姚顺雨 坐在马化腾旁边

Overall 1: Tencent has put AI at the center of its strategy and is using real investment and product data to show how AI is being deployed.

1. Yao Shunyu, born in 1998, went from Tsinghua's Yao Class and Princeton to OpenAI; within less than a year at Tencent, he took charge of foundation models, AI Infra, and AI Data, covering the full Hunyuan model chain. At 28, he appeared on Fortune's business elite list. This is a typical case for understanding the AI talent war among big tech companies.

2. In Q2 2026, Tencent's R&D spending reached RMB 27.28 billion, up 35% year on year; capital expenditure reached RMB 52.78 billion, up 176%, mainly directed at computing power. This shows AI infrastructure is still receiving intensive investment, and related jobs and applications will continue to grow.

Overall 2: The practical value for ordinary users comes from AI products that can be used directly.

1. Yuanbao's monthly active users surpassed 100 million, while WorkBuddy and CodeBuddy reached tens of millions of users. AI office tools, coding, and Q&A have entered the mainstream adoption stage.

2. Hunyuan Hy3 saw weekly call volume grow more than 68 times versus the previous generation within one week of launch; Hy4 preview was open-sourced and supports a 1M context. AI processing power for long-text and complex tasks has improved significantly, helping ordinary users save time in document analysis, coding assistance, and information organization.

Overall 1: What brands should watch is not a particular AI company but the way users are shifting to AI-driven workflows.

1. Tencent's Yuanbao surpassed 100 million MAU, and WorkBuddy and CodeBuddy reached tens of millions of users. This shows a large group of users has already integrated AI tools into their workflows. Brands need to think about how they can be recommended, cited, and reviewed inside these AI entry points.

2. Tencent has chosen to place AI in high-value scenarios and use business feedback to drive model iteration. This means AI products will become increasingly business-aware. Brands can apply the same closed-loop logic to AI transformation in customer service, content, and shopping guidance.

Overall 2: There are clear signals in product development and ecosystem selection.

1. Hunyuan Hy3's weekly call volume grew more than 68 times within its first week; Hy4 preview was open-sourced, with 770B total parameters, 49B activated parameters, and context length exceeding 1M. When choosing an AI foundation, brands should pay attention to iteration speed, openness, and long-context capability, as these directly affect content generation quality and user interaction.

2. Guangdong has proposed strengthening both general and vertical large models and expanding application promotion, creating policy and infrastructure opportunities for brands building vertical AI applications.

Overall 1: AI is being pushed up by both policy and big tech, and sellers can find incremental market opportunities from it.

1. Guangdong's core AI industry scale exceeded RMB 300 billion in 2025, up more than 40% year on year, accounting for about one-fourth of the national total. A total of 122 large models passed national filing, and the province ranks first nationally with 147 AI-specialized 'little giant' firms. This indicates rapid growth in regional demand for AI-related products and services.

2. The Guangdong provincial party secretary asked Tencent to deeply participate in the province's AI infrastructure construction, increase computing power deployment, strengthen general and vertical large models, and intensify application promotion. Sellers can watch for business opportunities in training, agency services, content services, and model customization brought by AI application promotion.

Overall 2: Tencent's AI approach gives sellers a reusable business playbook and risk reminders.

1. Tencent did not blindly diversify into other fields. Instead, it deployed AI in high-value scenarios such as Yuanbao, WorkBuddy, and CodeBuddy, using business feedback to drive model iteration. Sellers can also embed AI tools into their own main business scenarios and use real user feedback to optimize product selection, customer service, and marketing.

2. A key risk to note: AI investment is extremely heavy. Tencent's Q2 R&D spending was RMB 27.28 billion and capital expenditure was RMB 52.78 billion, mainly due to computing power procurement. Small and medium sellers should not build models or own computing infrastructure with heavy assets. They should prioritize mature AI tools and platform capabilities to control cash flow risk.

Overall 1: Full-industry-chain intelligentization is an upgrade direction factories can seize, and the article provides policy, technology, and data signals.

1. Guangdong has made AI infrastructure construction, computing power layout, and application of general and vertical large models key priorities. This means manufacturing factories can more easily find policy support in intelligent scheduling, quality inspection, and supply chain coordination.

2. Tencent's Hunyuan Hy4 preview was open-sourced, with 770B total parameters, 49B activated parameters, and a 1M context length. Factories can explore private deployment and industry-specific fine-tuning based on open-source models, avoiding handing all core production data to external platforms.

Overall 2: Tencent's R&D spending and product path offer lessons for factory digitalization.

1. Tencent's Q2 2026 R&D spending was RMB 27.28 billion, and capital expenditure was RMB 52.78 billion, mainly invested in computing power. As platform-level computing infrastructure expands, the cost for factories to use AI applications may decline.

2. Tencent uses business feedback to drive model iteration rather than building a model first and then finding a use case. Factory digitalization should similarly start with high-frequency pain points such as equipment maintenance and process optimization, letting AI iterate quickly on real production data while avoiding showcase-oriented projects.

Overall 1: Industry trends and client pain points are concentrated, and the opportunity for service providers is to help enterprises embed AI into a business closed loop.

1. Ma Huateng admitted that Tencent's early AI capabilities were not outstanding, saying it felt like 'getting on the boat and finding it was leaking.' Many clients face a similar situation. Service providers should first diagnose the client's real scenarios and then design AI integration plans, rather than directly selling computing power or models.

2. Tencent placed AI in Yuanbao, WorkBuddy, and CodeBuddy and used business feedback to drive model iteration. This model can be directly replicated in service provider projects: use client business data to continuously optimize model effectiveness and create quantifiable value.

Overall 2: Technology iteration and market space determine the direction of service providers.

1. Hunyuan Hy3's weekly call volume grew more than 68 times within one week of launch; Hy4 preview was open-sourced with context length exceeding 1M. Long-context capability makes enterprise needs such as contract analysis, industry research, and complex knowledge bases more solvable. Service providers should prioritize mastering these capabilities.

2. Guangdong's core AI industry scale has exceeded RMB 300 billion, and there is an explicit requirement to strengthen application of general and vertical large models. Service providers can focus on computing services, vertical models, AI application integration, and training to enter regional markets.

Overall 1: Tencent's latest AI layout is a complete platform strategy sample.

1. Platform approach: Tencent has not crossed into unrelated territory. Instead, it embeds AI into high-value scenarios such as Yuanbao, WorkBuddy, and CodeBuddy, and uses business feedback to drive model iteration. The key for platform companies is to integrate AI deeply into existing advantageous businesses.

2. Organizational support: Yao Shunyu leads the foundation model department, AI Infra department, and AI Data department simultaneously, covering the full Hunyuan chain. To advance AI quickly, platforms should give technology leaders cross-department resources and clear authorization, avoiding fragmented management.

Overall 2: Ecosystem building, developer recruitment, and risk control are key concerns for platform companies.

1. Hunyuan Hy4 preview was open-sourced; Yuanbao's MAU surpassed 100 million; and WorkBuddy and CodeBuddy reached tens of millions of users. Platform companies can use open-source models and high-traffic AI products to recruit developers and service providers, build an application ecosystem, and provide supporting data, computing power, and distribution.

2. Platform companies must control risk: Tencent's Q2 2026 R&D spending was RMB 27.28 billion and capital expenditure was RMB 52.78 billion, mainly due to computing power procurement, and Tencent explicitly said it will no longer use single-quarter profit to constrain AI innovation. This high-investment strategy places great pressure on cash flow. Platform companies should set clear boundaries between computing investment and commercialization returns.

Overall 1: This article provides a first-hand case for studying platform-based AI industry transformation and generational talent replacement.

1. New industry trend: Tencent moved from acknowledging weak AI capabilities to heavily investing in AI infrastructure, using high-value business scenarios to drive model iteration. This forms a research sample of the platform-based AI business model and contrasts with the cross-sector land-grab approach.

2. Talent and organization: Yao Shunyu went from Tsinghua's Yao Class and Princeton to OpenAI, and at 27 became the No. 1 person in charge of Tencent AI, leading three departments: foundation models, AI Infra, and AI Data. Researchers can study how big tech companies break the seniority ceiling to compete for young AI talent, and how this authorization model affects organizational efficiency.

Overall 2: Policy and regional industry data provide material for further research.

1. Guangdong's core AI industry scale exceeded RMB 300 billion in 2025, up more than 40% year on year, accounting for about one-fourth of the national total; 122 large models passed national filing and 147 AI-specialized 'little giant' firms made Guangdong the national leader. These figures can be used to analyze regional AI industry clustering and policy effectiveness.

2. The Guangdong provincial party secretary asked Tencent to participate in AI infrastructure construction, expand computing power layout, strengthen general and vertical large models, and intensify application promotion. This provides concrete policy direction for studying how local governments and enterprises collaborate across the computing, model, and application value chain.

3. A notable issue: Tencent's Q2 2026 capital expenditure reached RMB 52.78 billion, up 176% year on year, mainly directed at computing power, and the company no longer uses single-quarter profit to constrain AI innovation. How to balance massive AI investment with long-term commercialization returns is an important industrial economics topic.