听埃隆·马斯克在 SpaceX 业绩电话会上的发言，仿佛是一部畅销科幻小说的蓝本。但这家火箭、人工智能和全球互联公司在现实世界中所做的，却是实打实的大生意。这对我们投资组合中一大批权重股都有深远影响。周三，SpaceX 股价飙升 7%，此前该公司公布了强劲的第三季度业绩，并对年底及 2027 年给出了乐观展望。
Listening to Elon Musk on SpaceX's earnings call sounded like the basis of a best-selling science fiction novel. But what this rocket, artificial intelligence, and global connectivity company is doing in the real world is big business. And there are far-reaching implications for a bunch of the biggest stocks in our portfolio. SpaceX shares are soaring 7% on Wednesday after the company reported strong third-quarter results with a robust outlook through year-end and into 2027.
马斯克这个庞大的心血结晶——他希望将其与旗下电动车和太阳能公司特斯拉合并——让 SpaceX 能够接触并洞察当前推动股市的几乎所有行业，几乎覆盖了英伟达首席执行官黄仁勋所谈的“五层 AI 蛋糕”的每一层，从电力到半导体，到数据中心基础设施，再到 AI 模型和应用。加上马斯克的远见卓识和点燃投资者热情的能力，这场业绩电话会堪称任何市场参与者的必读材料。
Musk's sprawling brainchild, which he wants to merge with his EV and solar company Tesla, gives SpaceX access and insight into so many of the various industries currently powering the stock market — pretty much every level of the five-layer AI cake that Nvidia CEO Jensen Huang talks about, from power to semiconductors, to data center infrastructure, to AI models and applications. Layer in Musk's visionary nature and ability to get investors excited, and you're left with an earnings conference call that should be considered required reading for any market participant.
毕竟，AI 如今触及市场和更广泛经济的方方面面，无论你是自上而下的宏观投资者，还是自下而上的基本面投资者，都需要对 AI 支出未来走向持有某种看法。现在，让我们透过马斯克的发言，来推演“俱乐部”持仓股在 AI 蛋糕各层面的映射。数据中心供应商马斯克表示，与 AI 繁荣早期不同，半导体供应不再是制约算力供给的主要问题。
After all, AI now touches so much of the market and economy more broadly that whether you're a top-down macro investor or bottom-up fundamental investor, you need to have some view on where AI spending goes from here. Now, let's examine what Musk said as a read-through to the Club names that sit across the various layers of the AI cake. Data center suppliers Unlike the earlier days of the AI boom, semiconductor availability is no longer the major issue constraining compute supply, Musk said.
“我认为人们开始意识到，部署超级智能的瓶颈更多在于电力、散热和基础设施，而非芯片，”他说，并使用了唐纳德·特朗普总统偏爱的人工智能术语。9 月，特朗普签署行政命令，指示所有行政部门和机构用“SI”取代“AI”。
"I think people are starting to realize the constraint on deployment of super intelligence is more in power and cooling and infrastructure than it is in chips," he said, using President Donald Trump's preferred term for artificial intelligence. In September, Trump signed an executive order instructing all executive branch departments and agencies to use "SI" instead of AI.
马斯克已在 SpaceX 内部接受了总统的重新命名，并在本文的多处引述中使用了该术语。马斯克关于制约因素的评论与《商业内幕》本周早些时候的一份报告高度一致，该报告称亚马逊正争相购买 AI 硬件以确保供应充足，但却缺乏将这些算力真正上线所需的电力基础设施。马斯克表示，SpaceX 已重新部署了部分工程师——正是这些人推动了公司火箭飞船创新的巨大成功——来解决电力和散热瓶颈。
Musk has embraced the president's rebrand internally at SpaceX, and he uses the term in multiple quotes throughout this story. Musk's constraint commentary is very much in line with a Business Insider report from earlier this week, which said Amazon is racing to buy AI hardware to ensure it has enough, even as it lacks the power infrastructure required to actually get that compute online. Musk said SpaceX has redeployed some of its engineers — the same ones behind the company's wildly successful rocket-ship innovations — to solve the power and cooling bottlenecks.
“这是一个能够取得比所有太空竞争对手加起来还要大一个数量级成果的团队，现在将同一个团队应用于陆地算力的部署。所以，这有点像纽约洋基队去打少棒联赛。这不公平。”结果，他声称 SpaceX “能够以比任何其他公司快得多的速度部署电力散热基础设施，并且在某个时刻可能会比其他所有公司加起来还要快。”
"It's a team that has been able to achieve results that are an order of magnitude greater than the rest of all competitors in space combined and applied that same team to the deployment of terrestrial compute. So, this is a bit like the New York Yankees playing in Little League. It's not fair." As a result, he claimed SpaceX is "able to deploy power cooling infrastructure at a rate that is vastly faster than anyone else and may, at some point, be faster than everyone else combined."
据马斯克透露，SpaceX 计划在今年年底前投入 2.5 到 3 吉瓦的算力。展望明年，马斯克表示他“相当有信心”能在 2027 年再增加 10 吉瓦算力，但暗示可能实现的规模远不止于此。1 吉瓦电力可为超过 75 万户家庭提供一年的用电。“存在一种非零可能性……即 SpaceX 在明年拥有超过 20 吉瓦的超级智能算力上线。可能性很小。过去是零，现在不再是零了，”他说。
SpaceX plans to bring on between 2.5 and 3 gigawatts of compute power by the end of the year, according to Musk. Looking to next year, Musk said he is "fairly confident" in being able to bring 10 GW of additional compute in 2027, but suggested much more could be possible. A gigawatt of power can provide electricity to more than 750,000 homes for a year. "There's a non-zero possibility ... of SpaceX having over 20 gigawatts of super intelligence compute online next year. A small possibility. It used to be zero.
我们为什么关心？因为如果他要实现目标，将需要更多所需组件。虽然 SpaceX 似乎能自行生产部分所需部件——如燃气轮机用的叶片和导叶——但这些雄心勃勃的目标也指向了通用电气 Vernova 和伊顿 等公司的潜在需求上行。
It's not zero anymore," he said. Why do we care? Because if he is going to achieve his goals, he is going to need a lot more of the components needed to do it. And while SpaceX appears able to make some of what it needs in-house — like blades and vanes for gas-powered turbines — the ambitious targets also point to upside demand for companies like Club names GE Vernova and Eaton .
GE Vernova 是一家领先的重型燃气轮机制造商，而 Eaton 则提供各种用于电网及数据中心内部的电气和冷却设备。SpaceX 的盈利情况也为 Meta Platforms 周三晚发布的报告提供了重要的参考依据。在人工智能领域，SpaceX 拥有社交媒体平台 X、最近收购的编程辅助工具 Cursor，以及聊天机器人 Grok。今年 8 月，SpaceX 正式推出了名为 Grok Bot 的人工智能产品，该产品与 Meta 的 Muse 类似。
GE Vernova is a leading maker of heavy-duty gas turbines, and Eaton provides all sorts of electrical and cooling equipment used across the grid and inside data centers. Agentic demand SpaceX's earnings also served as an appetizer of sorts for Meta Platforms' report on Wednesday night. Within its AI business, SpaceX has social media site X, recently acquired coding assistant Cursor, and the Grok chatbot. In August, the company officially launched an agentic AI offering known as Grok Bot, which is similar to Meta's Muse.
这类智能系统能够在几乎无需人工干预的情况下完成多步骤任务，相比几年前的生成式聊天机器人来说，已经取得了显著的进步。马斯克表示：“Grok Bot 的发展速度是我见过的最快的——我从未见过任何产品能以如此快的速度成长。”Gwynne Shotwell（SpaceX 的总裁兼首席运营官）在电话会议上指出：“Grok Bot 的用户数量每周都在增加，而非减少；随着我们不断改进该产品的功能，用户的使用频率也在不断提高。”
These agentic systems are capable of completing multi-step tasks with little to no human intervention, and they represent a sizable advancement over the generative AI chatbots of a few years ago. "We're seeing the fastest growth of a product that I've ever seen in my life for Grok Bot," Musk said. I've never seen anything grow this fast. We're also seeing that the usage of Grok Bot grows over time per user. It does not decline. So, as we make Grok Bot more useful, we're seeing users use it more per week, not use it less." Grok Bot "already had paid weekly users of 800,000 earlier this month, more than doubling over the last month," President and COO Gwynne Shotwell said on the call.
马斯克还补充说：“我认为 Grok Bot 最终可能会成为我们超级智能计算资源的主要使用者。”我们将会将马斯克的这一观点与 Meta 周三晚关于 Muse 的说明进行对比。Meta 在 9 月推出的 Muse 广受好评，该应用迅速登上了苹果应用商店下载排行榜的榜首。不过，对于投资者来说，产品的用户留存率才是关键因素；长期来看，使用频率远比初始下载量更为重要。
Musk also said, "I think Grok Bot may end up being what uses most of our Super Intelligence compute." We'll be comparing that commentary against what Meta has to say about Muse on Wednesday night. The Muse launch in September was met with much praise as the app quickly surged to the top of Apple's App Store downloads list. However, commentary on retention will be critical for investors. Usage matters far more to the long-term AI story than initial downloads.
市场也在关注 Meta 是否会公布关于 Muse 使用情况的具体数据。我们非常希望获得这些信息，希望 Meta 能像 SpaceX 一样主动公开相关信息。
The market is also wondering whether Meta will disclose any concrete figures on Muse usage and paid sign-ups. We would love to get that information, so let's hope Meta follows SpaceX's lead in this regard.
同样值得一提的是，我们很想了解Muse对Meta未来的算力需求和支出计划意味着什么。芯片制造商和云服务提供商马斯克对Grok Bot使用情况的评论也指向了市场对支撑AI的半导体持续的需求。它们不再是核心制约因素，并不意味着它们不重要。它们当然重要。这既包括英伟达的图形处理器（GPU）和Alphabet的张量处理器（TPU）等AI加速器，也包括承担主力任务的中央处理器（CPU）和存储芯片。
On a similar note, we're interested to hear what Muse means for Meta's compute needs and spending plans going forward. Chipmakers and cloud providers Musk's commentary on Grok Bot usage also points to sustained demand for the semiconductors that make AI possible. Just because they aren't the central constraint anymore doesn't mean they're not important. They certainly are. This includes both AI accelerators like Nvidia's graphics processing units (GPUs) and Alphabet's tensor processing units (TPUs), along with workhorse central processing units (CPUs) and memory chips.
代理式AI所需的CPU和存储芯片远多于前代系统。就我们的半导体标的而言，这对未来一年是个好兆头。这里说的是GPU领域的英伟达、CPU领域的英特尔、高带宽内存领域的美光科技，以及定制芯片业务（包括谷歌的TPU）的博通。英伟达和博通在网络领域也将受益，因为算力关乎单一芯片或服务器的时代早已过去。最近几周的一大担忧是，所有这些AI基础设施支出能否带来足够的回报，以证明继续加大支出的合理性。
Agentic AI requires way more CPUs and memory than predecessor systems. That bodes well for the year ahead as far as our semiconductor names are concerned. Here we're talking about Nvidia on the GPU front, Intel on the CPU front, Micron for its high-bandwidth memory, and Broadcom for its custom silicon work, including on Google's TPUs. Nvidia and Broadcom also stand to benefit on the networking front, as we are long past the days when compute was about a single chip or server. One of the big concerns in recent weeks has been whether all this spending on AI infrastructure will deliver adequate returns to justify more spending.
在这场争论中，SpaceX的言论表明，算力卖家的定价能力只会越来越强。首席财务官布雷特·约翰森表示：“我们在市场上看到的情况非同寻常。按当前定价，新算力部署的回本周期不到一年。目前定价略高于我们上个季度讨论的每瓦30至50美元区间的高端，这进一步印证了这些地面SI算力投资的吸引力。”这对硬件厂商来说是个好兆头，因为这意味着需求与三个月前一样强劲，甚至更强。
On this debate, SpaceX's comments indicate that pricing power for those selling compute is only improving. "What we're seeing in the market is extraordinary," CFO Bret Johnsen said. "At current pricing, new compute deployments are generating a payback of less than a year. Pricing is currently tracking slightly above the high end of the $30 to $50 per watt range we were discussing just last quarter, which further supports the attractiveness of these terrestrial SI compute investments." That bodes well for the hardware players because it means demand is as strong, if not stronger, than it was just three months ago.
对于拥有云业务的超大规模计算公司，如微软、Alphabet和亚马逊而言，这意味着资本支出的回报可能比华尔街预期的来得更快。下周这些公司发布财报时，我们肯定会留意类似的评论。底线是：尽管我们目前并不激进看多，因为AI交易仍在应对其他不利因素（即政治方面的因素），但SpaceX的财报对AI主题的未来走势构成了极大的支撑。
For the hyperscaler players with cloud businesses like Microsoft , Alphabet, and Amazon , it means the return on capex could come even quicker than Wall Street was expecting. We will certainly be looking out for similar commentary as those reports come in next week. Bottom line While we're not aggressively bullish right now as the AI trade still works through other headwinds (namely those of the political variety), SpaceX's earnings report is incredibly supportive of the theme going forward.
我们承认，马斯克总是提出非常宏大的目标和时间表，比如明年让20吉瓦的算力上线。然而，这并不能否定一个事实：SpaceX眼下在实际层面所看到的，是AI解决方案需求正在加速。这反过来又为那些能够提供算力的厂商带来了强大的定价权，无论是半导体公司，还是让芯片得以运行并保持运行所需的电力和冷却解决方案提供商。
We acknowledge Musk always puts out very ambitious targets and timelines, like the ability to get 20GW of compute online next year. However, that doesn't negate the fact that what SpaceX is seeing on the ground, in the here and now, is accelerating demand for AI solutions. That, in turn, is driving strong pricing power for those able to supply the compute, whether that's the semiconductor companies or the power and cooling solutions providers needed to get and keep chips running.
将此与我们本周早些时候分享的观点——即我们认为这些芯片的长期可用寿命没有变化——结合起来，周二晚间的SpaceX电话会议对多头而言是又一场胜利，因为它进一步证实了AI投资周期的基础面依然非常完好。（吉姆·克莱默的慈善信托持有NVDA、MU、ETN、GEV、META的多头头寸。）
Combine this with the view we shared earlier this week — that we see no change in the long usable life of these chips — and Tuesday night's SpaceX call served as another win for the bulls because it provided further confirmation that the fundamentals of the AI investment cycle remain very much intact. (Jim Cramer's Charitable Trust is long NVDA, MU, ETN, GEV, META. )