Key Takeaways:
Tencent Holdings Ltd. is commercializing two of its advanced AI models, a decisive break from the cash-burning, scale-at-all-costs strategy that has defined China's AI race and a move that sent its stock up more than 4 percent.
"Tencent has switched ships," founder Pony Ma declared at a May shareholder meeting, signaling a strategic departure that the market is now seeing in action.
The company’s cloud division announced that its Hy3 preview and DeepSeek-V4-Pro models will end their free public beta on May 27, transitioning to a formal commercial service with usage-based charges. The move is validated by early market data; Tencent’s Hy3 Preview model held the top position on the global model router OpenRouter for three consecutive weeks, even after it transitioned from free to paid, showing a clear willingness to pay for quality.
This pivot to monetization reflects a harsh new reality in the AI sector, most recently illustrated by ByteDance’s decision to cut roughly 30 percent of its AI application projects. ByteDance’s AI inference costs in 2025 reportedly exceeded RMB 8 billion, approximately 2.3 times its incremental revenue from AI products. The episode demonstrates that the mobile internet playbook of indiscriminately scaling users is not a viable path forward in the era of costly AI.
For the past several years, Chinese tech giants like Tencent, Alibaba, and Baidu followed a familiar strategy: launch dozens of parallel products and bet on a breakout hit. That logic fails in AI for three structural reasons.
First, unlike mobile apps where marginal costs approach zero, AI inference costs scale directly with usage. Every query incurs real compute and storage costs, meaning user growth can deepen losses. Second, the power of foundational models is swallowing the application layer, with companies like OpenAI and Google often embedding features that smaller tools are built on within months. Third, user switching costs have collapsed, with no social graphs or data ecosystems to lock users in.
Tencent's commercialization is a clear signal that the industry is shifting from scale-first experimentation toward defensibility-first execution. The company's stock jump and a recent HKD500 million share repurchase underscore its confidence in the new strategy. The new moats in AI are not user numbers, but proprietary data, deep integration into workflows, and control over distribution channels.
The AI application market is now bifurcating. On one side are capital-intensive, high-risk general-purpose platforms. On the other, a more durable segment is emerging: precision AI systems built for specific workflows with high retention and clear return on investment. By charging for its high-performing models, Tencent is positioning itself to capture value in this new, more grounded market. The move puts pressure on competitors like Baidu and Alibaba to prove their own AI ventures can become sustainable revenue drivers, not just costly experiments.
This article is for informational purposes only and does not constitute investment advice.
