Summary of Key Points
Tencent has made two significant moves in the AI field: firstly, it invested 52.8 billion yuan (a year-on-year increase of 176%) to build its AI infrastructure, which resulted in a negative free cash flow (for the first time since 2005); secondly, it participated in the investment of the Swedish company Lovable, valued at $13.3 billion. Lovable's unique feature is its ability to enable non-programmers to create applications using simple language, and it has impressive user data but relies on the Claude model. Domestic companies in this space (such as Lingzhu) are still only in their early stages of funding (angel round), highlighting the significant gap between them. The vibe coding industry presents both opportunities (solving the problem of ordinary people creating software) and challenges (relying on models, technical debt, and competition from large corporations).
I. Tencent's "Dual Approach" to AI: Investing in Infrastructure + Capturing Application Entry Points
Tencent's financial report for the second quarter of this year includes two key figures: capital expenditure of 52.8 billion yuan (mainly for purchasing AI computing power) and a free cash flow of -13.8 billion yuan (all funds were invested). This indicates that Tencent is making a substantial bet on AI infrastructure, essentially stockpiling "AI fuel" in advance.
Investing in Lovable is not just about throwing money around; it aims to fill a gap in the application layer. Tencent already has products like CodeBuddy for professional developers and Tuzi for ordinary creators, but Lovable targets users who have no coding experience at all. By combining these offerings, Tencent covers all aspects of AI development (professional, general, and zero-code). This investment is more like a "call option": if vibe coding becomes popular, Tencent's infrastructure will be valuable, allowing it to provide the necessary resources for those creating software.
II. Why Does Lovable Worth $9 Billion? It Solves the Problem of "Creating Software Without Knowing How to Code"
Lovable's core technology is "vibe coding"—users can simply request something like "create a small store with login and membership features," and it generates a complete application (frontend, backend, and deployment) without the need to install any additional tools. This is different from Cursor, which is designed for professional developers. Lovable aims at the 99% of people who don't code but have ideas.
Its value is evident in its data: it has created 60 million projects since its launch, with 900 million monthly visits, and two-thirds of employees from Fortune 500 companies are using it. Its annual revenue has increased from 200 million to 500 million yuan (with a target of reaching 600 million yuan by August). Although it relies on the Claude model, it doesn't just use it superficially; it encapsulates the model into a user-friendly product and integrates it with databases, payment systems, and deployment tools, making it part of business processes (for example, Adidas uses it to build internal tools). This is why Lovable is valued at $13.3 billion—it has secured a crucial entry point for non-programming users.
III. The "Sweetness and Troubles" of the Vibe Coding Industry: High Valuation but Many Challenges
The industry is promising: Cursor was acquired by SpaceX for $6 billion, Lovable's valuation has doubled, and Bolt.new from StackBlitz generated 40 million in annual revenue within five months.
However, there are also many challenges:
1. The Sword of Damocles of Model Dependencies: Both Lovable and Cursor rely on models (Claude and GPT/Claude respectively). If the model providers adjust prices or restrict access, it could impact the stability of the applications.
2. Technical Debt: AI-generated code may contain hidden bugs, which could lead to long-term issues like building a house on unstable foundations.
3. User Loss Risk: After Cursor changed its pricing model, heavy users faced a monthly fee increase of $200, leading many to switch to Claude Code.
4. Competitive Shift: The industry is shifting towards end-to-end autonomous agents (such as Claude Code), which could dilute the value of vibe coding.
IV. Why Can't Similar Companies in China Emerge? Four Major Barriers
The reason domestic companies like Lingzhu are still in their early stages is due to several barriers:
1. Traffic Monopoly by Large Corporations: WeChat, TikTok, and Alipay control user access, limiting the ability of independent products to break through.
2. Poor Payment Habits: Chinese users are less willing to pay for AI tools, so large corporations prefer free options to retain users, leaving independent companies with limited revenue opportunities.
3. Closed Infrastructure Loop: In the US, there are independent tools like Supabase (backend), Stripe (payment processing), and Vercel (deployment), but in China, these services are often integrated with cloud services and payment platforms, making it difficult for independent developers to complete the entire process.
4. Cautionous VC Investment: Chinese VCs tend to invest more in model layers, hesitating to support independent application layers due to concerns about being acquired by large corporations or model providers.
Conclusion
The essence of vibe coding is to enable non-programmers to create software, but the fate of independent companies in this space may be to be acquired by large corporations or to rely on model providers. Tencent's investment in Lovable reflects its belief in the growing demand for zero-code applications. For domestic players to succeed, they need to overcome barriers related to traffic and payment models. The future of this industry depends on both user demand and the behavior of model providers and large corporations.