Chart Preview | Earnings Season Begins for Chinese Tech Stocks! Tencent and JD.com Report First, Facing Dual Challenges of AI and Profitability
In August, with earnings releases from major tech and internet giants such as Tencent, JD.com, Alibaba, and Meituan, market focus on China’s internet leaders has shifted from whether they would continue facing headwinds to whether they could further validate a recovery in profitability. Institutions expect further divergence in performance among large internet platforms, with AI revenue growth, profit margins, and capital expenditures emerging as key focal points.
Major tech firms are 'concentrating their forces' to capture entry points, suddenly crowding the AI-powered office productivity space.
① Major tech firms are concentrating resources to secure entry points, marking the end of fragmented, trial-and-error ‘horse-race’ approaches in AI-powered office solutions; ② The competitive landscape extends beyond major players, with ‘real-world business scenario delivery capability’ emerging as a critical differentiator; ③ AI-driven office applications are accelerating demand across the supply chain, directly benefiting upstream providers of inference chips, servers, and high-speed storage—though compute demand per office task remains limited, resulting in steady, gradual growth rather than an immediate surge.
Volcengine has launched the Seedance 2.5 API service, supporting direct generation of 30-second videos.
Volcengine, a subsidiary of ByteDance, announced the official launch of its Seedance 2.5 API service. Compared to Seedance 2.0, Seedance 2.5 delivers significant improvements in instruction following, long-form narrative generation, human-like realism, and audiovisual quality. The model natively supports direct generation of 30-second videos and can reference up to 50 full-modality assets, offering more precise and stable editing capabilities while supporting over ten languages. The output quality of Seedance 2.5 brings these foundational video elements close to cinematic-grade standards for long-form storytelling, unlocking broader commercial applications such as creative advertising and industrial simulation.
ByteDance is reportedly planning to train a large AI model with over 5 trillion parameters.
Chinese media, citing sources, reported that ByteDance is currently discussing the development of a model with over 5 trillion parameters—surpassing Alibaba’s Qwen 3.8-Max (2.4 trillion parameters) and Moonshot AI’s K3 (2.8 trillion parameters)—making it the largest-parameter model known to date in China. The plan remains in its early stages, and the model may not ultimately be released. It is understood that the new model will be led by Xiang Liang, head of Seed Foundation, in collaboration with Shen Ke, who oversees pre-training data for large language models. To this end, Seed is currently reorganizing its structure, clarifying responsibilities, and allocating resources. Seed team
Tencent and Alibaba's 'Yi Zhongtian' Bet on NPO Commercialization Next Year as Advanced Optical Interconnects Enter a Multi-Path Race
① The industrialization timeline for NPO is becoming increasingly clear, with commercial deployment expected by 2027; ② While the evolutionary path from LPO to NPO and then to CPO appears straightforward, the industry’s choice is not simply a matter of 'new technology replacing old technology'—the technology curve may be steeper than anticipated; ③ Currently, next-generation technology pathways have yet to converge, and advanced interconnects have entered a period of technological 'tug-of-war'.
Market Chatter: Alibaba, Baidu-Backed Vast Seeks New Capital at Nearly $2 Billion Valuation