9月。一个月后。
马斯克在SpaceX内部全员会议上放了一句话：公司的AI收入将于9月超过其他所有业务收入之和，并且在第四季度大幅领先。
这句话的分量，怎么强调都不过分。SpaceX是什么公司？是全球最大的商业火箭发射公司，是星链运营商，是估值几千亿美元的航天巨头。它的"其他所有业务"——火箭发射、星链、星舰——加在一起，年营收规模在数百亿美元级别。
而AI业务，在两年前还几乎不存在。
8月13日，SpaceXAI正式发布了Grok 4.6。这次升级的核心方向不是推理能力的常规提升，而是长时运行的智能体任务、复杂交互、视觉工作和编码能力的强化。换句话说，Grok正在从"聊天模型"变成"能干活的模型"。
从成本中心到利润中心的转身
两年前，xAI还只是马斯克名下一个烧钱的AI实验室。Grok 1的发布更像是一个玩具——性能落后于GPT和Claude，用户主要是X平台的硬核粉丝。
但事情变化的速度远超所有人预期。2026年2月，SpaceX吸收合并了xAI——这不是简单的组织调整，而是一次战略层面的资源整合。合并之后，xAI获得了SpaceX的算力基础设施、工程团队和分发渠道，而SpaceX则拿到了AI时代的入场券。
现在回头看，这步棋走得极其精准。
Grok的商业化路径，走的不是OpenAI那种纯API+订阅的路线，而是直接嵌入SpaceX的整个产品矩阵：
第一，编程入口。SpaceX以600亿美元收购Cursor之后，Grok成为Cursor的默认模型之一。数千万开发者的注意力和数据，直接转化为Grok的训练燃料和收入来源。
第二，硬件入口。Grok Build、Grok Code等工具直接对接开发者工作流，首周双倍额度的促销策略快速拉新。
第三，企业入口。通过xAI的企业版方案，Grok正在切入企业级AI市场。
"火箭是马斯克的梦想，AI是马斯克的生意。当AI收入超过火箭收入的那一刻，SpaceX就不再是一家航天公司了——它是一家AI公司，顺便造火箭。"—— 一位科技行业分析师
如果马斯克的预测兑现——9月AI收入超过其他所有业务之和——那意味着SpaceX的收入结构将发生根本性的变化。一家以火箭发射起家的公司，最终变成了一家以AI为主要收入来源的公司。这个转变，可能比任何技术突破都更有象征意义。
AI收入为什么涨得这么快？
很多人可能会问：Grok的性能明明不是最强的，为什么收入能涨这么快？
答案在于：AI收入的关键不在模型本身有多强，而在分发渠道有多广。
OpenAI的ChatGPT用户量巨大，但它的分发渠道本质上只有一个——自己的产品。Anthropic的Claude能力很强，但分发渠道主要靠AWS和第三方开发者。
而马斯克的AI版图有什么？X平台（数亿用户）、Cursor（数千万开发者）、SpaceX算力（自建超算）、Tesla（车载入口）、Neuralink（脑机接口）——这些渠道每一个都可以直接转化为AI产品的用户。
更重要的是，这些渠道之间形成了正向飞轮：X平台的数据训练Grok模型，Grok能力提升让Cursor更好用，Cursor的用户数据又反过来改进Grok，Grok变强了又吸引更多企业客户——整个闭环在马斯克的体系内就能完成。
这就是为什么Grok在技术参数上可能还不是全球第一，但收入增长速度却可能是最快的。因为它不是一个孤立的AI产品，而是一个庞大生态系统中的一个环节。
当然，马斯克的收入预测是内部讲话内容，不是已经实现的财报数据。实际情况是否能达到预期，还需要等财报验证。但即便数字有出入，大方向也是明确的：AI正在从SpaceX的"未来业务"变成"当前主业"。
当一家火箭公司的AI收入超过火箭收入的时候，我们可以确定一件事：AI时代的权力结构，已经彻底改写了。
明天见。
September. One month from now.
At a SpaceX all-hands meeting, Elon Musk dropped one line: the company's AI revenue will surpass all other businesses combined in September, and will lead decisively in Q4.
The weight of that statement can't be overstated. What is SpaceX? It's the world's largest commercial rocket launch company, it's Starlink, it's a multi-hundred-billion-dollar aerospace giant. "All other businesses" - rocket launches, Starlink, Starship - combined, represent tens of billions in annual revenue.
And the AI business? Two years ago, it barely existed.
On August 13, SpaceXAI officially launched Grok 4.6. The core of this upgrade isn't routine reasoning improvement - it's strengthened long-running agent tasks, complex interaction, visual work, and coding capability. In other words, Grok is evolving from a "chat model" to a "model that gets work done."
The Pivot From Cost Center to Profit Center
Two years ago, xAI was just a money-burning AI lab under Musk's umbrella. Grok 1's launch felt more like a toy - performance lagging behind GPT and Claude, users mostly hardcore fans on the X platform.
But things changed faster than anyone expected. In February 2026, SpaceX absorbed xAI - this wasn't a simple org change, it was a strategic resource consolidation. After the merger, xAI gained SpaceX's compute infrastructure, engineering teams, and distribution channels, while SpaceX got its ticket to the AI era.
Looking back now, this move was extraordinarily well-timed.
Grok's commercialization path isn't OpenAI's pure API + subscription route. It's directly embedded in SpaceX's entire product matrix:
First, the coding entry point. After SpaceX acquired Cursor for $60B, Grok became one of Cursor's default models. Tens of millions of developers' attention and data directly convert into Grok's training fuel and revenue.
Second, the product entry point. Grok Build, Grok Code, and other tools plug directly into developer workflows. First-week double-credit promotions accelerate user acquisition.
Third, the enterprise entry point. Through xAI's enterprise solutions, Grok is penetrating the enterprise AI market.
"Rockets are Musk's dream; AI is Musk's business. The day AI revenue exceeds rocket revenue, SpaceX stops being an aerospace company - it's an AI company that happens to build rockets." - A Tech Industry Analyst
If Musk's prediction holds - AI revenue surpassing all other businesses combined in September - it means SpaceX's revenue structure undergoes a fundamental shift. A company built on rocket launches ends up with AI as its primary income source. This transformation may be more symbolically significant than any technological breakthrough.
Why Is AI Revenue Growing So Fast?
Many people might ask: Grok clearly isn't the strongest model - why is revenue growing so quickly?
The answer: AI revenue's key driver isn't how strong the model is - it's how broad the distribution channels are.
OpenAI's ChatGPT has massive users, but its distribution is essentially one channel - its own product. Anthropic's Claude is very capable, but distribution relies mainly on AWS and third-party developers.
What does Musk's AI landscape have? The X platform (hundreds of millions of users), Cursor (tens of millions of developers), SpaceX compute (self-built supercomputers), Tesla (in-vehicle entry point), Neuralink (BCI) - each of these channels can directly convert into AI product users.
More importantly, these channels form a positive flywheel: X platform data trains Grok models, Grok's improvements make Cursor better, Cursor's user data in turn improves Grok, stronger Grok attracts more enterprise customers - the entire loop completes within Musk's ecosystem.
This is why Grok may not be the global leader in technical specs, yet its revenue growth could be the fastest. It isn't an isolated AI product - it's one link in a massive ecosystem.
Of course, Musk's revenue prediction is from an internal speech, not realized financial results. Whether reality matches expectations awaits earnings confirmation. But even if the numbers differ, the direction is clear: AI is shifting from SpaceX's "future business" to "current main business."
When a rocket company's AI revenue exceeds its rocket revenue, one thing is certain: the power structure of the AI era has been completely rewritten.
See you tomorrow.
Rockets are Musk's dream; AI is Musk's business. The day AI revenue exceeds rocket revenue, SpaceX stops being an aerospace company.
- A Tech Industry Analyst
