Baidu Cloud is striving hard, yet Alibaba and Tencent are outpacing it with even greater strength.

Authored by | Landong Business Zhao Weiwei

How do Baidu, Tencent, and Alibaba define their AI strategies? The answers will gradually unfold over the next month.

Firstly, Baidu will kick off its 2025 Baidu Cloud Intelligent Conference on August 28. Tencent Cloud is set to host its 2025 Global Digital Ecosystem Conference on September 16, followed by Alibaba Cloud's 2025 Cloud Town Conference on September 24. Interestingly, the chronological order of these events mirrors the escalating scale of BAT's cloud businesses, roughly maintaining a 1:3:5 ratio in terms of market share from last year.

Among the trio, Baidu Cloud lags behind.

Each cloud provider's annual conference serves as a platform to showcase the latest advancements and practical applications of AI cloud infrastructure, as well as an opportunity to attract potential clients.

From the currently booming robotics and smart hardware sectors to more traditional industries like finance and education, nearly every sector seeks to connect with internet users through cloud providers. Especially with the rapid adoption of AI, cloud providers are continuously innovating in underlying computational power and application development, striving to offer the technological prowess of large enterprises to small and medium-sized businesses on a paid basis.

Take the automotive industry as an example. At the Baidu Cloud Intelligent Conference, models like the Li Auto I8, Ledao L60, and Geely Galaxy M9 were showcased alongside Baidu's own Luobo Kuaipao, highlighting Baidu Intelligent Cloud's implementation in the automotive sector. Baidu Intelligent Cloud claims to cover the top 15 best-selling brands in the Chinese automotive market, achieving a 95% penetration rate among mainstream automakers.

In reality, every cloud provider is vying for major automotive clients.

"Baidu's strengths still lie in maps and search, and now its end-to-end voice capability is also a leading point in smart cockpits," explained a Baidu employee in the automotive industry. Meanwhile, a Li Auto staff member at the event mentioned that Baidu had hoped to collaborate with Li Auto on in-car maps but was unsuccessful. Currently, Li Auto still relies on AutoNavi Maps.

"In the field of intelligent driving, Huawei remains the strongest. Alongside Horizon Robotics, Zhuoyu (formerly DJI Automotive), and Momenta, these three intelligent assisted driving companies are collectively known as 'Di Da Hua Mo.' Huawei Cloud is also building the infrastructure for automotive intelligence and holds the largest market share in China's automotive cloud sector," said a technical leader from an automaker invited to the Baidu Cloud Intelligent Conference. He expressed interest in seeing Baidu's new technologies in automotive intelligence but remained cautious.

The automotive industry epitomizes the fierce competition among Chinese cloud providers. Most automakers are reluctant to tie themselves to a single cloud provider.

Take NIO as an example. Its exclusive partner for the in-car voice assistant NOMI GPT is Baidu's Wenxin large model, deployed in nearly 200,000 vehicles. Users can interact with NOMI for intelligent vehicle consultation, ambient magic light control, and encyclopedic Q&A. However, to comply with global in-car data security and user privacy regulations, NIO has integrated Volcano Engine's Jeddak AICC confidential computing platform.

A multi-cloud strategy has become the mainstream choice for automakers today. On one hand, it's due to the fierce competition among cloud providers offering more diversified market options. On the other hand, it's a consideration for the clients' own security and interests. From hardware-based intelligent manufacturing to software-assisted driving development, no one is willing to risk everything on a single provider.

The trend of interdependence among cloud providers' services is also inevitable in the future. "Some of our cloud services for overseas expansion in Southeast Asia also utilize Volcano Engine's services," revealed a Baidu Cloud employee to Landong Business. The rationale is to provide enterprise clients with higher cost-effectiveness.

All are transforming towards AI cloud. The Baidu Cloud Intelligent Conference also showcased its technological prowess, with the Baidu Baige AI Computing Platform releasing version 5.0 for AI computing and the Baidu Intelligent Cloud Qianfan Development Platform 4.0 for AI development.

Greater expectations are placed on Tencent and Alibaba. What releases will they bring to their respective cloud stages? How will they convert their technological advantages into sustainable commercial value?

Multi-Agent Collaboration: A Must-Win Battlefield

Last year, Li Yanhong appeared via online speech at the Baidu Cloud Intelligent Conference, stating that Agents are his most favored direction for AI application development.

This year, Baidu Cloud continues to advance based on this judgment. Shen Dou, Executive Vice President of Baidu Group, believes that in complex scenarios, breaking down a task and having multiple Agents work collaboratively can yield better results.

To enable multi-Agent collaboration to surpass the capability limits of a single Agent, Baidu's Qianfan Large Model Platform has been upgraded to version 4.0.

Among its features, over 150 models determine the upper limits of Agent capabilities. A notable recent addition is Baidu's video generation model, the Steam Engine. In terms of vertical model optimization, the Qianfan Huijin Financial Industry Model and Qianfan Visual Understanding Model can achieve the effects of a thousand-billion-parameter model with a scale of only a hundred billion parameters, suitable for applications in finance, e-commerce, and education.

To enhance the underlying capabilities of Agents, Baidu leverages its strengths in search by equipping Agents with AI Search MCP Server.

This addresses the urgent need for real-time search. Previously, large models had a cutoff date for training data and couldn't acquire the latest knowledge. With AI Search, AI toys can provide real-time answers to users' latest questions, and mobile assistants can offer timely and reliable responses to the latest weather changes, such as "What's the weather like in Beijing this weekend? Help me plan the best bird-watching route."

To enable multi-Agents to handle complex tasks collaboratively, Baidu follows the path pioneered by Google.

Agents need to communicate to automatically execute tasks. In April this year, Google released the Agent-to-Agent (A2A) communication protocol. Now, Baidu Cloud has become the first cloud provider in China to support the A2A protocol. This capability means it can interoperate with other A2A protocol-compliant Agent platforms and can also connect to enterprise Agents using non-standard access methods through protocol adaptation functions.

Currently, Qianfan 4.0 has integrated third-party premium services such as Alipay Payment and Tongcheng Travel, enabling Agents to leverage more scenario-based capabilities. With the collaboration capabilities of Agents, where are the additional scenarios for Agents to become "productivity tools"?

Baidu demonstrated a retail analysis and consulting Agent customized for retail clients, capable of quickly generating high-quality reports. In the manufacturing industry, by uploading a standard operating video, Baidu Cloud's Visual AI application "Yijian" can generate a Standard Operating Procedure (SOP) detection task to check for errors by employees on the production line, thereby improving product yield.

Additionally, Qianfan offers various collaborative scenario showcases for multiple Agents, aiming to unlock business scenarios and enable enterprises to efficiently develop their own Agents.

Currently, the most scenario-rich Agents are undoubtedly the digital employees created by cloud providers. This is also a competitive track pursued by Baidu, Volcano Engine, and Alibaba Cloud, essentially using Agents to collaborate with employees to enhance enterprise efficiency.

Volcano Engine has the HiAgent Digital Employee Dispatch Station, while Baidu's digital employees have been implemented in seven scenarios, including marketing and recruitment, offering out-of-the-box functionality backed by a suite of AI capabilities such as end-to-end voice large models and Huiboxing Digital Humans.

Shen Dou shared a series of numbers: Baidu Intelligent Cloud also serves over 65% of central enterprise clients and more than 460,000 enterprises, having created 1.3 million Agents. Behind this, Baidu Intelligent Cloud silently shoulders the responsibility of supporting the family. In the just-released second-quarter financial report, driven by the growth of the cloud business, Baidu's AI new business revenue exceeded 10 billion yuan.

The intelligent cloud, aiming at the AI cloud direction, is undoubtedly the potential for Baidu's growth. Building an Agent ecosystem is its key task.

Because in the internet era, super Apps mean monopolizing user time, while in the AI era, the work content and time of Agents will far exceed the limits of Apps, meaning they can bring higher economic value to cloud providers. Therefore, the current ecosystem design around Agents determines the future dominance of cloud providers.

What Will BAT Do in the Cloud?

After the latest quarterly financial report from Alibaba, the cloud capabilities of BAT have once again reached a point for horizontal comparison.

As the leader in domestic cloud services, Alibaba Cloud reported revenue of 63.5 billion yuan in the first half of this year, a year-on-year increase of 21.8%. Especially in the latest quarter, revenue reached 33.4 billion yuan, with a growth rate of 26%, the highest in nearly three years. Revenue from AI-related products has achieved triple-digit year-on-year growth for eight consecutive quarters.

The cost behind this high growth is high investment. In Alibaba's financial report, cash flow generated from operating activities decreased by 39% year-on-year, with a free cash flow outflow of 18.818 billion yuan, compared to an inflow of 17.372 billion yuan in the same period last year. The main reason behind this is the increased investment in cloud business infrastructure and "Taobao Flash Sale."

At Alibaba's financial report meeting, the last question raised was whether investing 50 billion yuan in the instant retail market instead of AI would bring better returns, given the faster growth and larger potential market in the cloud sector.

Jiang Fan, CEO of Alibaba's China E-commerce Business Group, responded that both the consumption behind AI and instant retail are historical opportunities that Alibaba must seize. From the perspectives of capability, resources, and cash reserves, Alibaba has sufficient resources for "saturated investment." The key issue is how to balance short-term and long-term returns, "as investment in AI will drive cloud growth, which will further improve in the coming quarters."

Now, let's look at Tencent Cloud. Although Tencent no longer discloses Tencent Cloud's revenue separately in its financial reports as it did five years ago, the increment brought by AI to Tencent Cloud is also significant.

Tencent Cloud's revenue is reflected in the "Financial Technology and Enterprise Services" section of the financial report. Compared to the over 20% revenue growth of Alibaba Cloud and Baidu Cloud, Tencent's Financial Technology and Enterprise Services growth is not as pronounced, with a year-on-year increase of 10% to 55.5 billion yuan in the second quarter of this year.

Tencent doesn't compare much with its peers but rather focuses on its own progress.

Tencent's financial report clearly states that "the growth rate of enterprise service revenue has accelerated compared to recent quarters," mainly due to "increased demand from enterprise clients for AI-related services, including GPU leasing and API token usage, as well as growth in merchant technical service fees."

In the AI large model battle, Tencent was not initially fast. It wasn't until the popularity of DeepSeek that Tencent accelerated its layout.

Compared to Alibaba and Baidu, one of the most important characteristics of Tencent Cloud is its "stability." Tencent's financial report highlights that the focus of Tencent Cloud services is on gross profit. In the past quarter, due to revenue growth and increased cost of benefits in cloud services, the gross profit of the Financial Technology and Enterprise Services segment, where Tencent Cloud is located, reached 29 billion yuan, with the gross profit margin increasing from 50% in the previous quarter to 52%.

How much is Tencent investing in AI? This can be seen from the capital expenditures in free cash flow.

Alibaba's dual-line battle in flash sales and AI has led to a decline in cash flow, and Tencent's free cash flow has also been affected, mainly due to payments of 22.9 billion yuan to support the development of AI-related businesses. In the second quarter of this year, Tencent's capital expenditures reached 19.1 billion yuan, a year-on-year increase of 119%. In the first half of the year, Tencent's capital expenditures totaled 46.583 billion yuan.

Looking at the AI investments of BAT, the differentiation has become very clear. There is no shortcut to overtake in this competitive field. Alibaba Cloud and Tencent Cloud are in a phase of high investment for high growth, while Baidu and Volcano Engine are competing on the value points of intelligent "AI cloud."

Alibaba is taking daring steps, as evidenced by its clear declaration during its financial report meeting that, over the past four quarters, it has poured more than 100 billion yuan into AI infrastructure as well as the research and development of AI products. In contrast, Tencent remained tight-lipped about the exact amount of its AI investment over the past four months at its financial report meeting. Nevertheless, according to pertinent data, Tencent's total capital expenditures in the AI sector have also surpassed 84.9 billion yuan. Tencent's President, Liu Chiping, emphasized that they will persist in boosting investment, adding, "However, we must also allocate funds judiciously, rather than simply pledging to spare no expense."

The burgeoning demand for AI is set to drive incremental growth in the cloud market in the foreseeable future. However, it will still take three to five years for infrastructure investment in AI to yield stable profit growth. This presents a prolonged challenge for all three BAT (Baidu, Alibaba, and Tencent) companies, with none eager to lag behind. The products and ecosystems that extend beyond Alibaba and Tencent's financial reports will soon come to the fore, revealing their true potential.