Wer wird die Ära der KI-Spiele regieren? Tencent, NetEase und Mihoyo kämpfen heimlich um die Technologie.

The "production capacity ceiling" of the gaming industry is being shattered by AI technology.

Tencent, NetEase, miHoYo... All game manufacturers are betting on AI. Tencent Hunyuan has launched the Hunyuan Game Visual Generation Platform, which can generate high - precision game characters within minutes. NetEase's Egg Party has achieved the popularization of UGC creation with the help of AI. The new work of miHoYo has reconstructed the narrative logic with real - time multi - modal dialogue.

Whether it is Tencent's "Jili" serving as players' AI teammate, or the martial arts world constructed by 400 intelligent NPCs of NetEase, the industry giants are using technology to prove that AI can not only boost the efficiency of art and development hundreds of times, but also enable games to leap from "preset scripts" to a "dynamic world" — an era where exclusive storylines and companions are generated for each player in real - time has begun.

01

One Name, Two Directions

At the 2025 Global Game Developers Conference (GDC) held in San Francisco, USA in March this year, the trend of AI empowering games was already very obvious. Among the 16 theme summits held during the conference, almost all the forum sharing sessions by game companies and tool providers were related to AI.

Meanwhile, game manufacturers are accelerating their AI layout. For example, 37Games invested in more than 5 AI technology companies as early as 2024. Manufacturers such as Tencent, NetEase, Giant Network, and Kingnet Network have built all - around basic capabilities centered on self - developed large models.

According to the estimation of market.us, the global market size of generative AI in games was $1.43 billion in 2024 and is expected to reach $11.11 billion in 2033, with a compound annual growth rate of 25.6%. Undoubtedly, AI - powered games have become a must - have for game manufacturers.

Although they are all called AI - powered games, there are actually two completely different application directions.

One is on the production side. With the empowerment of AI, more complex and innovative game content can be developed in a shorter cycle and at a lower cost. As a productivity tool, it continuously promotes cost reduction and efficiency improvement, which can be roughly called "games made with AI".

Public information shows that 37Games integrated Shaolin martial arts into the game Puzzles & Survival using AI motion capture technology, shortening the development cycle by 40%. At the same time, 70% of user problems are handled through the intelligent customer service system, reducing labor costs by 60%.

NetEase CEO Ding Lei even said directly at the earnings conference that technologies such as DeepSeek have increased game R & D efficiency by 30%, especially automating processes in NPC dialogue design, level testing and other aspects. Taking Back to Blood as an example, the time for AI to generate mission scripts has been shortened from 2 weeks to 3 days, and labor costs have been reduced by 40%.

The Hunyuan Game Visual Generation Platform launched by Tencent Hunyuan has four core capabilities: AI art pipeline, real - time canvas, AI 2D art generation, and professional character multi - view generation, all of which serve to improve the efficiency of game production.

Moreover, Tencent Games has launched a no - code programming game editor called "Light Game Dream Workshop". Developers can easily edit the logic of levels, items, skills, characters, etc. without writing code, further lowering the development threshold for game production.

The other is on the experience side. Through generative AI technology, more diverse game content is developed and tailored to players' personal preferences, thus providing a gaming experience that is closer to players' needs, which can be roughly called "games driven by AI".

In February this year, Peacekeeper Elite under Tencent announced that it had connected its digital spokesperson "Jili" to DeepSeek. Players can interact with Jili through the "Jishi Tong" entrance in the game lobby. In the game, Jili can also serve as players' AI teammate.

Based on technologies such as intelligent voice, animation, and NLP, NetEase has long been experimenting with the application of AIGC in popular games such as Egg Party and Back to Blood.

For example, the AIGC function is built into the UGC creation tool of Egg Party, and intelligent Abby, AI creatures, etc. that can have intelligent conversations are launched. Since 2023, there has been a new AI gameplay in Egg Party almost every two to three months.

02

Still Waiting for the Singularity

The impact of AI - powered games on the industry is obvious. Take Grand Theft Auto as an example. This globally popular game has been iterated to Grand Theft Auto V. However, it has been 12 years from Grand Theft Auto V to Grand Theft Auto VI, with numerous delays, and there is still no exact release date for the new work. The biggest reason is that the production cycle of excellent works is too long, and the final effect is difficult to control.

As a player, how many 12 - year periods can one have in a lifetime? With the addition of AI, the game production efficiency has been greatly improved. If the production time of a new version of Grand Theft Auto is shortened from more than 10 years to 2 - 3 years, and the game is evolved through high - frequency iteration, it will not only reduce the production cycle of such traditional large - scale games but also generate more topics in marketing.

During the latest earnings conference call, when answering investors' questions, Tencent also announced the company's understanding of AI - powered games: "We believe that the improvement of AI technology will directly and indirectly benefit the gaming industry."

Directly, game developers can use AI to create more content more quickly and serve more users more efficiently.

Indirectly, it may be a long - term thing over decades. As humans use AI more widely, people who are currently empowered by AI will have more time and stronger willingness to engage in highly autonomous activities.

In game R & D, excellent content can be produced in a shorter time, so that high - quality content in games can be updated more frequently. In PVE, when there are more intelligent robots, the game will become more exciting and more like PVP.

In PVP, when games apply AI, many tasks such as matching, balancing, and guiding new users can be completed in a better way. All these AI applications can really help popular large - scale games become even more popular and more attractive to users.

However, from a broader perspective, if AI is only used to reduce game development costs and improve development efficiency, its development space will soon reach a ceiling. Only when it can bring about an increase in revenue and the number of users will it be truly meaningful. Huang Yi, the secretary of the board of directors and vice - president of Century Huatong, believes that "only when AI can make games more fun and creative will games in the AI era emerge."

However, the capabilities of the current base models are not sufficient to drive a real experience revolution in AI - powered games.

Yan Kuan, an investor from Lisi Capital, mentioned in a salon that there is currently no game that combines the playability of traditional games with AI - driven features. The core problem is that the capabilities of the base models have not reached the critical point. For example, in terms of reasoning ability, whether intelligent NPCs can really achieve long - term memory.

The crux of the problem lies in the essential difference between using large models to infer the behavior logic of NPCs and using traditional game programs to regulate their behavior logic, which is a knowledge - base issue. How to store the "memories" of NPCs and complete one - link - after - another reasoning through the connection of these "memories" is a great test for large models.

For example, the AI practices of Yanyun 16 Sounds and Back to Blood have exposed an obvious "disconnection effect".

The former uses self - developed AI to build a dynamic martial arts world ecosystem, allowing NPCs to remember and respond to players' actions, thus enhancing players' immersion in the martial arts world. The latter, although it has created an intelligent NPC named Shen Qiusuo with the help of DeepSeek, has become a marketing gimmick due to the lack of rich interaction scenarios.

It is obvious that there is still a long way to go in the technology of AI - powered games. While game manufacturers are exploring, they are also waiting for the arrival of the singularity in large - model technology.

03

Conclusion

Currently, large AI models are penetrating the surface of the gaming industry. The AI competition among game manufacturers has gradually evolved from the initial technology stacking to a three - dimensional war of "technology ecologicalization", "emotional valueization", and "industrial collaboration". The year 2025 will be a watershed for AI - powered games.

Referring to the development path of mobile games, casual games may be the first to explode. RPGs and open - world games are still in the exploration stage due to their high interaction complexity. From a longer - term perspective, the ultimate winners need to have the capabilities in three aspects: an industrial base for efficient game production, emotional interaction design, and a global collaborative ecosystem.

When AI is no longer just a "cost - reducing tool" but an "experience engine", games will ultimately transcend the category of "entertainment products" and become an extension of human emotions and creativity. The competition among Tencent, NetEase, and miHoYo is actually a battle to lay the foundation for the next - generation digital civilization.

This article is from the WeChat official account "Insight New Research Club" (ID: DJXYS - 0309), author: Chen Wen. It is published by 36Kr with authorization.