这三者之间如何选择，取决于投资者最看重什么。在算力获取方面，OpenAI 最突出：它已提前锁定 10 GW，而 SpaceX 目前的设施容量相当于 1 GW；与此同时，Anthropic 还在以据称每年 150 亿美元的价格向 Musk 的公司租用容量。在产品变现方面，Anthropic 可能最强：其 Claude 编码工具被广泛使用，预计本季度收入为 109 亿美元，若实现将比上季度翻倍。Musk 的 AI 业务在今年前三个月实现了 8.18 亿美元。
SpaceX 可能会在可选性上获胜。如果 AI 的盈利能力不如预期，SpaceX 至少已有 Starlink，这一业务贡献了公司的利润和几乎全部收入，而且小行星采矿和火星殖民仍提供上行空间。三家公司的边界正在变得模糊，但目前这场竞争仍未分胜负。一些投资者会押注算力或收入等基本面；另一些则可能只是选择拥有最好气质的创始人。
SpaceX, OpenAI and Anthropic are all circling the same IPO market, a setup that could echo the overheating episodes of 2000 and 2021. The key risk is scale: if each company sells just 5% of its implied valuation, the market could absorb more than $180bn of new equity, far above the $106bn raised in 2000, or about $200bn in today’s money.
The choice between them depends on what investors value most. On compute access, OpenAI stands out with 10 gigawatts already secured ahead of schedule, compared with SpaceX’s current 1GW footprint, while Anthropic is paying Musk’s company an implied $15bn a year to rent capacity. On product monetisation, Anthropic may look strongest: its Claude coding tools are widely used, and its expected quarterly revenue of $10.9bn would be about double the prior quarter. Musk’s AI segment made $818mn in the first three months of the year.
SpaceX may win on optionality. If AI disappoints, Starlink already provides the company’s profit and nearly all of its revenue, and there is still upside from asteroid mining or Mars. The boundaries between the three firms are blurring, but for now the contest is still open. Some investors will back fundamentals such as capacity or revenue; others will simply choose the founder with the best vibe.