An acquisition, four post-2000s young people, achieved financial freedom.

Elon Musk has made another heavy investment in AI. It is reported that SpaceX has signed a final merger agreement with Anysphere, the developer of the AI programming tool Cursor, and will complete the acquisition through an all - stock transaction. The implied valuation of Cursor in this transaction is as high as $60 billion (approximately RMB 406.2 billion).

In fact, this acquisition was disclosed to the public as early as April this year. Coincidentally, the signing of this agreement happened to coincide with SpaceX's listing. After the news was announced, SpaceX's stock price soared by more than 16% during intraday trading, and its market value briefly exceeded $2.94 trillion, surpassing Microsoft. By the close of trading, the company had overtaken Amazon and ranked as the fourth - largest company in the United States by market value. Calculated at the IPO issue price of $135 per share, its cumulative increase has been nearly 50%.

Cursor is considered one of the most dazzling AI applications globally in the past two years. The company was co - founded in 2022 by four post - 2000 geniuses from Silicon Valley, namely Michael Truell, his MIT classmate Sualeh Asif, Arvid Lunnemark, and Aman Sanger.

As an AI application, Cursor has shown amazing profitability. The official website shows that in January 2025, the company's annualized revenue exceeded $100 million; in June, it reached $500 million; in November, it exceeded $1 billion; by February 2026, this figure had been updated to $2 billion.

However, limitations have also emerged. In May 2025, Claude Code was released, and the discussion of "models devouring applications" became more intense. As the underlying models become more powerful, where is the moat for the application layer? Is it just a shell of the model's capabilities? The fate of Cursor may ring alarm bells for many application companies.

A $60 - billion Heavy Investment

According to the 8 - K filing submitted by SpaceX to the U.S. securities regulatory agency, the existing common and preferred shareholders of Anysphere, the developer of Cursor, will receive Class A common shares of SpaceX through a share - exchange based on the implied equity valuation of the company at $60 billion. This transaction does not involve any cash payment. The transaction is expected to be completed in the third quarter of 2026, subject to regulatory approval and the satisfaction of other customary closing conditions.

In fact, this acquisition was not achieved overnight. In April this year, SpaceX disclosed that it had obtained the pre - emptive right to purchase Cursor. However, due to the approaching IPO at that time, the acquisition plan was temporarily put on hold.

However, Musk still showed great sincerity at that time. The two parties agreed on a transaction arrangement with an option. That is, what SpaceX obtained was not an immediately effective acquisition agreement, but an acquisition option: if it decides to exercise the option, Cursor will be officially incorporated into Musk's business map at a valuation of $60 billion; if it ultimately gives up the acquisition, it will need to pay Cursor $10 billion as consideration for deepening strategic cooperation. This means that regardless of the final outcome, Cursor can get at least $10 billion as a "deposit".

With the signing of this final agreement, this transaction framework has been officially implemented. According to the announcement, the shareholders of Cursor will receive the transaction consideration entirely in the form of SpaceX stocks. The specific share - exchange ratio will be calculated based on Cursor's $60 - billion valuation and the volume - weighted average price (VWAP) of SpaceX stocks over seven consecutive trading days before the completion of the transaction.

As for why Musk made such a large - scale investment, the reason lies in strengthening his own key capabilities. Previously, he publicly admitted that his xAI was lagging behind competitors in the field of AI programming tools and started to recruit engineers from Cursor. After the completion of this acquisition, SpaceX will combine Cursor's product capabilities and distribution channels for professional software engineers with its Colossus supercomputer, which has a million - level H100 equivalent computing power, to create "the world's most practical AI model" and further strengthen its capabilities in the software development field.

Musk's "AI + aerospace" business map has clearly emerged.

Once the Most Profitable AI Tool

Looking back at Cursor itself, I personally feel a bit disappointed with such an ending.

As one of the fastest - growing and most commercially successful AI applications in recent years, it was once regarded as the next Adobe.

Cursor was founded in 2022. Its founders, Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger, are all post - 2000s who dropped out of MIT to start a business. In the second year of its establishment, institutions such as OpenAI's venture fund and the startup accelerator Neo led its seed round of approximately $8 million. Cursor has a very high conversion rate of paid users. By June 2025, its annual recurring revenue (ARR) had exceeded $500 million.

It is worth mentioning the company's revenue - generating ability. In just three years since its establishment, its annualized revenue has exceeded $2 billion. The speed from zero to $2 billion has refreshed the record of the B2B software industry, and its per - capita revenue even exceeds that of most Silicon Valley star companies. Before the transaction, Cursor was still preparing for a new round of financing, rumored to plan to raise approximately $2 billion, with a valuation of over $50 billion. However, Musk finally offered a price of $60 billion, a premium of over 20%.

However, as a programming application tool, there has always been an unavoidable contradiction in Cursor: although it has an excellent product experience, it does not possess the most core capabilities.

Cursor's business model is quite simple. Users pay subscription fees, and Cursor then purchases the Claude model API from Anthropic. At its peak, Cursor contributed approximately 40% to 50% of Anthropic's revenue. It is more appropriate to say that it played the role of Anthropic's best sales channel rather than an independent AI company.

This is where the danger was planted. In May 2025, Anthropic released Claude Code. In just half a year, the annualized revenue of Claude Code exceeded $1 billion. By the end of the year, its user base officially exceeded that of Cursor. At the beginning of 2026, Anthropic announced that it would tighten third - party access rights, which was obviously another heavy blow.

Regarding the difference between the two, a developer posted on Reddit, "Compared with Claude Code, Cursor's operating efficiency is much lower; when searching for relevant content in the code library, its performance is also significantly worse. The reason is that if Cursor wants to use the Claude model, it must pay API fees. To save costs, Cursor has to reduce costs by optimizing or deleting some relevant data, but this will slightly reduce its efficiency. In addition to costs, because models and functions are usually interrelated and used together. For example, the multi - agent processing method adopted by Claude is actually more suitable for the Claude model itself."

In the process, Cursor didn't just sit back and do nothing. Founder Michael Truell urgently held an all - staff meeting, hoping to create a programming model with performance comparable to Claude but at a lower cost. A few months later, Composer 2 was released. Official data shows that it outperformed Claude Opus in multiple programming benchmark tests, once exciting the developer community.

But a dramatic scene soon occurred. A developer found that the model ID behind Composer pointed to "kimi - k2p5 - rl - 0317 - s515 - fast" when debugging the API. Musk later replied personally, "Yeah, it’s Kimi 2.5."

Models Devouring Applications

In the past few years, the narrative of "models devouring applications" has always fluctuated in the market.

Cursor's rise proves that AI applications can completely create amazing revenue, ultimate user experience, and huge commercial value; but its final fate also exposes another side: when the core capabilities are built on the underlying models, the application layer always has difficulty getting rid of the constraints of suppliers.

Different from the traditional SaaS model of charging by account or seat, AI original manufacturers tend to adopt a business model of charging by usage volume and effect. For example, Claude Opus 4.5 is billed per million Tokens, and manufacturers such as Sierra even directly charge according to the problem - solving rate and task - completion effect and participate in value sharing. For large - model manufacturers, there is a complete opportunity to bypass the software middle layer and directly reach the end customers.

This has even directly led to concerns in the capital market about the software industry. Although the scale of this market is extremely large. The global software industry output value represented by North America has approached $500 billion. IDC predicts that by 2029, the global SaaS market scale will approach $1 trillion. However, now, this huge cake is facing the risk of being re - divided by AI original manufacturers.

At the beginning of this year, the explosion of the OpenClaw ecosystem once suppressed the above - mentioned pessimistic view, but it was still the model manufacturers that first enjoyed the dividends of this round. In the Hong Kong stock market, the valuations of companies such as MiniMax and Zhipu AI have been continuously rising; the Dark Side of the Moon has seen rapid growth in overseas API revenue due to the wide - spread use of its K2.5 model in the OpenClaw ecosystem. The company also completed a cumulative financing of more than $1.2 billion in the following more than a month, with its valuation exceeding $10 billion and continuing to move towards a higher level.

Now, the market heat of OpenClaw has subsided to some extent, but the earnings of model companies have not cooled down synchronously. The traffic may ebb, but the usage volume, developer ecosystem, and commercial revenue are still accumulating and continuously translating into a higher pricing in the capital market for the underlying model capabilities. Recently, Kimi has launched a new round of financing, with a pre - investment valuation of $30 billion, a significant increase from before.

This reality may be a bit cruel for most application companies. After all, not every company can get a $60 - billion option like Cursor.

This article is from the WeChat official account "Dongshisi Tiao Capital" (ID: DsstCapital). The author is Wei Xianghui, and it is published by 36Kr with authorization.