We Think the SpaceX IPO Is Overvalued
We Think the SpaceX IPO Is Overvalued
我们认为 SpaceX 的首次公开募股（IPO）估值过高
Why Is Morningstar So Bearish on SpaceX’s IPO? We value SpaceX at $63 per share, a 53% discount to the upcoming IPO’s offering price. Our valuation is the result of mathematics more than skepticism. With such a wide range of possible outcomes for the company’s financial future, we created forecasts and valuations for three scenarios and probability-weighted them. 为什么晨星（Morningstar）对 SpaceX 的 IPO 如此看空？我们将 SpaceX 的估值定为每股 63 美元，较即将到来的 IPO 发行价折让 53%。我们的估值更多是数学计算的结果，而非单纯的怀疑。鉴于该公司未来财务状况存在多种可能性，我们针对三种情景分别进行了预测和估值，并进行了概率加权。
Even at $63 per share, we give SpaceX a lot of benefit of the doubt in two of the three scenarios, in which we assume the company can achieve a rapidly reusable Starship rocket enabling multiple launches per week and successfully commercialize data centers in space. Neither of these engineering problems has been solved, and we don’t expect them to be until at least 2028. 即使在每股 63 美元的估值下，我们在三种情景中的两种里，依然给予了 SpaceX 很大的宽容度。在这些情景中，我们假设该公司能够实现星舰（Starship）火箭的快速重复使用，从而实现每周多次发射，并成功实现太空数据中心的商业化。这两个工程难题目前都尚未解决，我们预计至少要到 2028 年才能实现。
In our most optimistic “moonshot” scenario, the company would be worth $1.97 trillion, or $154 a share. That’s 14% above the offering price and a level the shares might even reach in the short term after their public launch, given widespread investor enthusiasm about SpaceX, artificial intelligence infrastructure, and the IPO. However, we assign this scenario, in which both Starship is reusable and scaled orbital data centers are highly successful, a 7% chance of happening, which is one reason our final fair value estimate of $63 is much lower than $154. 在我们最乐观的“登月”情景中，该公司的估值将达到 1.97 万亿美元，即每股 154 美元。这比发行价高出 14%，考虑到投资者对 SpaceX、人工智能基础设施以及此次 IPO 的广泛热情，股价在上市后短期内甚至可能达到这一水平。然而，我们认为这种情景（即星舰实现重复使用且大规模轨道数据中心取得巨大成功）发生的概率仅为 7%，这也是我们最终给出的 63 美元公允价值远低于 154 美元的原因之一。
What Would It Take for SpaceX to Get a ‘Buy’ Rating From Morningstar? Morningstar’s consistent and independent equity research methodology is designed to help long-term investors ascertain a stock’s intrinsic value and weigh it against the prevailing market price. What’s more, with a Very High Uncertainty Rating for SpaceX, we would only think the shares offered a compelling risk-adjusted return (and thus garner a 5-star Morningstar Rating) below a 50% discount to our fair value estimate. To illustrate, to get 5 stars from Morningstar at the $135 IPO offering price, our fair value estimate, all else equal, would have to be $270 per share. Here are the assumptions we made to arrive at our more realistic valuation. SpaceX 需要满足什么条件才能获得晨星的“买入”评级？晨星一贯且独立的股票研究方法旨在帮助长期投资者确定股票的内在价值，并将其与当前市场价格进行权衡。此外，由于 SpaceX 的不确定性评级极高，我们认为只有当股价低于我们公允价值估算的 50% 折让时，该股才具备令人信服的风险调整后回报（从而获得晨星 5 星评级）。举例来说，若要在 135 美元的 IPO 发行价下获得晨星 5 星评级，在其他条件不变的情况下，我们的公允价值估算必须达到每股 270 美元。以下是我们得出这一更现实估值所依据的假设。
What Are the Building Blocks of the Fair Value Estimate? Our analysis of fair value focuses on the profit drivers and forecasts for the company’s businesses described in our full report. Some items that add up to our fair value estimate don’t change with our scenarios. We assume the firm will raise $85.7 billion for 639 billion shares offered in the IPO, which amounts to $6.50 per share of our fair value estimate. Add to that its existing $1.80 of cash and investments, minus $2.30 per share of debt. 公允价值估算的构成要素是什么？我们的公允价值分析侧重于我们在完整报告中描述的公司业务的利润驱动因素和预测。构成我们公允价值估算的一些项目不会随情景变化。我们假设该公司通过 IPO 发行 6390 亿股股票筹集 857 亿美元，这相当于我们公允价值估算中每股 6.50 美元。再加上其现有的每股 1.80 美元的现金和投资，减去每股 2.30 美元的债务。
Our valuation for the core space and connectivity businesses adds around $40 per share (we believe a slower-growth scenario for Starlink after 2028 would reduce that estimate by $5). Our probability-weighted average of the three wide-ranging AI scenarios adds $16.50 to our overall valuation estimate, which we view as more akin to the value of a call option on the commercialization of orbital AI infrastructure. The math: $6.51+$1.80-$2.30+$40.00+$16.50=$62.51 (we rounded up to $63.00). 我们对核心太空和连接业务的估值增加了约每股 40 美元（我们认为如果星链在 2028 年后增长放缓，该估值将减少 5 美元）。我们将三种广泛的人工智能情景进行概率加权平均，为我们的整体估值增加了 16.50 美元，我们认为这更类似于轨道人工智能基础设施商业化的看涨期权价值。计算公式为：6.51 + 1.80 - 2.30 + 40.00 + 16.50 = 62.51 美元（我们向上取整为 63.00 美元）。
What Assumptions About SpaceX Drive the Forecast? In all three scenarios, we apply base-case forecasts provided by PitchBook for the space and connectivity businesses, in which we assume that by 2035, SpaceX will be able to launch 340 Starship missions (nearly one a day), and that the reusability rate of the rockets on these missions reaches 85%, extending cost and time savings beyond the reuse of the boosters to the upper stage spacecraft. Then we built forecasts for three scenarios to ascertain what the AI business might achieve. 哪些关于 SpaceX 的假设驱动了这一预测？在所有三种情景中，我们都采用了 PitchBook 为太空和连接业务提供的基准预测。我们假设到 2035 年，SpaceX 将能够执行 340 次星舰任务（几乎每天一次），并且这些任务中火箭的重复使用率达到 85%，从而将成本和时间的节省从助推器扩展到上面级航天器。随后，我们针对三种情景建立了预测，以确定人工智能业务可能取得的成就。
Moonshot: In the first, our most optimistic scenario, SpaceX’s orbital AI platform works, achieves operating cost advantages over terrestrial computing, and eventually deploys and commercializes one-fifth of our forecast of AI infrastructure computing capacity (excluding Russia and China) by 2040. We apply the estimates provided by SpaceX that it can engineer and deploy satellites with the equivalent of 100 kilowatts of AI processing capacity each and eventually fit more than 100 of them into the fairing of a Starship, which we forecast to result in an orbital computing cluster of some 59,000 satellites by 2035, providing the equivalent of 11.6 gigawatts of AI computing capacity and generating $225 billion of annual revenue. 登月：在第一种也是我们最乐观的情景中，SpaceX 的轨道人工智能平台运行良好，在运营成本上比地面计算更具优势，并最终在 2040 年前部署并商业化我们预测的人工智能基础设施计算能力的五分之一（不包括俄罗斯和中国）。我们采用了 SpaceX 提供的估算，即它可以设计并部署每颗具备 100 千瓦人工智能处理能力的卫星，并最终将 100 多颗卫星装入星舰的整流罩中。我们预测，到 2035 年，这将形成一个由约 59,000 颗卫星组成的轨道计算集群，提供相当于 11.6 吉瓦的人工智能计算能力，并产生 2250 亿美元的年收入。
No Go: In our downside scenario, orbital data centers won’t work or offer any advantage over terrestrial ones. We surmise that the company, having invested tens of billions to find this out, would cut bait on the project sometime around 2028, the way management walked away from plans to build multiple small-car factories at Tesla. We assume SpaceX would continue to commercialize its terrestrial Colossus data center but would not take a meaningful share of global computing capacity. 不行：在我们的下行情景中，轨道数据中心将无法运行，或无法提供优于地面数据中心的任何优势。我们推测，该公司在投入数百亿美元发现这一点后，会在 2028 年左右放弃该项目，就像管理层当年放弃在特斯拉建设多个小型汽车工厂的计划一样。我们假设 SpaceX 将继续将其地面的 Colossus 数据中心商业化，但不会占据全球计算能力的显著份额。
Minimum Viable Product: In our most likely scenario, orbital data centers prove viable, subject to some capacity constraints, but also benefit from SpaceX’s decreasing cost to launch large payloads on Starship, and the company successfully deploys and commercializes around 4% of our forecast AI computational capacity—probably best serving those use cases that can tolerate higher data transmission latency. It’s a project that requires some unproven engineering to succeed, but we do see SpaceX as the best positioned to pursue it. In this scenario, we estimate SpaceX would engineer and deploy satellites with the equivalent of 50 kilowatts of AI processing capacity each and eventually fit more than 90 of them into the fairing of each Starship launch, resulting in an orbital computing cluster of some 48,000 satellites by 2035, providing the equivalent of 2.4 gigawatts of AI computing capacity and generating $47 billion of annual revenue. 最小可行性产品：在我们最可能发生的情景中，轨道数据中心被证明是可行的，尽管受到一些容量限制，但也受益于 SpaceX 使用星舰发射大型载荷成本的降低。该公司成功部署并商业化了我们预测的人工智能计算能力的约 4%——这可能最适合那些能够容忍较高数据传输延迟的用例。这是一个需要一些未经证实的工程技术才能成功的项目，但我们确实认为 SpaceX 是最有能力推进该项目的公司。在这种情景下，我们估计 SpaceX 将设计并部署每颗具备 50 千瓦人工智能处理能力的卫星，并最终在每次星舰发射中装入 90 多颗卫星，到 2035 年形成一个由约 48,000 颗卫星组成的轨道计算集群，提供相当于 2.4 吉瓦的人工智能计算能力，并产生 470 亿美元的年收入。
How Do the Scenarios Add Up to the Fair Value Estimate? We give the optimistic Moonshot scenario a 7% chance of happening. This represents the combined probability that Starship is reusable 85% of the time… 这些情景如何汇总为公允价值估算？我们认为乐观的“登月”情景发生的概率为 7%。这代表了星舰 85% 的时间可重复使用的综合概率……