With the growing popularity of DeepSeek, the Tongyi Large Model team faces mounting pressure. One Tongyi employee noted that their model, while well-regarded, has not yet gained widespread acceptance. As an open platform, Alibaba Cloud can deploy various models, but maintaining the leadership of its self-developed large models and "defending its open-source status" have become new mandates from senior management.

Alibaba Cloud is pivotal to Alibaba.

In June 2023, three months after Alibaba's restructuring and six months after the emergence of ChatGPT, Daniel Zhang, former Chairman and CEO of Alibaba Group, announced his resignation from group positions, retaining only the role of CEO of Alibaba Cloud. The end of Alibaba's "Daniel Zhang era" is inevitable, but his decision to focus on Alibaba Cloud was not entirely negative. Surprisingly, by September, Zhang was once again sidelined, with the CEO position of Alibaba Cloud being taken over by Wu Yongming.

According to the Financial Times, Wu Yongming actively pursued this position.

Wu Yongming conveyed to Alibaba's co-founders, Jack Ma and Joe Tsai, that as the group's CEO, he should oversee all aspects. Since the cloud business is crucial for unlocking Alibaba's future as an AI company, it is evident how vital Alibaba Cloud is to the Alibaba Group, with the group's CEO also serving as the CEO of the cloud business.

Two years later, this perspective has been clearly validated. With the influence of the DeepSeek model spreading in the international developer community, validating China's AI capabilities, foreign capital returning to China, and Alibaba Group's public declaration of "investing 380 billion yuan in AI," its share price has surged by nearly 80% within the year.

The market no longer views Alibaba solely as an e-commerce company. Among the nearly $100 billion increase in market value, the cloud business has emerged as the most significant growth driver.

How did Alibaba Cloud transition from aiming for an independent listing (valued at over $40 billion) but not being recognized by investors to becoming an Alibaba business that investors and the market are once again optimistic about?

01 Over the past two years after the "ChatGPT moment," the organization and strategy have become clearer

Back in 2023, although the new opportunities brought by AI were globally recognized, investors focused on short-term benefits were skeptical about the cloud business due to macroeconomic concerns.

At that time, the United States had expanded export restrictions on advanced computing chips, on which Alibaba Cloud's business heavily relied. Moreover, Alibaba Cloud's revenue growth was only in the single digits year-on-year.

More importantly, China's AI competition was still in its nascent stage. It was uncertain whether the most capable large models would emerge from Alibaba, Baidu, ByteDance, or whether only startups could bring surprises.

Everything takes time.

For Alibaba and Alibaba Cloud, AI represents a once-in-a-lifetime opportunity, potentially the most crucial for recovering from the setbacks of the e-commerce wars and antitrust investigations.

Alibaba acted swiftly. Soon after Wu Yongming took over the Alibaba Group, he announced a strategic shift to "AI first" with a contraction of the frontline. Consequently, Alibaba Group began exploring the possibilities of integrating AI into both its cloud and e-commerce businesses.

However, e-commerce is a mature industry, and its focus is more on applying AI capabilities to enhance business efficiency, such as helping merchants generate product details intelligently, deploying intelligent customer service, and providing automatic translation for cross-border transactions. Alibaba Cloud, on the other hand, faced a complete transformation to adapt to the future.

During this period, Alibaba Cloud saw three CEOs in just a few months, transitioning from Zhang Jianfeng with a technical background to Zhang Yong with a financial background, and finally to Wu Yongming, who had been an investor for many years.

Frequent business and organizational adjustments led to a sense of "turbulence" among most employees. One mid-level manager who experienced these adjustments and eventually left described the situation as chaotic, with everyone feeling disoriented, unsure of upper management's objectives.

Besides confusion, there was also fear, as organizational adjustments were accompanied by significant layoffs. For instance, in Alibaba Cloud's Hybrid Cloud Business Unit, some teams laid off 30% of their staff; the IoT hardware integration business line laid off over 50% of its staff; and the entire autonomous driving department of the DAMO Academy was disbanded, with only some personnel being integrated into the Cainiao Group.

Organizational transformation is harsh for those who are no longer needed. However, looking back, these adjustments by Alibaba Cloud were necessary, and the changes in three CEOs clarified and concentrated the business focus.

When Zhang Yong took over Alibaba Cloud, his primary adjustment was to streamline organizational relationships to better serve B-end customers. Simultaneously, Zhang Yong emphasized the importance of the public cloud business and reduced project-based sales orders.

However, during Zhang Yong's tenure, he was more focused on transforming Alibaba Cloud into an organization that better serves customers. In an internal sharing session, Zhang Yong stated, "We have spent a lot of time considering business growth but not enough on designing an internal operating system that provides services to customers and generates synergy between organizations."

Zhang Yong believed that technology should address the issue of unit cost. An Alibaba Cloud employee recalled that at that time, Zhang Yong's requirements for technology were "stability, security, performance, and cost." However, for Alibaba Cloud, advancing into the AI era meant that the most crucial aspect of technology was its advanced nature.

After taking over, Wu Yongming continued to amplify and focus Alibaba's investment in "public clouds." Compared to project-based orders, the public cloud model allows for the most efficient utilization of computing resources, which is more suitable for the demand for computing resources during AI training and inference processes. The public cloud can flexibly adjust resource allocation to meet peak demand for computing power while reducing waste during off-peak periods.

Moreover, he established the Infrastructure Business Unit and Infrastructure Committee within the organization, personally overseeing and coordinating the planning and construction of the group's underlying technical infrastructure.

At this juncture, Alibaba Cloud became clearer in its organizational and business strategy for the AI era, completing its initial transformation. While streamlining its organization and strategy, Alibaba Cloud also gradually found its positioning – to excel in infrastructure, building an AI ecosystem through public clouds, openness, and advanced large models.

02 Why is Alibaba Cloud aiming to be the "water, electricity, and coal" of the AI era?

During the Spring Festival of 2025, DeepSeek went viral on social media, and the Chinese people welcomed their own "ChatGPT moment." One investor described that if in the past, only 100 million people in China paid attention to and used large models, by 2025, at least half of the Chinese population began to feel the power of AI.

Cheaper and astonishingly performant large models ushered in a new dawn for the application ecosystem. Tencent's Yuanbao integrated with DeepSeek and began promoting it in lower-tier markets. Car brands, adept at marketing, quickly announced that their intelligent driving systems utilize DeepSeek, not to mention the numerous entrepreneurs developing "shell" applications.

Alibaba Cloud employees also had a busy Spring Festival. They were already supporting the Spring Festival Gala, and with the popularity of DeepSeek, their workload increased further. Speed was crucial, and by the sixth day of the lunar new year, Alibaba Cloud announced the official launch of the DeepSeek model, supporting users to deploy DeepSeek-V3 and DeepSeek-R1 on the cloud with one click. To drive adoption, Alibaba Cloud also introduced a new discount policy, providing 1 million free tokens to new users.

The launch of the DeepSeek model is a microcosm of Alibaba Cloud's "openness" strategy in the AI era. In addition to DeepSeek, Alibaba Cloud had earlier integrated high-quality pre-trained models from domestic and international AI open-source communities and provided the "Step" series of open-source models through "LeapStar." In January this year, Alibaba Cloud also announced a strategic partnership with ZeroOne to establish the "Industry Large Model Joint Laboratory."

According to an Alibaba Cloud employee, with the launch of DeepSeek, the usage of Alibaba Cloud's own Tongyi Qianwen model decreased by 20%. "However, large models are not Alibaba's competitive advantage. It's fine for everyone to use whatever they want. This is already clear," the employee summarized. Alibaba Cloud's strategy is to attract users with models and services, but the ultimate goal is to generate revenue through the cloud.

But with the popularity of DeepSeek, the Tongyi Large Model team faces more pressure. One Tongyi employee noted that their model's biggest issue is that it is well-regarded but not widely accepted. Although Alibaba Cloud, as an open platform, can deploy other models, maintaining the leadership of its self-developed large models and "defending its open-source status" have become new mandates from senior management. It is understood that to win this competition, they will soon release a new generation of models, Qwen 3.0, with enhanced performance.

By integrating other open-source models, developing more advanced large models, and providing them to developers for free, or even significantly reducing prices and giving free tokens to new users, Alibaba Cloud's ultimate goal is to expand its user base, establish scale effects, eventually flatten infrastructure construction costs, continuously reduce the average cost of using the cloud, and lay the foundation for longer-term benefits.

In the AI era, this is a viable layout with a high competitive advantage.

In the mobile internet era, Alibaba Cloud struggled to attract as many public cloud customers as AWS. Large companies across various industries in China were accustomed to controlling their own infrastructure, leading to low service and technology reuse across different customers and persistently high marginal costs. However, now, if large customers want to build their AI infrastructure and train large models from scratch, it would be an unaffordable investment.

In other words, in the AI era, the demand for public cloud computing platforms has intensified, and the larger the scale and the longer the term, the lower the marginal costs.

Based on this perspective, Alibaba Cloud's investment in infrastructure construction is resolute.

At the 2024 Cloud Town Conference, Zhou Jingren introduced a series of upgrades to Alibaba Cloud's AI infrastructure. At the beginning of 2025, Alibaba announced that it would invest 380 billion yuan in infrastructure construction over three years.

Over the past two years, with the establishment and improvement of the entire infrastructure system, open-source large models, and model-as-a-service strategies, Alibaba Cloud's customer demand has surged. At last year's Cloud Town Conference, Wu Yongming announced that over 50% of their new computing power demand is "AI-driven."

Now, Alibaba Cloud has formed a virtuous cycle of "investing in infrastructure – providing free models – cloud revenue recycling." Better still, as inference models become the new direction, the usage and demand for the cloud will increase significantly.

One cloud industry practitioner explained that in the past, the pre-training of large models was essentially a one-time process, but each calculation made by an inference model consumes computing power. "This demand will grow larger and larger, far exceeding that of training, and it is continuous."

Positioning as infrastructure is a choice that requires a long-term mindset. For Alibaba Cloud, it is just beginning to see the start.

03 The value of the cloud is recognized, and Alibaba's market value is returning

"Alibaba Cloud used to have zero valuation, but now it's fully valued." A US dollar fund investor told us that the market also expects that for every 2 percentage point increase in cloud revenue growth driven by AI, Alibaba's share price will rise by 1%. Currently, JPMorgan predicts that Alibaba Cloud's valuation may be comparable to that of Microsoft.

This is, of course, based on real revenue growth for Alibaba Cloud. According to Alibaba Group's financial report, by the last quarter of 2024, Alibaba Cloud's AI-related revenue had achieved triple-digit year-on-year growth for six consecutive quarters.

On the business front, whether it is the more than 200 models of different sizes released by Tongyi Qianwen or the recent announcement by the large model service platform Bailian that it will be the first in the industry to launch full-lifecycle MCP services, Alibaba Cloud has been continuously increasing its investment in ecological layout. Jack Ma's frequent appearances at the Hangzhou headquarters since 2025 also confirm the value of Alibaba Cloud today.

Supporting the market value is also Alibaba Cloud's technical capabilities. After the release of DeepSeek, Alibaba Cloud's Qwen2.5-Max model also received high international praise and outperformed DeepSeek-R1 in multiple benchmark tests. The cooperation with Apple also proved once again the reliability of Alibaba Cloud's technical capabilities and services.

In the past, 80% of Alibaba Group's revenue came from e-commerce, and the cloud business was once a "cost center." Today, with the gradual arrival of scale effects, the public cloud business platform has gradually become a "profit engine."

During the second quarter of 2024, Alibaba Cloud reported revenues of 26.549 billion yuan, while its adjusted EBITA profit surged by 155% year-on-year to reach 2.337 billion yuan. By the fiscal year's end, this profit had further increased by 33% to 3.1 billion yuan, positioning Alibaba Cloud as the sole profitable cloud service provider in China. In comparison, Taobao and Tmall Group experienced a growth rate of merely 8% during the same period, with the cloud business's profit margin (9%) closely trailing that of the e-commerce segment (12%).

A US dollar fund investor remarked that Alibaba's transformation from an "e-commerce company" to an "AI core player" marks a significant shift in its valuation model. "It dawned on everyone that Alibaba's previous valuation did not account for its cloud operations."

JPMorgan's estimates suggest that if Alibaba Cloud were valued using the average multiple of 6.5 times for US SaaS companies, its worth could hit $115 billion, potentially lifting Alibaba's overall valuation to $320 billion. Currently, Alibaba's market capitalization stands at $314.1 billion (Note: Due to external factors and volatility in the US stock market, Alibaba's market value based on the latest closing price is approximately $250 billion).

While Alibaba Cloud has propelled the Alibaba Group out of its undervalued quagmire, the journey of commercial competition in the AI era is still in its infancy. The commercialization of AI remains unproven, and the anticipated explosion of AI applications is yet to materialize.

Alibaba's Tongyi Qianwen may not be the most powerful AI offering among tech giants, but it stands out for its openness. In the competitive landscape against Huawei Cloud and Tencent Cloud, Alibaba Cloud has maintained its distinct identity and market share. However, given the substantial investments in long-term infrastructure, Alibaba Cloud is advancing with considerable weight. Combined with the fierce rivalry from Huawei Cloud and Tencent Cloud, the road ahead is fraught with challenges and significant uncertainties.