Summary of Key Points
This article focuses on Tencent's AI strategy and compares it with the choices made by companies like Alibaba and Apple. The main argument is that Tencent will not follow the path of becoming an “AI token factory” (renting out computing power to provide basic services). Instead, it will prioritize using its computing power for its own developed large models (such as Hunyuan) and internal AI applications (like WorkBuddy), only considering renting out its power as a last resort. Tencent's strategy is to leverage the strengths of its consumer products and earn money through differentiated AI applications, rather than engaging in cost-competitive battles in areas where Alibaba and Huawei excel. Although the current investment in AI is substantial, Tencent has a backup plan; it will not bet its all like Alibaba and will not dilute its shares.
1. Tencent’s AI Strategy Priorities: “Make Chips First, Then Sell the Products”
Tencent has a clear order of priority for allocating its computing power:
1. Top Priority: Developing the Hunyuan Large Model – Just as you need to grind flour before making bread.
2. Second Priority: Internal AI Applications – Such as WorkBuddy (to improve employee efficiency), CodeBuddy (for coding assistance), and WeChat Mini Programs (intelligent assistants), which are direct products that serve its core businesses.
3. Third Priority: Renting Out Computing Power – This is a backup option, like selling the flour that doesn’t get sold.
Why this order? Because renting out computing power is like “selling potatoes” (a basic product), where you have no control over prices and can only compete on cost. In contrast, developing models and applications is like “selling chips” (processed products), which allow you to earn money through user experience and differentiation. Tencent clearly states, “We could recoup our investment by renting out computing power immediately, but that’s not what we want to do.”
2. Why Doesn’t Tencent Become an “AI Token Factory”?
An “AI token factory” specializes in producing and selling AI computing power (for example, a service that generates 1000 words/images for a fee). Alibaba has made this its primary strategy, with the CEO personally leading the Token Hub division. However, Tencent doesn’t want to go down this path for three reasons:
1. Fierce Homogeneous Competition: Tokens are standardized products, and the only way to compete is on cost. Tencent doesn’t have its own GPU chips (while Alibaba, Huawei, and Baidu do), so it can’t compete on this front.
2. Mismatch in Business Focus: Over its 27-year history, Tencent has never succeeded with cutting-edge technologies (like chips); it excels at understanding user needs (through products like WeChat and Honor of Kings). Trying to compete with Alibaba and Huawei in the B2B computing power market (tendering, channels, implementation) would mean using its weaknesses against their strengths. For example, Tencent Cloud lost 20 billion yuan in competing with Alibaba for projects.
3. Low Profit Margins: The operating margin of China’s cloud business is around 10% (Alibaba Cloud aims for 20% in five years), while Tencent’s consumer businesses (WeChat, games) have margins close to 40%. The revenue from Honor of Kings alone exceeds what Alibaba Cloud has earned in 17 years. There’s no need to earn money in a less profitable way.
3. Why Tencent Needs to Develop Large Models but Not Chips
Many people ask if Tencent’s lack of its own GPU chips will affect its competitiveness. The answer is: Developing models is essential, but not chips.
- Why Models? If Tencent only focused on AI applications (like game AI) without its own models, its applications would lack differentiation. Using someone else’s models would make its products indistinguishable from competitors’ and result in lower prices.
- Why Not Chips? Tencent doesn’t aim to compete on the lowest chip costs; as long as the costs are “low enough,” it’s more cost-effective to buy chips from companies like Huawei or Cambricon. Its goal is to use models to support differentiated applications, not to offer the cheapest computing power.
4. Tencent vs. Alibaba vs. Apple: Different Strategies Based on Different “Competence Circles”
The AI strategies of these three companies are fundamentally different; they each focus on what they do best:
- Alibaba: Can only rely on the “token factory” strategy because its e-commerce growth has slowed, and it needs new businesses for growth. It also has a strong B2B foundation, so it bets on computing power.
- Apple: Doesn’t invest in large data center models or rent out computing power; it focuses on small, device-specific models (such as AI features in iPhones). With over a billion iPhone users, it controls the user access point, allowing it to charge a premium regardless of the quality of models.
- Tencent: Takes a middle path, developing models to support its applications while avoiding the cost competition of the “token factory” model. Its strengths lie in its consumer ecosystem (WeChat, games, video platforms), and using AI to enhance the user experience of these services, which is more profitable.
5. Risks of Tencent’s AI Investment: Betting on Growth with a Backup Plan
Tencent has made significant investments in AI this year (52.8 billion yuan in Q2 capital expenditures, with an additional 51.4 billion yuan in prepaid for computing power), resulting in a negative free cash flow for the first time, and it has reduced share repurchases. However, the article suggests it won’t issue new shares. The reasons are:
1. Temporary Investment: Liu Chiping (Tencent’s CEO) has stated that AI-related capital expenditures are concentrated this year and next year, not annually.
2. Backup Plan: If AI applications don’t succeed, Tencent can reduce its investments (for example, by buying fewer chips next year) and continue to rely on its profitable consumer businesses.
3. Comparison with Alibaba: Alibaba is “betting its all” on the token factory strategy, with no backup plan, while Tencent is “betting on growth” and can withdraw if necessary.
The article also points out that the current AI investment may not align with Buffett’s “1 Dollar Rule” (creating at least 1 dollar in value for every dollar retained), but Tencent’s bet is on the future monetization of AI applications (such as AI advertising and games), which could be more profitable than internal tools like WorkBuddy.
Conclusion
Tencent’s AI strategy is about leveraging its strengths and avoiding its weaknesses. It focuses on its consumer products and using models to support differentiated applications, avoiding homogenized computing power competition. Although the investment is substantial, it has a clear prioritization and a backup plan. It won’t invest everything like Alibaba and won’t dilute shareholder equity. For ordinary users, they may see AI features in WeChat and games in the future, but they won’t see Tencent selling computing power services. This is Tencent’s “smart move”: it avoids competing in areas it’s not good at and focuses on what it does best.