"China Outpacing Us on AI": Why Musk Bought an Autonomous Coding Startup

modinfo1 pts0 comments

"China Outpacing Us on AI": Why Musk Bought an Autonomous Coding Startup - Seoul Economic DailySkip to main content<br>Seoul Economic Daily

Seoul Economic Daily

KST<br>Seoul Economic DailyKorea's Leading Business News Since 1960

한국어

Menu

Saved ArticlesNewslettersAI KEYSettings

|Summary|Save<br>AAA

"Anthropic is currently No. 1, OpenAI is No. 2, Google is No. 3, and the Chinese open-source model is No. 4. xAI is only around No. 5."<br>Elon Musk, CEO of SpaceX and Tesla, reportedly made this assessment of global AI capabilities during his testimony in a recent civil lawsuit he filed against OpenAI. He diagnosed that while U.S. Big Tech firms are competing in the frontier AI race, Chinese companies are also catching up fast. Musk also offered a self-assessment that Grok, the AI model of SpaceX subsidiary xAI, ranks a notch below China.<br>Not long after, a major deal was unveiled. After Musk coolly acknowledged his lagging ranking, SpaceX announced Monday that it would acquire Anysphere, the parent company of U.S. autonomous coding agent startup "Cursor," in a deal worth $60 billion (approximately 92 trillion won).<br>It is worth taking a close look at why Musk, who said he ranks fifth in the AI race, would spend an astronomical sum to bring this company into his fold at this point in time. Analysts say the move conceals a cost competition that has swept the global AI market, as well as a strategic bid to seize the business-to-business (B2B) AI ecosystem.<br>Surge in Chinese AI Use Amid Token Cost Burden<br>To understand the backdrop of Musk's sense of crisis, one must examine the changing "way" global companies use AI. Until last year, U.S. large corporations and startups mainly used high-performance AI models (frontier AI models) such as OpenAI's ChatGPT or Anthropic's Claude.

DeepSeek. Reuters-Yonhap News

Elon Musk, CEO of SpaceX and Tesla. Reuters-Yonhap News

This year, the cost issue has come to the fore. The token (the basic unit of AI sentence processing) costs that snowball as the scope of AI use expands have become a serious financial burden for companies. It is precisely this gap that low-cost Chinese open-source AI models such as Alibaba and DeepSeek have penetrated. Recently, U.S. companies have begun adopting tools that automatically switch between AI models depending on the difficulty of the task, instead of handling all work with expensive top-tier models.<br>This is a hybrid approach that deploys overwhelmingly cheaper Chinese open-source models for simple repetitive tasks or primary text processing, and calls on ChatGPT or Claude only when complex reasoning is required. In fact, the share of DeepSeek AI usage on the platform of startup Vercel jumped sharply from 1% in April to 17% in May.<br>According to OpenRouter, a startup that processes AI queries, DeepSeek models have been called on the most since mid-May. OpenRouter said that among high-paying customers, the number of tokens processed by open-source models grew four times faster than proprietary models such as ChatGPT, Claude, and Gemini between the fall of 2025 and the spring of 2026.<br>This situation is also related to OpenAI's financial position. According to a report by The Information, OpenAI spent $3.7 billion in the first quarter, more than half of its total revenue of $5.7 billion. Its first-quarter operating loss was $9.3 billion, and its net loss was $21.3 billion.<br>Musk Takes Aim at Developer-Oriented AI<br>Musk's acquisition of Cursor is a strategic detour to escape the margin burden of AI companies. He has turned toward a definite target, such as developers, rather than the business-to-consumer (B2C) AI chatbot market.<br>Cursor is an autonomous coding tool that global companies such as Nvidia, British Airways, and Deloitte, as well as major AI research labs, use as an essential part of software development. It has effectively established itself as the standard in the field of coding agents that write program source code and catch errors.<br>It can be seen that Musk has devised a strategy to combine xAI's core technology with Cursor's autonomous coding engine. This is because, if he dominates the work environment of developers worldwide, he can build a lock-in effect that cannot easily be changed no matter how hard low-cost Chinese models attack.<br>Of course, it remains uncertain whether this gamble will pay off. Musk's AI business division posted $3.2 billion in revenue last year, but suffered a loss of $6.4 billion, twice that amount. In the first quarter of this year, it also recorded a $2.5 billion deficit on revenue of $818 million. On top of this, there are mountains of areas requiring further investment, such as the construction of a semiconductor plant called Terafab.<br>The market is optimistic about Musk's AI drive. Goldman Sachs and Morgan Stanley, the underwriters of the SpaceX initial public offering (IPO), forecast that SpaceX's annual revenue will reach approximately $160 billion in 2028, driven by AI growth.<br>※ By subscribing to "Kim Ki-hyuk's...

musk models billion coding chinese spacex

Related Articles