· 个股论点

看好全栈新云$NBIS利润率优势,减持矿工股。

涉及标的:

中文翻译

市场开始定价: 像 $NBIS 这样的 AWS 式全栈新云(Neocloud)与 $CIFR、$BITF、$CLSK 等矿工之间的分化。 这就是我卖出 $IREN、$WULF 等并集中持仓 Nebius 的原因。 为什么? 💹 利润率 > ⚡️ 产能。 就在今天: Nebius 的 CTO 表示: “市场上几乎没有真正的‘新云’。所谓的新云不过是一堆裸金属服务器,上面堆砌了大量第三方软件。” 耸人听闻?完全不是——如果你认真对待“云”这个概念,而不仅仅是堆砌硬件盒子。 Danila 解释了裸金属运营商很快就会发现 Nebius 已经在大规模层面解决的那些挑战。 这正是我指出 $NBIS 拥有最高非对称上行潜力的原因,因为矿工在规模化时可能会在利润率上挣扎。 我们还看到 Nebius 联合创始人 Roman Chernin 在《福布斯》的文章中总结,AI 经济需要一种新的、全栈的、AI 原生基础设施,而不是拼凑起来的系统: “你可以从一个供应商购买机架,将它们连接起来……但从长远来看,经济性、灵活性和速度取决于你是否控制自己的全栈。” “如果你控制全栈,你就控制产品、性能和经济效益。” _ 这并不意味着像 $WYFI、$APLD、$WULF 等其他高性能计算(HPC)公司没有增长空间,新云板块才刚刚开始上涨(尤其是当我们看到 $META 或 $AMZN 的财报时)。 然而,在投机阶段,我们已经为矿工的产能建设定价——大多数已经上涨了 50–200%+。 但谈到执行层面,那些控制全栈的人将在利润率上获胜。 我们看到 $CRWV 亲自体会到了这一点,花费数十亿美元进行软件收购,利用率仍低于 $NBIS(根据白皮书)。 我们也看到一家市值 8000 亿美元的超大规模云厂商(Oracle)未能建立其全栈,毛利率仅为 14%。 如果 Oracle 都无法完全整合其软件,CoreWeave 仍在花费数年时间尝试这样做,小型矿工如何转型? “仅 Type-1”运营商的论点将艰难地发现规模化时的利润率压缩是什么样子的。 正如《福布斯》文章总结: “归根结底,如果你只是连接盒子或在无法控制的基础设施上构建服务,你在规模和效率的游戏里将受到限制。” $NBIS 今天就已经拥有全栈。

英文原文

Markets are starting to price in: A divergence between AWS-style full-stack Neoclouds like $NBIS and miners such as $CIFR, $BITF, $CLSK, and others. This is the reason I sold off $IREN $WULF and others, and consolidated into Nebius. Why? 💹 Margins > ⚡️Capacity. Just today: The CTO of Nebius stated: “Almost no Neocloud in the market is actually a cloud. As in a bunch of baremetal with a ton of third-party software slapped on top.” Provocative? Absolutely not — if you take the term “cloud” seriously, not just as slapping on a bunch of hardware boxes. Danila explained how baremetal operators will soon discover the challenges Nebius has already solved at scale. And that’s exactly why I stated $NBIS has the highest asymmetrical upside, since miners may struggle with margins at scale. We’ve also seen a Forbes article by Nebius Co-Founder Roman Chernin, summarize how the AI economy requires a new, full-stack, AI-native infrastructure, not the stitched-together systems: “You can buy racks from one supplier, cable them together... but in the long term, economics, flexibility and speed depend on you controlling your own stack.” “If you control the stack, you control the product, the performance, and the economics.” _ This doesn’t mean there’s no room for other HPC companies like $WYFI, $APLD, $WULF, and others to grow, and the Neocloud segment is just beginning its run-up (especially when we look at $META or $AMZN earnings). However, during the speculation phase, we’ve already priced in capacity buildout with miners - most are already up 50–200%+. But when it comes to execution, those who control the full stack will win on margins. We’ve seen $CRWV figure that out firsthand, spending billions on software acquisitions and still maintaining lower utilization than $NBIS (as per the whitepaper). We’ve also seen how an $800B hyperscaler (Oracle) failed to build out its stack, running at 14% gross margins. If Oracle couldn’t fully integrate its software, and CoreWeave is still spending years trying to do so, how will small miners pivot? Arguments from “Type-1 only” operators will find out the hard way what margin compression at scale looks like. As the Forbes article concludes: “At the end of the day, if you’re just cabling the boxes or building service on top of infrastructure you don’t control, you’re limited in the game of scale and efficiency.” $NBIS already has that full stack today.

在 X 上查看原推 ↗