Looking back at the end of 2025, China's AI battlefield has evolved from a "parameter war" in 2024 to a strategic competition centered on capital efficiency, infrastructure dominance, and traffic gateways. On November 27, Goldman Sachs released an in-depth research report on China's internet and AI development, analyzing the fierce competition among major players—Alibaba, ByteDance, and Tencent—and their distinct strategic choices.
**Alibaba's "Full-Stack" Gamble** Alibaba has opted for the most capital-intensive path, with an 80% YoY surge in capital expenditure (CAPEX) to RMB 32 billion in Q3, aiming to build a Google-like "full-stack" dominance in China's AI market. Despite chip supply volatility, Alibaba remains committed to expanding its AI infrastructure, integrating foundational models with multimodal capabilities to strengthen its B2B market leadership. This strategy is paying off: Alibaba Cloud’s external revenue grew 29% YoY in Q3, with AI-related revenue posting triple-digit growth for the ninth consecutive quarter. Goldman Sachs forecasts further acceleration to 38% growth in Q4. On the consumer side, Alibaba’s newly launched "Tongyi App" surpassed 10 million downloads in its first week, while Ant Group’s "Lingguang" hit 2 million downloads in six days.
**ByteDance’s "Traffic Breakthrough"** ByteDance leverages its massive traffic advantage, consuming 30 trillion tokens daily—nearing Google’s 43 trillion and far outpacing domestic peers like Baidu and DeepSeek (around 10 trillion). Its AI app "Doubao" leads domestic user engagement, while overseas education app Gauth saw a 394% YoY revenue surge. This demand has propelled ByteDance’s MaaS (Model-as-a-Service) strategy, with its Volcano Engine capturing 49.2% of China’s public cloud market for large models, according to IDC.
**Tencent’s "Ecosystem Integration"** Tencent maintains its pragmatic approach, reducing CAPEX while embedding AI seamlessly into its WeChat ecosystem. Its AI assistant "Yuanbao" now integrates with WeChat Pay, enhancing efficiency for SMEs. Though less flashy, this strategy offers high implementation certainty.
**U.S.-China AI Race: A 3-6 Month Catch-Up Cycle** Goldman Sachs notes a resilient "dynamic追赶" pattern: after U.S. models (e.g., Google’s Gemini 3 Pro) advance, Chinese counterparts typically close the gap within 3-6 months. Chinese firms also excel in cost control—e.g., Kuaishou’s "Kling" video model undercuts global rivals—and benefit from a robust open-source ecosystem, with 80% of local AI startups using shared models.
**Valuation Outlook** Goldman Sachs sees no AI bubble in China. Tencent and Alibaba trade at 21x and 23x 2026E P/E, respectively, below Google (24x), Amazon, and Microsoft (28-30x).
Comments