒，并把它和 CRDO/ALAB 的连接层逻辑并列
感谢提醒 $LITE。 我之前没有意识到它对 $NVDA Blackwell、$GOOGL TPU 和 $AMZN Trainium 芯片有多关键。 离谱的是，它竟然处在每家 hyperscaler 的单 GPU/TPU/ASIC 核心环节里（这也是为什么它是高信念标的）。 至于 $CRDO，它和 $ALAB 类似，都是高 Nvidia 式利润率、三位数同比增长，以及依赖 hyperscaler 连接需求的逻辑。
英文原文
Thanks for the $LITE shoutout. I didn't realize how instrumental it was to $NVDA blackwell, $GOOGL TPUs, and $AMZN Trainium chips. Insane how it's part the core of single GPU/TPU/ASIC from every hyperscaler (hence why it's high-conviction) As for $CRDO, similar ground to $ALAB regarding high-Nvidia like margins, triple digit Y/Y growth, and hyperscaler dependencies for connectivity.
-
说明稳定币与 BTC 下行周期相对隔离
好问题。稳定币其实和 BTC 下跌趋势是分开的，尽管算法层面会影响该行业公司/资产的市值。 即使加密价格下跌，USDC/Tether 的供应量也会增加。机构会在加密下行周期中把 BTC 卖成 USDT/USDC，而不是放在银行账户里。 尤其现在在风投领域，Lightspeed、Sequoia 等都在把钱投向和稳定币相关的公司，比如 neobank；其中很多投资会在 6 个月到 1 年后开始产生回报。 我会在另一篇帖子里写更深入的 thesis，但总体上，我看好稳定币和 Bitcoin；即使 Bitcoin 处于下行趋势，稳定币和发行方也是相对隔离的。
英文原文
Good question, stablecoins are actually separate from the BTC downtrend, even though algorithmically this affects the marketcaps of companies/assets in the sector. USDC/Tether increases in supply even as crypto prices go down. Institutions sell BTC for USDT/USDC while crypto is in a downtrend instead of holding it in bank accounts. Especially in venture capital right now, lightspeed/sequioa/etc. are pouring money into stablecoin related companies like Neobanks, and lot of these investments pay dividends 6 months.-1 year later. I'll write up a more in-depth thesis in another post but generally, I'm bullish on stablecoins and Bitcoin, and even if Bitcoin is in a downtrend, stablecoins + issuers are isolated.
-
列出 Circle 的几个主要空头情景
几个空头情景： - 美联储推出自己的 CBDC，比如 Fed Coin，或者推出另一个稳定币来直接竞争。 - 美联储继续降息到 1%，而 $CRCL 必须不断印 USDC 来弥补收入损失。 - 监管，比如下一届政府任命 Gary Gensler 这类人物，或者 USDC 在其他国家被禁。 但如果事情继续按现在的方向发展，Circle 未来应该是一家 500 亿美元以上的公司。
英文原文
Few bear cases: - Fed creates it's own CBDC like Fed Coin or another stablecoin to compete directly. - Fed keeps cutting rates down to 1% and $CRCL needs to keep printing USDC to make up for lost revenue. - Regulation - eg. next administration puts in a gary gensler or usdc gets banned in other countries But if things go the way they're going now, Circle deserves to be a $50B+ company down the road.
-
解释此前回避 Circle 是因为流通盘动态，现在重新看多
可以。所以我之前一直说，在 Circle 市值 600 亿美元时，要避开 Circle，或者做空 $CRCL、做多 $COIN。 原因是大家没理解流通盘动态。 大部分流通股被锁定，不能卖，类似于 BULL IPO。等流通股释放后，Circle 股价就崩了。 降息也会伤害稳定币铸币利息带来的盈利能力，但流通盘动态才是压倒性的最大因素。 我一直看好 USDC，但市值非常关键。现在它回到 180 亿美元，就变成了极强的买入机会。
英文原文
Sure, so I've been saying avoid Circle or short $CRCL long $COIN when Circle was trading at a $60B marketcap. This was because people didn't understand float dynamics. Majority of the float was locked up and couldn't be sold, similar to BULL IPO. When float was released, the price on Circle crashed. There's stuff like rate cuts too hurting profitability from interest on minted stablecoins, but float dynamics was the biggest factor by a longshot. I've always been bullish on USDC, but market cap plays a huge role. Now it's an extraordinarily strong buy back at $18B
-
说明 USDC 发行和公链网络价值捕获的关系
@dubiousnoob 所以 $CRCL 和 $COIN 是 USDC 的发行方，你说它是在 Ethereum 上发行是对的。 但它也在 Solana、Base、Polygon 等网络上发行。具体网络没那么重要。 Ethereum 网络是基础设施，但 USDC 并不会给 Ethereum 代币带来太多价值捕获。
英文原文
@dubiousnoob So $CRCL and $COIN are the issuers of USDC, you’re correct in saying it’s issued on Ethereum. But it’s also done on Solana, Base, Polygon, and so on. Network doesn’t matter much. Ethereum’s network is the infrastructure but USDC doesnt accrue much value to Ethereum’s token.
-
认为 USDC 已成为结算标准，价值远超当前估值
$CRCL 是稳定币，和 Ethereum 或 Solana 这类 L1 没什么关系，因为它们并不重要。 thesis 在于，它正在成为一种结算标准（Visa、PayPal、YouTube 等），而这种网络效应的价值远远超过它 170 亿美元的估值。 这既符合美国政府把更多人纳入美元/美国债务体系的战略利益，也符合全球公司的利益。 人们当然可以铸造 PayPal USD、Paxos USD 等，但 USDC 基本已经成了大家认可的标准，这一点极其、极其有价值。
英文原文
$CRCL is stablecoins, has nothing to do with L1s like Ethereum or Solana since it doesn’t matter. Thesis involves that it’s becoming just a standard for settlement (visa, PayPal, YouTube, etc.) and that network effect is worth infinitely more than its $17B valuation both in us government strategic interest to get people into the US dollar/debt system as well as the global companies. People can mint PayPal usd, paxos usd, etc but USDC has basically become the agreed upon standard and this is incredibly, incredibly valuable.
-
比较一组高增长标的的高估和低估程度
所以我会说，$RKLB 现在肯定是最被高估的。 $LITE 正接近超买区间，但考虑到它在 TPUv7、Trainium 和 Blackwell 芯片中的关键性，它确实值得重估。 $NBIS 和 $CRCL 是这组里最被低估的，然后是 $ALAB。
英文原文
So I’d say $RKLB is definitely the most overvalued right now. $LITE is edging toward the overbought territory but it definitely warrants a re-rating given how critical it is in TPUv7, Trainium, and Blackwell chips. $NBIS and $CRCL are most undervalued out of the bunch, then goes $ALAB.
-
列出更稳的十倍股和小盘 moonshot
如果说“更安全”的 10 倍股，我会说未来 5 年是 $RKLB 和 $NBIS。 我完全能想象 Rocket Lab 有一天到 3000 亿美元估值，Nebius 到 2000 亿美元估值。 市值越小，10 倍概率越高（比如低于 20 亿美元）。显然，对 $HOOD 这类公司来说，一年翻倍要难得多。 但在小盘股里，$AAOI、$WLAC 和 $TE 是我喜欢小仓位参与的那类 moonshot。
英文原文
So for a “safer” 10x. I’d say $RKLB and $NBIS next 5 years. I could definitely see Rocketlab at $300B valuation and Nebius at $200B one day. The chance of a 10x increases the smaller marketcap they are (eg. Sub $2B), and it’s obviously much harder for companies like $HOOD to double in a year. But for smaller caps, $AAOI, $WLAC, and $TE are those types of moonshots I like to dabble in.
-
公布 2026 年五只高信念多倍股
我在个人组合的极端增长高信念篮子里新增了 2 只股票。 在更长时间框架上，我对高信念股票从来没错过（比如 $HOOD 从 18 美元到 100+ 美元）。 我 2026 年的 5 只高信念多倍股现在是： $NBIS - AI 算力和 Robotaxi $RKLB - 太空 $ALAB - 连接 新增的两只是： $CRCL - 印钞机 $LITE - hyperscaler ASIC 我后面会写 Circle 和 Lite 的更长 DD thesis。 我有信心，只持有这 5 只，就会大幅跑赢 99.9% 的组合。
英文原文
I’ve added 2 new stocks to my extreme growth high conviction basket in my personal portfolio. I’ve never been wrong before about high conviction stocks on longer timeframes (eg. $HOOD $18 -> $100+) My 5 high conviction multi-bagger stocks for 2026 are now: $NBIS - AI Compute and Robotaxis $RKLB - Space $ALAB - Connectivity And the two new ones are: $CRCL - Money Printer $LITE - Hyperscaler ASICs I’ll do longer DD thesis on Circle and Lite on follow ups. I’m confident having just these 5 will heavily outperform 99.9% of portfolios.
-
用电竞和股票交易开玩笑
@goatyishere @PolymarketMoney 对我来说，成为电竞作弊器只是我交易 $TE 这类股票时的支线任务。 我在小组赛阶段就知道 T1 会赢。 有些东西就是写在星星里的，15 美元变 320 万美元很简单！
英文原文
@goatyishere @PolymarketMoney Being a cheat code for esports is just a side quest for me when I trade stocks like $TE. I knew t1 would win during group stages. There’s just some things written in the stars, $15->$3.2m is easy! https://t.co/DLAZnalCRM
-
祝对方欧洲旅行愉快，提到自己在亚洲旅行
@DeepValueBagger 祝你在欧洲玩得开心！ 我最近一直在东南亚附近旅行，比如关岛和日本。
英文原文
@DeepValueBagger Have fun in Europe! I’ve been spending my time traveling around SEA like Guam and Japan recently https://t.co/NS2cOQvDtl
-
调侃 FT 采访和 DUOL 回复
@RJCcapital @FT @retail_mourinho @zephyr_z9 @jiahanjimliu @DeepValueBagger @NighthawkTradez 当 FT 做我的个人采访时，我唯一的回复就是 $DUOL。 那时候他们就知道必须把 RJC 也加进名单里了。
英文原文
@RJCcapital @FT @retail_mourinho @zephyr_z9 @jiahanjimliu @DeepValueBagger @NighthawkTradez When FT did my individual interview my only reply was $DUOL. That’s when they knew they had to include RJC on the list too.
-
称 NBIS 即使更高也想买，但现金不足
@soulbiri1 对。如果 $NBIS 今天是 135 美元，我还是会买；但很遗憾，在下跌尾端我已经没有足够干火药继续部署了。 最后我买了光子玩家和 Circle 的长期仓位。
英文原文
@soulbiri1 Yep. If $NBIS was $135 today I’d still buying - but ran out of dry capital to deploy on the full tail end of the drop sadly. Ended up buying long term positions in photonics players and Circle though
-
认为 Nebius 基本面没变，情绪转向且收购叙事增强
像 $NBIS 这样的公司，基本面没有任何变化（除了继续改善）。变化的只是市场情绪突然切换了。 但现在，Wedbush 的 Dan Ives 这样的人突然出来说：“Nebius 可能在 2026 年被 $GOOGL、$MSFT、$AMZN 这样的 hyperscaler 收购”，这极其利多。
英文原文
Nothings fundamentally changed (aside from improving) about companies like $NBIS. Aside from the sentiment flipping the switch. But suddenly we have people like Dan Ives from Wedbush going out and saying "Nebius likely gets acquired by a hyperscaler like $GOOGL, $MSFT, $AMZN " in 2026, which is incredibly bullish.
-
OpenAI 高估值融资改善 AI 数据中心与供应链风险偏好
最新消息：OpenAI 在完成 5000 亿美元估值、100 亿美元以上的 Amazon 轮融资后，正在以 7500 亿美元估值继续融资。 $CRWV：+15.85% $NBIS：+10.28% 有了这笔交易，像 CoreWeave 和 Oracle 这样的公司，在为 OpenAI 的资本开支需求建设容量时，对手方风险在结构上降低了，因为 OpenAI 的资产负债表更强了。 像 Nebius 这种算法上和行业龙头绑定的公司，也因此上涨。 从 Rocket Lab 到 Bitcoin，高 beta 资产普遍上涨。 最近因为日元套息交易解除，加上今天大量未平仓合约到期，市场出现了极端波动。 不过，AI 交易的基本面（尤其是 Micron 神级财报显示了极强内存需求）以及 Neocloud 的基本面，仍然比以往更好。
英文原文
Latest news: OpenAI is raising funds at a $750B valuation after their $10B+ Amazon round at $500B. $CRWV: +15.85% $NBIS: +10.28% With this deal, companies like Coreweave and Oracle structurally have less counterparty risk with OpenAI's stronger balance sheet to fund capex requirements. Companies like Nebius that are algorithmically tied to the sector leaders are up as a result. High-Beta Assets from Rocketlab to Bitcoin are up across the board. There was extreme volatility recently with the Yen carry trade unwinding + large open interest expiring today. However, the fundamentals of the AI trade (especially with Micron's godlike ER showing extreme memory demand), and Neoclouds remain better than ever.
-
认为 TAE 若商业化聚变，DJT 估值上限很高
@Toy0tomi_H TAE 是硅谷聚变能源的皇冠明珠，所以它要和 $DJT 合并的消息真的让我措手不及。 如果 TAE 到 2030 年成功商业化聚变，1000 亿美元以上估值都算极其保守，哈哈。
英文原文
@Toy0tomi_H TAE is silicon valley's crown jewel for fusion energy, so the news about merging with $DJT really caught me off guard. If TAE successfully commercializes fusion by 2030, $100B+ is extremely conservative lol.
-
梳理 DJT 从社交媒体到聚变收购的荒诞估值难题
$DJT 刚刚变成股票市场的皇冠明珠了吗？ 我真的难以置信： 1. 被 X/Meta 封禁 2. IPO 一个社交媒体创业公司 3. 卖圣经、西装和运动鞋 4. 成为美国总统 5. 卖 10 万美元陀飞轮手表 6. 发行 memecoin 7. 发行 crypto ETF 8. 购入数十亿美元 Bitcoin 并建立战略储备 9. 发行 USD1 稳定币 10. 收购 Google 宠爱的核聚变公司，而且它恰好是世界领导者 （Google 真的派工程师去帮 TAE 反应堆稳定等离子体，高盛也是投资者） -> 股价上涨 46%。 $DJT 现在因为 TAE 真的变得可投资了吗（可能还结合总统审批来加速监管通关）？ 市场是不是正难以置信地看着下一个 Microsoft 形成？ 给 Trump 的公司估值，真是一个绝对头疼的问题。
英文原文
Did $DJT just become the crown jewel of the stock market? I’m genuinely in disbelief: 1. Get Banned from X/Meta 2. IPO social media startup 3. Sell Bibles, Suits, and Sneakers 4. Become US President 5. Sell $100K Tourbillion Watches 6. Launch memecoin 7. Launch crypto ETFs 8. Acquire billions in Bitcoin and build a strategic reserve 9. Launch USD1 Stablecoin 10. Acquire Google-Darling Nuclear Fusion company that’s happens to be the world leader. (Google literally sent their engineers to stabilize plasma in TAE’s reactors + Goldman Sachs are investors) -> Send stock up 46%. Did $DJT actually investable now with TAE (likely combing president approval for speed running regulation)? And are markets watching in disbelief of the next Microsoft in making? How you value Trump’s company is an absolute headache.
-
询问 Bitfinex 企业合约与费率锁定
@bitfinex 嘿，如果这是真的，有没有 Bitfinex 员工能联系我，讨论设置长期企业合约和费率锁定？ @paoloardoino
英文原文
@bitfinex Hey if this true, is there any bitfinex employee that can reach out regarding setting up a long term enterprise contract and a fee rate lock in? @paoloardoino
-
讽刺拿到 hyperscaler 合作反而市值蒸发
@ChiefLiquidity 没错，拿到像 $IREN 那样的 hyperscaler 合作 = 市值被抹掉 40%。
英文原文
@ChiefLiquidity True, get a hyperscaler deal like $IREN = -40% wiped off your market cap.
-
Hut8 与 Anthropic/Fluidstack 签署大规模租赁后上涨
$HUT 刚刚通过 Fluidstack 与 Anthropic 签署了一项新的三重租赁协议，容量至少 245MW，最高可达 2295MW。 该交易 15 年价值 70 亿美元，并由 $GOOGL 为交易提供兜底。 Hut8 在消息后上涨 25.28%，现在市值为 39.8 亿美元。
英文原文
$HUT has just signed a new Triple lease deal with Anthropic through Fluidstack for at least 245 MW and up to 2,295MW capacity. The deal is valued at $7 Billion over 15 years with $GOOGL backstopping the deal. Hut8 is up 25.28% on the news and now trades at a $3.98B market cap https://t.co/50O28EsJVp
-
Amazon 投资 OpenAI 对 AI 数据中心和 ASIC 供应链的连锁影响
最新：Amazon 的 100 亿美元 OpenAI 融资，以及 AI 供应链的涟漪效应。 $AMZN 将以 5000 亿美元以上估值向 OpenAI 投资 100 亿美元以上。 为什么这是 AI 股票的重大结构性变化： 1. AI 数据中心交易去风险：（$ORCL、$CRWV、$APLD、$CORZ） 结合前面提到的 SPEED 法案，影响 Neocloud 的主要问题是： 1. 数据中心延迟和递延收入 2. 不可持续的 CapEx -> 没有 FCF 3. OpenAI 传染/积压订单风险 SPEED 法案直接处理了第 1 和第 2 点，但没有处理 OpenAI 相关的第 3 点。 影响 Oracle、CoreWeave 等最大 Neocloud/数据中心提供商的主要担忧，是它们为 OpenAI 这个对手方投入了巨额 capex，而 OpenAI 本身没有足够资金承诺这些资本开支。 但现在，随着 Amazon 新融资，第 3 点开始被解决。 - 新增 100 亿美元资金，加上 Amazon 的资产负债表支持 ChatGPT 的创造者，OpenAI 对 Oracle 和 CoreWeave 的早期承诺开始获得兜底。 - 下游影响：这直接降低了 $CRWV 和 $ORCL 的风险，因为它们在为 OpenAI 建设容量。再往下两层，依赖 CoreWeave 作为租户的 $APLD 或 $CORZ 也会被去风险。 由于 CoreWeave 和 Oracle 被视为“行业龙头”，这会立刻改变整个 Neocloud 板块的情绪，包括 $NBIS、$IREN、$CIFR、$WULF 等。 2. Hyperscaler AI capex 浪潮（$AMZN、$MRVL、Alchip）： 我们之前看到，$AVGO ASIC backlog 之后，市场开始担心 hyperscaler 支出减弱，很多相关玩家因此大跌。 但这笔交易的关键条件之一，是 OpenAI 采用 Amazon 自研 Trainium ASIC。这代表非 Nvidia 集群会激进扩张。 - 设计与 IP：直接利好 Marvell（$MRVL）和 Alchip 等 ASIC 设计伙伴。 - 定制硅集群需要大量光互连和 HBM。这会为光子（$AAOI、$LITE、$COHR）和内存（$MU、SK 海力士）创造新的 capex 超级周期。 - 代工厂，比如 $TSM。 以及更多参与 hyperscaler ASIC 建设的相关公司。 唯一输家？Nvidia（$NVDA）。 Amazon 正在成功用它庞大的资产负债表，迫使领先 LLM 摆脱对 H100/Blackwell 的依赖，同时推动整条 AI 供应链。 核心结论是，AI 交易由世界上最富有的公司资助，比如 $AMZN；OpenAI 也正在证明，它可以通过出让股权来扩大资产负债表，以满足需求。 做多 AI 板块。
英文原文
Just In: Amazon's $10B OpenAI Funding and The AI Supply Chain Ripple Effect. $AMZN is set to invest $10B+ in OpenAI at a $500B+ valuation Why this is a MASSIVE structural shift for AI stocks: 1. De-Risking the AI DC trade: ( $ORCL, $CRWV, $APLD, $CORZ ) With the SPEED Bill mentioned earlier, the main issues affecting Neoclouds were: 1. DC Delays & Deferred Revenue 2. Unsustainable CapEx → No FCF 3. OpenAI Contagion/Backlog. the Speed bill directly addresses #1 and #2. But not #3 with OpenAI. The main fears affecting the biggest Neocloud/Datanceter providers like Oracle, Coreweave was their immense capex spend for a counterparty (OpenAI) that doesn't have the funding to commit to it's capex spend. But now, #3 is starting to be addressed with the new Amazon funding. - With a fresh $10 Billion and Amazon’s balance sheet backing the creator of ChatGPT, OpenAI's early commitments to Oracle and CoreWeave are now starting to be backstopped. - Downstream Impact: This directly derisks companies like $CRWV and $ORCL, who are building capacity for OpenAI. And by two hops, companies like $APLD or $CORZ that rely on Coreweave as a tenant. As Coreweave and Oracle are seen as the "sector leaders" this immediately changes sentiment across the whole Neocloud sector from $NBIS, $IREN, $CIFR, $WULF and others as well. 2. The HyperScaler AI capex wave ( $AMZN, $MRVL, Alchip): We've seen fears after $AVGO ASIC backlog about hyperscaler spending waning. And many related players tanked on the news. However, a key condition of this deal is OpenAI’s adoption of Amazon’s proprietary Trainium ASICs. This signals an aggressive scaling of non-Nvidia clusters. - Design & IP: Direct benefit to ASIC design partners like Marvell ( $MRVL ) and Alchip. - Custom silicon clusters require massive optical interconnects and HBM. This creates a new capex supercycle for photonics ( $AAOI, $LITE, $COHR ) and memory ( $MU, SK Hynix). - Foundries such as $TSM. and many more related companies involved in the buildout of hyperscaler ASICs. The only loser? Nvidia ( $NVDA ). Amazon is successfully using its massive balance sheet to force the leading LLM to diversify away from H100/Blackwell dependence and boosting the whole AI supply chain alongside it. The main takeaway is that the AI trade is funded by the richest companies in the world, such as $AMZN, and OpenAI is showing it can scale up its balance sheet to meet requirements by trading off equity. Go long on the AI sector.
-
吐槽 NYT 仍未更正对 Alexandr Wang 的描述
@alexandr_wang 我只是觉得好笑，@nytimes 到现在还没发更正。 显然，MIT 前竞赛程序员、运营 Meta Superintelligence Labs，这还不足以让他们把你称为工程师，只能叫“商业人士”和“连接者”。
英文原文
@alexandr_wang I'm just laughing that @nytimes still hasn't issued a correction. Apparently being a former competitive coder from MIT who runs Meta Superintelligence Labs isn't enough for them to qualify you as an engineer, but just a "business guy" and "connector". https://t.co/qu3Kxy3Gev
-
估算 AI 基建法案通过概率，并更看重行政命令
是的，就像我和 Sydney Sweeney 的机会是 50/50 一样，法案通过、然后 $NBIS、$IREN、$CIFR 起飞的概率也是 50/50。 要么发生，要么不发生。 玩笑归玩笑，我随便猜的话，考虑到 Bernie 等民主党人早期反对，概率大概是 30-35%。 我会更押注可能的行政命令强行推进，因为 $ORCL 创始人和前沿 LLM 公司与本届政府关系很近。 另外，AI 基础设施现在已经是国家安全风险。
英文原文
Yes like how my chances with Sydney Sweeney are 50/50, the chances that the bill passes and $NBIS, $IREN, $CIFR goes to the moon are 50/50 as well. Either it happens or it doesn’t. Jokes aside just a wild guess is 30-35% given early disagreement from Democrats like Bernie. I’d count more on likely executive orders ramming things through given how close the $ORCL founder and frontier LLMs are with the administration. Also given how AI infrastructure is now a national security risk.
-
日文说明 AI 数据中心仍有风险，但法案能缓解能源和许可瓶颈
正如你所说，AI 数据中心行业仍然存在信用收紧和物理约束等担忧。这个法案并不能解决所有问题。 但是，它仍然是这个板块最大的顺风之一，因为它能缓解能源供应和许可流程方面的瓶颈。这样一来，从建设到变现的时间风险会显著降低，$CRWV 等公司此前作为下调指引原因提到的“递延收入”确认也会加快。 我个人认为，这个法案比三次降息还要利多。因为相比利率下降，更重要的是能把庞大的资本开支（CapEx）更快转化为自由现金流（FCF）。
英文原文
おっしゃる通り、AIデータセンター業界には信用の引き締め（クレジットタイトニング）や物理的な制約といった懸念材料が依然として存在します。この法案ですべての問題が解決するわけではありません。 しかし、エネルギー供給と許認可プロセスのボトルネックを解消するという点で、このセクターにとって最大級の追い風であることに変わりはありません。これにより、収益化（マネタイズ）までの期間に伴うリスクが大幅に低減され $CRWV などがガイダンス下方修正の要因として挙げていた「繰延収益」の計上も加速することになります。 個人的には、この法案は3回分の利下げよりもポジティブだと考えています。なぜなら、金利低下以上に、膨大な設備投資（CapEx）をより迅速にフリーキャッシュフロー（FCF）へと転換できることの方が重要だからです。
-
认为近期抛售来自套息交易解除，中期降息利好风险资产
我的观点和之前一样。最近这轮抛售，是日本央行加息和美联储降息后重新建立仓位，再引发套息交易解除。 短期来看，套息交易解除的影响比三次降息大得多。 现在我们已经看到了这轮抛售的结果。 但 12 月 10 日的美联储降息，对 $RKLB 到 $NBIS 这类风险资产的中期，比如未来 3 个月，是极其利多的，尤其是在它们基本面还在增长的情况下。
英文原文
I’ve held the same view as before. Recent selloff is carry trade unwind from reload after boj hike and fed rate cut. Carry trade unwind is much more impactful than 3x rate cut near term. And now we’ve seen the result of that selloff. But Dec 10ths fed rate cut is incredibly bullish for risk assets from $RKlB to $NBIS medium term like 3 months out, especially when they have growing fundamentals.
-
认为 SPEED Act 缩短利用率滞后，比降息更直接改善 FCF
不同意。和 75 个基点降息相比，利用率滞后造成的是利润率破坏。 降息当然是巨大顺风，因为 Neocloud 需要用债务来资助扩张。 但如果美国政府通过 SPEED Act，把 $ORCL 或 $CRWV 的资本开支转化为收入的速度提前几周，甚至提前一个季度，尤其是在资产快速折旧的情况下，这对 FCF 的影响根本不是一个量级。 如果法案通过，这会让资本开支投资 thesis 更加可行。
英文原文
Disagree, utilization lag is margin destruction compared to .75 bps rate cut. Rate cuts are a massive tailwind given how Neoclouds debt to fund expansion. But, when the US gov guarantees faster revenue conversion on $ORCL or $CRWV on capex spend with the Speed act by a few weeks or even a quarter, especially when assets are depreciating rapidly, it’s not even a comparison for the difference in fcf. This made the capex investment thesis even more viable if the bill passes.
-
估计 SPEED Act 时间线并强调其对 Neocloud 的顺风
按时间线估计，大概是 12 月最后一周。 不过像 Bernie Sanders 这样的人正在反对它。我看到 @pepemoonboy 下面的评论笑了，而且他说得很对：Bernie 对 AI 和基础设施作为关键国家安全风险的理解很不足。 当中国正处在构建关键前沿模型的边缘时，根本没必要搞这么多官僚程序和州与州之间的政治博弈。 但如果它通过，这可能是 Neocloud + 数据中心领域最大的顺风，对整个板块都极其利多。
英文原文
Last week of December for a timeline estimate. Regards like Bernie Sanders are pushing back on it though. I laughed at @pepemoonboy comment underneath and it’s so true given Bernie’s lack of understanding of AI and infrastructure a pivotal national security risk. There’s no need for all the bureaucracy and state-by-state political games when China is on the precipice of building critical frontier models. But if it passes, it’s probably the biggest tailwind for the Neocloud + Datacenters space and incredibly bullish for the whole sector.