The era of AI cloud computing: Tencent "making money", Alibaba "spending money", Baidu "climbing the mountain"

In the past two days, Wu Hongsheng, the former core person - in - charge of Tencent Cloud, lashed out at his former employer, Tencent Cloud, on his WeChat Moments.

Most of the problems Wu Hongsheng complained about are related to products and services. These problems are not new, and may not even be considered major issues.

These problems are not necessarily exclusive to Tencent Cloud. For example, Alibaba Cloud also chases overdue payments, but it may not suspend services for users as quickly.

However, the fact that the former core person - in - charge of Tencent Cloud publicly "fired a shot" on his WeChat Moments is a bit of an embarrassment to his former employer.

From the inside out, should Tencent Cloud change in the AI era?

From the perspective of product development and service provision, there is room for Tencent Cloud to continue to improve.

For example, can the overdue payment limit be raised, and can the way of chasing overdue payments be more user - friendly? There is indeed room for improvement in the details of these products and services. Moreover, as the demand for cloud computing services is becoming increasingly essential, adjusting the service strategy may not necessarily affect the company's profits.

But for Tencent, it may not be that they can't do these things well, but that there is no sense of urgency.

From Tencent's perspective, the poor product and service quality is understandable.

Firstly, Tencent Cloud grew out of the internal needs of the WeChat ecosystem, and the priority of serving the internal is much higher than that of serving the external.

AI is one of Tencent's important current strategies, and the cloud business department has a unique ecological position.

AI has become a must - fight territory for tech giants. Alibaba Cloud and Baidu Smart Cloud both shoulder the important mission of implementing their groups' AI strategies. Perhaps for these giants, ensuring the implementation of the group's AI needs takes precedence over investing resources in serving small and medium - sized customers.

Moreover, Tencent proposed the commercialization goal of its cloud business early on and achieved full - year large - scale profitability in 2025.

The "zero - tolerance" policy towards overdue payments makes sense financially. Especially today, when the price of computing power is rising, each additional day that overdue users occupy computing power means that Tencent Cloud's revenue has to bear more opportunity costs. Under the "profit - first" strategy, even a decline in market share is acceptable.

Secondly, Tencent Cloud has no shortage of users.

Relying on the WeChat ecosystem, Tencent Cloud currently reaches over 130 million enterprise users. Its revenue structure is also very healthy, and in the second half of 2025, the enterprise service business mainly based on cloud achieved a 22% growth rate.

Ultimately, Tencent Cloud's business growth is driven by efficiency rather than scale.

Tang Daosheng has also emphasized "business operation efficiency, cost structure, and profit level" in the past. Therefore, the development focus of Tencent Cloud's IaaS and PaaS has always been to pursue profits, and the core goal of SaaS is also commercial success.

In the cloud computing industry, only Tencent Cloud can achieve a gross profit margin of around 50%. In terms of commercialization, Tencent Cloud is the most stable.

This makes it easy to understand why Wu Hongsheng was "urgently called" for owing only 7 cents.

To put it bluntly, one of the core KPIs of Tencent Cloud in the past was to "make money".

There is nothing wrong with making money, but in today's era where AI is the main line of competition, being too obsessed with pursuing efficiency and profit may not be the best solution.

The reason is simple: Tencent Cloud may not have many advantages in the infrastructure of the AI era.

In the AI era, the market competition dimension of cloud computing has changed.

In the past, the competition was about IaaS, PaaS, and SaaS services, and product prices. But today, the competition is about AI - native and Agent - native capabilities. The competition among cloud providers is essentially a comprehensive competition of model capabilities and service systems.

This is also one of the major challenges that Tencent Cloud needs to face.

The result of pursuing profits and having no shortage of users is that Tencent Cloud's market share in the IaaS and PaaS layers has been compressed by competitors. After all, everyone's budgets are tight these days. In the AI cloud market, Tencent may not have an absolute competitive advantage over Alibaba Cloud, Baidu Smart Cloud, and even ByteDance's Volcengine.

Therefore, the Hunyuan large - model team has been accelerating the model version upgrade and constantly trying to improve AI capabilities in order to make further breakthroughs in the AI cloud market.

According to Omdia's statistics, in the 2025 AI cloud market share, Alibaba Cloud ranked first with a 38% market share, Volcengine ranked second, Baidu Smart Cloud ranked third, and Tencent Cloud ranked fourth.

In terms of AI capabilities, Tencent's Yuanbao may still lag behind Baidu's Doubao and Alibaba's Qianwen.

It should be noted that the stronger the AI capabilities, the stronger the customers' demand for Tokens, which may further drive the demand for cloud services. This may also be one of the reasons for the changing landscape of the AI cloud market.

For Tencent Cloud, since it doesn't have an advantage in AI models, should it consider focusing on services?

From this perspective, Wu Hongsheng's voice on his WeChat Moments may be worthy of Tang Daosheng's attention.

Because this may just expose Tencent Cloud's "weakness": insufficient external service capabilities.

Profitability is indeed an advantage, but the continuous decline in market share, growth relying on specific fields such as finance, and insufficient penetration in emerging fields are all problems that Tencent Cloud needs to solve.

Does Tencent lack money? No.

In my opinion, Tencent lacks the determination and courage for continuous investment in AI. With the trend of AI, Tencent Cloud should not give up growth space in favor of short - term profits.

Next, can the Hunyuan AI surpass Qianwen and Doubao in product capabilities and gain greater industry influence? For another example, can the growth brought by Tencent's AI product WorKBuddy after it enters the market drive the growth of the small and medium - sized enterprise and even the personal cloud market?

These may be questions that Tang Daosheng needs to reflect on.

To put it bluntly, the success or failure of Tencent Cloud in the future cannot always rely on the internal and cannot always bet on the WeChat ecosystem. It also needs to find growth space externally.

In the past, Tencent Cloud's growth direction was internal, while in the future, the growth direction of cloud computing must be outside the Tencent system.

Beyond WeChat, how can Tencent build another infrastructure for the AI era? This is a question that Tencent's AI business will definitely answer in the future. At that time, how should Tencent Cloud grow externally and how should it cooperate with Tencent's AI strategy implementation?

This may be the mission of Tencent Cloud in the next stage.

Alibaba Cloud has to follow the path that Tencent Cloud has taken

Alibaba Cloud and Tencent Cloud are two different extremes. At the beginning, Alibaba Cloud focused on the external market.

Tencent pays too much attention to efficiency and profit, while Alibaba attaches too much importance to scale and investment.

Alibaba Cloud is the leader in the market. Its core strategy is to consolidate its leading position with scale advantages and full - stack technical capabilities. Especially in the AI wave, it acquires an absolute market share through aggressive investment.

Therefore, for a long time in the past, profitability was not the most important thing for Alibaba Cloud. Wu Yongming even stated that Alibaba will invest more than 380 billion yuan in AI infrastructure in the next three years.

Alibaba's investment has paid off.

According to IDC's statistics on the AI public cloud, Alibaba Cloud tied for first place with Baidu Smart Cloud with a 24.6% share. According to Omdia's annual statistics on the AI cloud, Alibaba Cloud's market share exceeds the sum of the second to fourth - ranked companies. In the financial report, the proportion of Alibaba Cloud's AI - related revenue exceeded 30% for the first time, reaching 8.971 billion yuan in a single quarter.

The experience of Internet business tells us that scale advantages are often achieved through substantial investment.

In Alibaba's financial report, the total capital expenditure in fiscal year 2026 was 126.063 billion yuan, mainly used for cloud infrastructure construction and instant retail investment.

With such a large - scale investment, the profit level of the cloud business has not changed much.

According to a research report by Morgan Stanley, the EBITA profit margin of Alibaba Cloud's business is approximately 8% - 9%. After the revenue generated by the new computing power is offset by depreciation costs, the actual improvement of Alibaba Cloud's EBITA profit margin seems not ideal.

The problem is clear. Next, the path ahead for Alibaba Cloud is clear, which is to follow the "profit - first" path that Tencent Cloud's business has taken.

Wu Yongming also set the tone at the earnings conference call. He said, "Alibaba's full - stack AI technology investment has officially crossed the initial cultivation stage and entered the positive cycle of large - scale commercial returns."

How should Alibaba Smart Cloud enter the next stage of commercial returns?

The answer may lie in AI models and chips.

Let's start with AI models.

The stronger the model capabilities, the easier it may be to expand customers.

It mainly depends on the Qianwen AI and the implementation of Alibaba's full - stack AI capabilities. In fact, the faster the implementation of AI solutions, the faster the growth of users' demand for Tokens. Then, the business demand for Alibaba Smart Cloud can be stimulated.

Objectively speaking, although the Qianwen model may not be the top - notch in the industry, Alibaba's ability to serve customers is very strong.

Currently, there are many customers of Alibaba Cloud among the leading enterprises in various industries such as finance and manufacturing, and it has also expanded overseas customers such as the NBA and Marriott. Next, how to further tap the Token demand of these users and convert Alibaba Cloud's customers into customers of AI solutions is an important proposition.

Secondly, it is about chips and computing power.

For Alibaba Cloud, another key to commercialization is actually cost reduction. For example, further increasing the deployment ratio of self - developed chips and further implementing the core domestic substitution solutions for large - model training/inference.

Currently, Alibaba's T-Head has established a complete chip product system integrating the edge and the cloud. However, at present, it is mainly positioned as the main force for inference and a supplement for training, and is deployed in coordination with NVIDIA GPUs.

What does this mean?

High - end NVIDIA GPUs are still the first choice for large - scale pre - training. Next, whether the Zhenwu PPU for training and inference can be further deployed and whether the production capacity can keep up are crucial.

Currently, the performance of T-Head's Zhenwu M890 chip has been increased by 3 times, and a total of 560,000 chips have been shipped. This quantity may still be far from enough. Next, can the core low - cost computing power be commercially available on a large scale?

This question is worth pondering.

Beyond AI models and chips, another challenge for Alibaba Cloud itself is the internal personnel changes.

In March 2026, Lin Junyang, the key figure behind Tongyi Qianwen, left the company. The outside world believes that there are differences within the company regarding the AI strategy. Personnel changes can be seen as a strategic adjustment at best, or a loss of talent at worst. The stability of core technical personnel may affect the business.

After all, the stability of core personnel is the guarantee for the stability of technical products.

Over the years, there have been many downtime incidents at Alibaba Cloud. For example, there were payment and order failures on Alipay, Taobao, and Xianyu in December last year. Another example is the service interruption at the Hong Kong data center in June last year...

During the critical period of AI implementation, the tolerance for T0 - level failures like these is getting smaller and smaller. Next, how to improve the service stability and quality is also crucial.

Beyond Alibaba and Tencent, Baidu Smart Cloud has to take the most difficult path

Compared with Tencent Cloud and Alibaba Cloud, Baidu Smart Cloud faces a more complex situation.

Among all Chinese cloud providers, Baidu Smart Cloud was the earliest to verify its route and has the deepest technical accumulation, but it is also in the most delicate situation.

When most of its peers were still selling IaaS computing power, Baidu set its sight on the "cloud - intelligence integration" strategy - packaging its self - developed Kunlun chips, PaddlePaddle deep - learning framework, and Wenxin large - model into a full - stack AI solution.

Baidu Cloud was the first to catch the biggest trend in the industry: AI, which allowed Baidu Smart Cloud to occupy a high - ground in the AI cloud service market for a long time.

No one expected that although Baidu verified the direction of the AI cloud, it was Alibaba and Tencent that reaped the fruits.

Currently, the biggest dilemma for Baidu Smart Cloud may be the loss of the scarcity of AI.

Firstly, in terms of models, its first - mover advantage has been leveled or even surpassed.

In the C - end application field, Doubao's monthly active users have doubled to 226 million, ranking first by a large margin; DeepSeek reached a peak of 187 million, ranking second; Tencent's Yuanbao, relying on the WeChat ecosystem, ranked third. It is difficult for Baidu's Wenxin to catch up.

After the equalization of AI, the impact on Baidu Smart Cloud's business cannot be ignored. Although the combination of the Wenxin large - model, Kunlun chips, and the Baige platform is still highly recognized by enterprise customers, Baidu's AI cloud's annual revenue is about 30 billion yuan.

While Alibaba Cloud's single - quarter revenue has reached 41.6 billion yuan.

Secondly, "full - stack AI" is no longer exclusive to Baidu. Both Huawei and Alibaba are following up.

"Full - stack self - development" was once the most distinctive label of Baidu Smart Cloud, but today, the ability of the full - stack self - developed AI technology system + Kunlun chips also seems to be losing its scarcity.

Baidu has Kunlun chips, while Huawei has Ascend, and Alibaba has T-Head. In 2024, Huawei's Ascend accounted for about 23% of the Chinese AI chip market, ranking first among domestic products, while Kunlun chips had a domestic market share of over 8%.

In addition, Alibaba is also developing full - stack AI.

Alibaba's "Tongyunge" system is replicating Baidu's "cloud - intelligence integration" path. In the AI cloud niche market, Baidu has also encountered Alibaba as a competitor.

In addition, the rise of Doubao has given ByteDance's Volcengine new opportunities.

Volcengine's strategy is extremely clear: stimulate the call volume with extremely low prices, expand the scale of Token distribution, and thus drive the consumption of computing power and the revenue of cloud infrastructure. This approach of "selling Tokens instead of servers" is also constantly bringing new competitive pressures.

Therefore, the only card that Baidu Smart Cloud can play may be the Kunlun chips.

According to the information on Tianyancha APP, the post - D - round valuation of Kunlun chips is about 13 billion yuan.