AI Contrast Among Big Tech Firms: Robin Li Preaches, Pony Ma Sounds the Alarm
In November 2026, two of the most influential founders in China's internet industry laid out their respective judgments on AI on the table at the same time.
One of them is Robin Li. At the Baidu World Conference on November 12, he reaffirmed the judgment he made ten years ago — "The next step is artificial intelligence", and defined the next phase of AI as "Intelligent Organization".
Robin Li is used to standing at the front stage, and he is the earliest and longest-serving AI evangelist among founders of large tech companies: he talks about trends, technological revolutions, and full-stack investment, with his consistent narrative never breaking off.
The other one is Pony Ma. The general office's LinkTime event for Tencent's 28th anniversary was made into an internal podcast, and the meeting minutes that spread out later used a severe tone: "In the great age of navigation, you either progress or fall behind; in the industrial revolution, you will get beaten if you lag behind", "Half of the team is responsible for holding the ground, and the other half will go to fight new battles".
One keeps evangelizing on the stage, the other repeatedly issues warnings internally. Putting the two of them in contrast just forms a mirrored observation of the AI routes of China's large tech companies: Baidu, which entered the AI track the earliest, faces the challenge of "getting up early but still arriving late" — it has a far-sighted vision but has not yet consolidated its position steadily; Tencent, which has always been prudent and restrained, entered the AI race half a beat later, so what does it rely on to catch up from behind?
I
Robin Li's Evangelism: The Prophet Did Not Reach the Destination First
Robin Li is an evangelist, and Baidu AI is a pioneer. This judgment is by no means undeserved praise.
As early as January 2013, Baidu announced the establishment of the Institute of Deep Learning (IDL), which is most likely the world's first corporate research institute named after deep learning in the industrial sector.
Four years later, Robin Li publicly wrote "All in AI" into the company's strategy, leaving that widely quoted judgment: The Internet is only an "appetizer", and artificial intelligence is the real "main course".
As a member of the National Committee of the Chinese People's Political Consultative Conference, Robin Li has submitted AI proposals at the Two Sessions for consecutive years, ranging from the "China Brain" project to autonomous driving, and Baidu has positioned itself as the "national team" in the artificial intelligence sector.
The prophet also took the first move: Baidu laid out all lines including models, chips, autonomous driving, and cloud. The cards of this full-stack layout are still on the table today.
Model layer: The first version of the ERNIE large model was released back in 2019; ERNIE Bot, launched by Baidu in March 2023, was the first generative AI product for C-end users among domestic large tech companies;
Chip layer: Kunlunxin started its development in 2011, withstood nearly a decade of investment that was almost a financial black hole, and has now entered the harvest period. It has submitted a listing application to the Hong Kong Stock Exchange, with an estimated target valuation of 50 billion US dollars according to sources;
Autonomous driving: Baidu Apollo Go has been in operation for more than ten years with a total investment of tens of billions of yuan, and has served users in 30 cities around the world;
Cloud and computing power layer: In the third quarter of 2026, Baidu's AI business revenue was about 13.2 billion yuan, accounting for 53% of the general business revenue, exceeding 50% for three consecutive quarters.
With early entry and comprehensive layout, Baidu's AI foundation has become increasingly solid; but the problem of "getting up early but arriving late" is also a lingering curse for Baidu.
The most typical case is its To C products. ERNIE Bot is the earliest released large model in China, but it has not converted its first-mover advantage into the user scale of independent applications. On the September monthly active list of AI-native apps released by QuestMobile, the ERNIE App did not even make it into the top 10. The only Baidu product that barely made the list was Baidu Buddy.
Baidu is also the earliest explorer in terms of commercialization. In November 2023, the professional version of ERNIE Bot tried to charge C-end members. But its C-end monetization capability has never been verified. In April 2025, ERNIE Bot fully switched to free access, with an interval of less than a year and a half from paid service to full free access.
The more realistic pressure is reflected in its financial report. In the third quarter of 2026, Baidu's total revenue was about 30.8 billion yuan, of which the general business revenue was 25 billion yuan, a slight year-on-year increase of 1%, while the traditional online marketing revenue fell by 22% year-on-year. The growth of AI is not enough to offset the decline of traditional advertising, and Baidu has not achieved a smooth transition between its old and new business narratives.
It is difficult to realize monetization on the C-end. Domestic large tech companies have collectively pinned their monetization hopes on the B-end. At this Baidu World Conference, Robin Li shifted the focus of his narrative to "Intelligent Organization" — Baidu Buddy, Baidu Famou, and the Kuku product series were all upgraded and unveiled collectively.
Baidu Buddy, the only Baidu product that made it into the top 10 of AI-native apps mentioned above, is highly anticipated. It has good user reputation and strong growth momentum, targeting the currently booming AI office track.
However, Baidu Buddy faces many challenges. The appetizer of AI office is To C, but the main course lies in To B. The model capability determines whether the AI is smart enough, and the context determines whether it understands a specific company; only with complete context can the AI become a digital colleague that is familiar with business, aligned with goals, and capable of following up tasks independently.
Alibaba's Qwen Office has DingTalk, Tencent's WorkBuddy has WeCom, and Doubao Work has Feishu, while Baidu has relatively limited accumulation in the office track in the past.
II
Pony Ma's Warnings: From Calm to Radical
The public's long-term perception of Tencent's technology narrative is that it is conservative and restrained, and it always regards new technologies as efficiency improvement tools for its existing ecosystem first, rather than rushing to overthrow and rebuild everything.
This attitude curve reversed dramatically in 2026.
At Tencent's 2025 annual staff meeting held in January 2026, Pony Ma still seemed very calm: "Every enterprise has different genes and different conditions, and Tencent's style is to advance steadily." At that time, the strategic role of AI seemed to be mostly improving efficiency and strengthening Tencent's existing stock business.
At Tencent's general meeting of shareholders in May 2026, Pony Ma changed his usual calm attitude, reflected that Tencent's early basic capabilities in the AI field were not outstanding, and publicly admitted that Tencent was making up lessons: "A year ago we thought we had boarded the ship, but later we found that the ship was leaking. Now we feel we are standing on it, but we still can't sit down, and we still hope the ship can go faster."
At the anniversary meeting last week, Tencent's attitude became more radical, and it announced that it had entered a "wartime state" with metaphors: "Land warfare turns into naval warfare, the army turns into the navy, everyone needs to carry supplies, update skills, and get on board to fight quickly." "If we don't invest and embrace the ongoing intelligent revolution, we will be subverted. This is a matter of life and death."
Tencent's strategic positioning of AI has also reversed, emphasizing the subversive role of AI: "Some core businesses seem to be in a stable position, but AI is developing so fast that the waves in the distance are swallowing hills one by one, you can't just sit there and appreciate it."
Once the cognition reverses, the actions will also change accordingly.
In the second quarter of 2026, Tencent's capital expenditure in a single quarter was about 52.78 billion yuan, a year-on-year increase of 176%, and its free cash flow turned negative to 13.8 billion yuan. This is the first time since 2005 that Tencent has recorded negative free cash flow in a single quarter. The senior management has made it clear that AI-related investment in 2026 will at least double.
Tencent is catching up in the AI capability layer, and its model has gradually approached the domestic first tier; WorkBuddy, as the first AI-native product of Tencent that has achieved outstanding performance, ranks first among office Agent products; Xiaowei, leveraging the stock advantage of WeChat, aims to become the public's personal AI assistant.
The organization is also undergoing major reshuffling. Tencent's general office calls on all employees to reset their mindset and throw away burdens, flatten the organization again, and take small teams as combat units; there will not necessarily be so many management positions in the future, and sectarianism and territorial awareness must be completely broken; half of the staff in core businesses are responsible for holding the ground, and the other half will fight new battles.
It must be admitted that Tencent's AI expansion has not been all smooth sailing. The position of its base model is still unstable; facing the suppression of ByteDance's Doubao in terms of user scale, Yuanbao is under pressure in terms of user growth; from the perspective of product layout, Tencent's AI focus tilts towards B-end scenarios and internal business empowerment, and whether this pragmatic shift can work still needs to be observed.
In fact, Pony Ma and Robin Li have relatively consistent cognition of the final outcome of AI. But why does Pony Ma focus on issuing warnings while Robin Li tends to evangelize?
Baidu took the first move in AI but failed to reach the destination first, with its performance not being outstanding enough. Robin Li needs to boost morale, emphasize Baidu's status as a prophet and first mover, and outline Baidu's future prospects.
In this regard, Robin Li performed very frankly. He took the initiative to respond to the public's questioning of Baidu's "getting up early but arriving late", saying that he psychologically accepts that 90% of attempts may fail. If the attempt does not succeed, the so-called "arriving late" mentioned by the outside world is the price that innovation needs to pay.
Tencent is still financially strong, with its revenue in Q2 2026 increasing by 11% year-on-year; securities firms predict that its Q3 revenue will increase by about 9%. Such steady growth can easily lead to slackness among internal employees, so Tencent needs to sound the alarm even more than Baidu.
III
All at the Poker Table, Moving in the Same Direction but Taking Different Paths
In this round of AI transformation, large tech companies are generally in a state of anxiety.
When the internet shifted to the mobile internet, most resources were transferable, and large tech companies still held their positions at the front stage. In the mobile era, the only new large tech company that emerged in China is ByteDance.
But in the AI track, "unorthodox moves beat seasoned veterans". The traffic entrances, user duration, and business moats accumulated over the past ten years are no longer unbreakable barriers in the face of the capability reconstruction brought by large models.
The A company founded by Dario Amodei, who once worked at Baidu, has a valuation of over one trillion US dollars, dozens of times Baidu's market value.
Domestic companies including Zhipu AI, Moonshot AI, DeepSeek, and overseas companies including OpenAI and Anthropic are all AI-native companies that have risen rapidly from 0 to 1.
Of course, the AI battle is far from reaching its final outcome, and the killer product that truly defines the AI era has not yet emerged. Whether it is Baidu AI which took the first move but failed to reach the destination first, or Tencent AI which is catching up from behind, they are all still at the poker table.
However, from the development paths of these two companies, we can also summarize several rules.
First, the first mover does not necessarily reach the destination first, and a correct direction does not mean you can reach the finish line.
Robin Li is one of the earliest people who saw clearly the AI paradigm. Baidu laid out its full-stack technology more than ten years in advance, but the first-mover advantage did not help it reach the destination first.
Pony Ma saw the trend later and made his move later. What really determines whether the latecomer can catch up is not only the model benchmark scores: WorkBuddy wins not because of its model capability, but because of its product capability.
Whether it is Tencent or Baidu, their current AI business has only won partial victories, and it is far from the time to open champagne for celebration.
Second, the hematopoietic capability is the confidence to make up for missed lessons.
The reason why Tencent dares to double its AI investment in 2026 largely relies on the revenue generated by its main business: its revenue in the second quarter of 2026 was about 204.8 billion yuan, a year-on-year increase of 11%, which is one of the few large tech companies that achieved double growth in both revenue and profit in the same period. Baidu is in the most difficult stage of switching between old and new growth drivers.
Third, entrances and ecosystem are the hardest cards in the hands of latecomers.
The most promising point for Tencent is that it holds China's most frequent social and content entrances, so its models and agents can penetrate into the daily life of users and enterprises along existing scenarios such as WeChat, WeCom, and Tencent Meeting.
But whether this card can be played out is exactly stuck at Tencent's consistent cautious point: whether it is willing to shake the foundation of its super App for AI.
Tencent's general office's emphasis on "resetting mindset", "flattening organization", "breaking sectarianism and territorial awareness", and "half of the staff fighting new battles" this time is essentially clearing obstacles for activating these entrances.
Fourth, the AI war has not reached its final outcome, and as long as you are at the poker table, you still have a chance.
Baidu's Kunlunxin and Apollo Go are both technical barriers that need to be verified on a ten-year scale. In the chip sector, Baidu has already achieved great results.
According to Reuters, Kunlunxin, which is sprinting for IPO, is seeking a target valuation of about 50 billion US dollars for its Hong Kong stock IPO, far exceeding Baidu's current market value, which is also the harvest of Baidu's long-term R&D investment.
Baidu's first move lies in its early cognition, calling out the direction when everyone is still skeptical; Tencent's hidden card lies in its strong capital strength, and the necessity of issuing warnings is to wake up the frogs in warm water.
The new technology wave will not reward those who raise their hands first, but only those who sail the ship to the opposite shore. For Baidu, the problem to solve is how to bridge the gap between being a prophet and reaching the destination, and improve the success rate of its products; for Tencent, the challenge is whether its determination to catch up from behind can break through inertial thinking.
Fortunately, the poker game of the intelligent revolution has only reached the middle stage, and Pony Ma and Robin Li are both willing to rush to the front line. Both of them have firm belief in AI, but they just choose different postures to reach the destination.
As long as the game is not over, there is still a chance to win.
This article is from the WeChat Official Account "Finance Story Collection" (ID: cjgshui), written by Wan Tiannan, edited by Chen Jiying, and published with authorization from 36Kr.