Chinese Consumer Tech Giants Build Own AI Systems

Chinese technology companies are increasingly developing their own artificial intelligence systems for everyday users. This trend extends beyond dedicated AI research laboratories, encompassing established consumer platforms. Organisations in e-commerce, gaming, social media, and travel are creating bespoke foundation models for their operational ecosystems.

Credit: UNSPLASH

This quieter wave of AI development contrasts with more publicised efforts of China's major AI firms. Digital platforms now integrate custom AI directly into business operations, rather than relying on external providers. The aim is to serve a broad user base with tailored applications and functionalities.

And, RedNote, a lifestyle platform often compared to Instagram, recently revealed its latest advancement. Its AI research division, Dots Studio, launched Dots3-Note Preview, an open-weight model with 280 billion parameters. Benchmark tests by RedNote indicated this system performed on par with or exceeded models from US-based OpenAI and Anthropic, alongside domestic leaders DeepSeek and Z.ai, across specific tasks.

This release marks a technical shift for a platform primarily recognised for fashion, travel, and shopping recommendations. Dots Studio stated its objective was to build AI systems that benefit individuals and assist in resolving common daily problems. The operational change across China's technology sector reflects this internalisation of AI development.

But, industry analysts view this internalisation as a response to AI becoming central for non-AI companies. Omdia chief analyst Su Lian Jye noted these consumer platforms are "digital native and extremely data-rich." Combining proprietary business data with custom AI models can streamline processes, forecast demand, personalise services, and generate new revenue.

Similar development approaches are evident throughout the sector. Meituan, a food delivery and local services company, invested substantially in its LongCat model family. The company seeks to incorporate AI across merchant services, logistics, customer support, and recommendation algorithms.

So, Trip.com, China's largest online travel agency, also developed its own AI system. Named Wendao, this travel-focused AI supports functions such as itinerary planning, customer service interactions, and travel suggestions. These in-house developments permit greater control and specialisation in specific service areas.

Chinese video streaming platform Bilibili and gaming developer miHoYo have similarly concentrated efforts on utilising AI. Their focus is on enhancing content creation capabilities and developing more interactive digital experiences. This strategy aims to deepen user engagement within their entertainment ecosystems.

And, Bilibili created the Index family of language models, including IndexTTS, a multilingual speech synthesis system. This system has found application in AI subtitles, translation, video creation, and voice generation. miHoYo developed Glossa, an in-house generative AI model designed to make game characters more interactive. It is reportedly used in AI features for *Honkai: Star Rail*, its globally popular role-playing game.

The gaming company also explores additional AI applications for game development, including content generation, voice synthesis, and production workflows. According to the SCMP report, this extensive internal development of AI infrastructure poses significant operational challenges.

But, building and expanding such systems comes with steep hurdles. Omdia's AI Market Maturity Survey identified several significant barriers. These included insufficient or poor-quality business data, regulatory compliance, a scarcity of qualified in-house talent, and integration challenges with existing systems.

Investing in AI can introduce "budget pressure" from high infrastructure and personnel costs. However, Omdia's Su suggested "the long-term impact can be strongly positive" for these organisations.

• Chinese consumer platforms are increasingly developing in-house artificial intelligence capabilities.

• Organisations like RedNote, Meituan, Trip.com, Bilibili, and miHoYo are building tailored foundation models for specific business needs.

• These bespoke AI systems aim to streamline operations, personalise services, and enhance user experiences.

• Operational hurdles include data quality, regulatory issues, talent shortages, and system integration.

• Despite high initial costs, the long-term benefits of AI investment are projected to be substantial.

Source: SCMP