02/21 2025
473
Tencent CEO Pony Ma once remarked:
"In the grand scheme of an industrial revolution, bringing out the light bulb a month earlier isn't as crucial. What matters is solidifying the underlying algorithms, computational power, and data."
From the rise of ChatGPT to DeepSeek empowering China's internet landscape, Tencent has finally decided to act after a lengthy period of observation.
Tencent is now fully committing all its products and teams to AI.
Why the emphasis on AI at the dawn of 2025? Tencent is dedicating all its products and team efforts to AI.
Following the transfer of Tencent's Yuanbao from the TEG (Technology and Engineering Group) to CSIG (Tencent Cloud and Industrial Internet Group), products and applications such as Sogou Input Method, QQ Browser, and ima have also officially moved to CSIG, collectively forming a potent force driving Tencent into the era of large models.
From a product perspective, Sogou Input Method and QQ Browser boast core advantages in the AI era: their long history and vast user base. Their application scenarios seamlessly integrate large model applications like AI writing and AI search. Their mature user base and content accumulation also help Tencent reduce costs on promotion and user habit cultivation.
As an AI-native application incubated within Tencent in October 2024, ima's positioning as an "intelligent workbench," coupled with its interface advantages of direct integration with official accounts and WeChat, offers users a novel experience of "integrated search, reading, and writing." This allows Tencent to expand the reach of its AI products while enhancing existing products with AI technology, striking a balance between "old products" and "new paths."
Whether it's the swift adaptation of products to AI concepts or the agile adjustment of team organizational structures, these actions underscore Tencent's determination to establish an AI ecosystem.
Since the Spring Festival, news of Tencent products like Yuanbao, Tencent Docs, Tencent Maps, and QQ Music adapting to AI has been relentless. This signifies that after DeepSeek sparked a technological upgrade in China's internet sector, Tencent has ended its "observation phase" in the AI era and pinpointed different focal points for each product.
In the near future, Tencent is likely to further bolster its efforts in organizational structure and resource support to fully embrace the era of large models for its brand and products.
The key to surpassing competitors is for established entities to foster new growth
Pony Ma once stated: "For an industrial revolution, bringing out the light bulb a month earlier isn't critical in the long run. The essence lies in solidly developing the underlying algorithms, computational power, and data."
This implies that, compared to the speed of launching AI-related products, Tencent prioritizes user experience and cultivating and maintaining users' AI usage habits in current and future application scenarios. For Tencent, while exploring new avenues is important, it cannot forsake the idea of empowering old products with new technology to achieve "old trees sprouting new shoots."
Take the national-level app WeChat as an example.
As a chat app, WeChat has faced challengers.
Concepts like instant messaging and voice message interaction can only facilitate its transformation from 0 to 1, helping it establish a foothold in the "chat app" category. To evolve from an "app" to a "national-level app," it's essential to flexibly apply the "old trees sprouting new shoots" strategy across different internet product cycles, maintaining user stickiness and reinforcing its product position. This was evident with the advent of mini-programs, mini-games, video numbers, and live streaming accounts.
Today, after experiencing multiple impacts of the "AI empowerment" concept, and after a prolonged period of observation, WeChat has finally decided to integrate DeepSeek.
Related reports indicate that after adding the "AI search" function, WeChat search will incorporate content from the entire official account module into the AI search content database. This not only aids users in solving problems more effectively but also allows the popularity of official account posts to transcend the communication cycle, reinforcing their content-oriented development model and further invigorating creators' enthusiasm.
Ultimately, relying on flexible and precise AI recommendation and answer mechanisms, the platform will achieve a dual upgrade in "user and creator" stickiness, crafting a positive cycle for the platform.
It's worth noting that as the current national-level app, every decision and update made by WeChat draws significant attention to Tencent itself. This high-profile move by WeChat, besides promoting new features, also breaks the misconception that Tencent is slow in the realm of large models by placing a "double bet" on technology and user experience, positioning itself as a "calm boxer" with a late-mover advantage.
Who will discover the next killer app first?
As the term "Tencent, Huawei, etc., lose over 400 million yuan per month after integrating DeepSeek" trended on social media, people have become somewhat perplexed about large companies' bets on AI.
Relevant content reveals that though domestic large companies have integrated DeepSeek, they primarily adopt free or low-cost operating models. (The full-fledged version of DeepSeek-R1 charges only 16 yuan per million tokens)
If a company outputs 100 billion tokens daily, its monthly machine cost based on DeepSeek services will reach 450 million yuan, resulting in a loss of 400 million yuan; even with AMD chips yielding a monthly income of 45 million yuan and reducing the monthly machine cost to 270 million yuan, there will still be a loss of over 200 million yuan.
Put simply, current large companies are caught in a commercial vicious circle of "the more users, the more losses."
Given the monthly losses, why haven't any large companies halted their AI layout plans?
Because everyone is vying to "find the killer app."
While AI chatbots and AI search functions represent significant interpretations of AI to consumers, their essence lies in different ways for companies to explore AI potential and empower their products with AI. To date, there has been no genuinely "killer AI application" on the market.
For large companies laying out the AI track, their core task is to solidify their AI transformation pace, as market and technological trends all suggest that AI holds the key to the next business iteration. After initiating the transformation plan, the primary objective for large companies becomes discovering and finding the content that users truly desire and can represent the application 2.0 experience.
Just as WeChat is betting on AI search, everyone is using their unique approach to seize the technological advantages of the AI era and explore what users desire in this era.
Once a company achieves these two points ahead of others, the future balance will undoubtedly tilt in their favor. Therefore, compared to dominating the discourse power of future tracks, the current loss figures cannot impede the progress of large companies.
References:
Science and Technology Innovation Board Daily: Four major product lines converge, Tencent is rapidly building an AI application matrix
Securities Times: "e-company observation" Super apps successively partnering with DeepSeek, "technological community-style" interconnection or interconnection may become a new trend
TMTPost APP: Internet giants under the impact of DeepSeek: Who is rising, who is falling behind