If we step out of the predefined perspectives of this year's Cloud Town Conference, such as "Intelligent Leap in Cloud Computing, Industrial Transformation" and "The Third Wave of Cloud Computing," and observe the conference from the angle of "AI holds too many unknowns," what would we see?

We would see that for Alibaba Cloud at present, AI does not solely bring opportunities. It also necessitates facing four major challenges: the impact of the popularity and price reductions of public clouds on performance, uncertainties in investment returns, and how to establish a more mature and reasonable To B service system. Additionally, we would observe that domestic cloud service giants like Alibaba Cloud lag significantly behind international top cloud service providers in terms of revenue scale and technological innovation.

This reveals that while the technological revolution brought about by AI appears grandiose, there are hidden undercurrents beneath the surface. As AI's large models are gradually implemented and venture into "deep waters," the journey for giants like Alibaba Cloud will be fraught with challenges.

01

Alibaba Cloud's "Flywheel Effect"

The 2024 Cloud Town Conference, themed "Intelligent Leap in Cloud Computing, Industrial Transformation," boasts a massive scale and is touted as "the edition with the most AI hard technologies ever gathered." Leading AI startups funded by Alibaba, such as Zhipu AI, Moon's Dark Side, Baichuan AI, and Zero-One Everything, also made their debut at the conference.

Given its scale and lineup, this year's Cloud Town Conference not only echoes Alibaba Cloud's strategy of "AI-driven, public cloud priority" from a year ago but also underscores its ambition in "cloud + AI."

The keynote speech by Wu Yongming, CEO of Alibaba Group, Chairman and CEO of Alibaba Cloud Intelligence Group, garnered significant attention. His core viewpoints included: "Firstly, in the past 22 months, AI has developed faster than at any other time in history, yet we are still in the early stages of the AGI revolution. Secondly, the greatest potential of AI lies not in mobile screens but in taking over the digital world and transforming the physical world. Thirdly, AI computing is rapidly evolving and becoming the dominant force in the computing system."

This was a rare public appearance for Wu Yongming since becoming Alibaba's helmsman last year, underscoring the importance he places on Alibaba Cloud. Through his speech, Alibaba Cloud articulated its vision for AI for the first time, indirectly outlining what Alibaba Cloud's AI will achieve and its future leadership position.

Just days before the Cloud Town Conference, OpenAI unveiled its latest generation of o1 series models, hailed as "the most powerful and consistent to date." "With OpenAI widely recognized as the leader, Wu Yongming's remarks carried hidden meanings," observed cloud industry insiders. In Alibaba Cloud's narrative, the AI world will have clear divisions of labor, and Alibaba will occupy a place, not necessarily following in OpenAI's footsteps.

Alibaba Cloud has its own understanding of AI. Its AI strategy encompasses three parts: first, chips, where Alibaba's "T-Head" is responsible for AI chip development to enhance Alibaba Cloud's computing power for training large models; second, cloud services, which are the focus of Alibaba's AI strategy, with Alibaba Cloud aiming to become an AI computing infrastructure through the opportunities presented by large models; and third, large models and AI services, such as toolchains for training and deploying large models.

An industry expert told the author that Alibaba Cloud's approach is clear, positioning itself to excel in infrastructure and then fostering an AI application ecosystem through external investments and open large model development. "In contrast, AI applications on mobile phones appear far less ambitious."

Apart from large models, autonomous driving and robotics were also prominently showcased at this year's Cloud Town Conference, with Wu Yongming emphasizing these two areas in his speech. "The broadest application scenarios for AI lie in reshaping various industries, emphasizing productivity transformation and upgrades, with autonomous driving and robotics being two typical examples in this context," the industry expert opined. "Alibaba Cloud's message is that AI is essential for reshaping different industries, and public cloud services and AI capabilities are indispensable."

Through the conference, Alibaba Cloud aims to convey to the outside world that cloud service providers play an indispensable role in the AI landscape. For instance, large models with billions of parameters represent a systematic engineering encompassing computing power, algorithms, networks, big data, and machine learning, requiring ultra-large-scale AI infrastructure support. The combination of AI and public clouds becomes a rigid demand, upon which cloud service providers can create a flywheel effect.

Alibaba Cloud's flywheel effect manifests in two ways. Firstly, AI reshapes both infrastructure and application software, further unleashing the value potential of AI Infra, with new applications and demands reinforcing each other to form a "flywheel." Secondly, through "open-source models + open ecosystems," Alibaba Cloud enables models, computing power, applications, users, and ecosystems to drive each other forward, also creating a "flywheel."

Several cloud and AI industry experts compared Alibaba Cloud with Baidu Intelligent Cloud when discussing Alibaba Cloud. Baidu entered the cloud service market later and aims to leverage AI as an opportunity for overtaking competitors.

Interestingly, following the Cloud Town Conference, Baidu's Cloud Intelligence Conference debuted this week, creating a sense of competition between the two events. Unlike Alibaba Cloud, which uses the cloud to drive AI, Baidu leverages AI to drive the cloud, constructing a new "cloud + AI + industry" ecosystem through its four-tier architecture of chips, frameworks, models, and applications.

According to Tianyancha APP, Baidu Intelligent Cloud's services encompass underlying infrastructure, large model development and application, and end-to-end AI-native application development, enabling continuous upgrades to the "cloud-intelligence integration" architecture and improving revenue structure.

Baidu Intelligent Cloud aims to achieve revenue growth through computing power, models, platforms, and applications, and it has already seen results: the latest financial report shows that in Q2 2024, Baidu Intelligent Cloud's revenue reached 5.1 billion yuan, a year-on-year increase of 14%, with sustained profitability. Currently, only Alibaba Cloud and Baidu Intelligent Cloud have achieved non-GAAP profitability in the domestic cloud service market.

02

Four Tricky and Complex Challenges

Beyond showcasing its strengths, this year's Cloud Town Conference also revealed numerous challenges facing Alibaba Cloud.

The first challenge stems from public clouds.

AI's development heavily relies on public clouds. Public clouds, with higher computing power utilization, more open ecosystems, and greater potential for technological innovation, represent the future of cloud services.

After over a decade of development, the government and enterprise sectors have emerged as the primary demand drivers for cloud services, becoming a new business focus for cloud service providers. Consequently, cloud service providers shifted their focus to the government and enterprise markets several years ago, offering public, dedicated, private, and hybrid cloud services.

Due to policy and security considerations, private and hybrid clouds have seen significant growth in China, while public clouds have encountered numerous obstacles during the transformation and upgrading of traditional industries, leading to issues such as low computing power utilization, decentralized cloud infrastructure, low standardization, and high costs. Therefore, the widespread adoption of public clouds will take considerable time and cannot be achieved overnight, nor can it be transformed by AI large models alone.

This implies that cloud service providers, including Alibaba Cloud, will continue to face high investments, low growth, and low profits, further widening the gap between domestic and international cloud service providers.

The second challenge arises from product price reductions.

At the Cloud Town Conference, Alibaba Cloud announced further price reductions for three of its Tongyi Qianwen flagship models on its Bailian Platform. The price of Qwen-Turbo dropped by 85%, with a cost of 0.3 yuan per million tokens, while Qwen-Plus and Qwen-Max saw reductions of 80% and 50%, respectively. After the reductions, Qwen-Plus was 84% cheaper than industry prices at the same scale.

Alibaba Cloud believes that reducing prices through AI infrastructure is crucial for fostering future application growth. This strategy not only impacts the market landscape but also serves as a vital means of expanding the public cloud market, accelerating the popularization of AI applications.

By lowering the prices of large models, Alibaba Cloud stimulates a significant increase in invocation counts, thereby spreading out computing costs and gradually generating profits. In the long run, the revenue generated by the massive growth in invocation counts will outweigh the business losses caused by price reductions.

However, industry experts argue that enterprise customers are more concerned with whether large models and AI applications can solve practical problems. "Cloud service providers cannot rely solely on price reductions to create a positive business and revenue cycle for public clouds, large models, and AI applications. In fact, the capabilities and effectiveness of large models are currently unsatisfactory, which is the primary reason for price reductions," they noted.

Since May this year, cloud service giants including Alibaba Cloud, Tencent Cloud, Volcano Engine, and Baidu Intelligent Cloud have significantly reduced the prices of large model invocations. "Price reductions impact the industry, so everyone follows suit. Excessive price cuts can lead to a decline in existing revenue and an increase in incremental revenue. If incremental revenue fails to cover existing revenue, it will affect Alibaba Cloud's overall revenue," they added.

The third challenge comes from external investments.

Google CEO Sundar Pichai believes that in the face of large models, "the risk of underinvesting far outweighs the risk of overinvesting, even if we ultimately prove to have overinvested." Alibaba shares this sentiment. Investment professionals believe that Alibaba's substantial investments in large model startups are driven by the prospects and value of these investments, as well as strategic considerations for early business layout .

Leading large model startups like Zhipu AI, Moon's Dark Side, Baichuan AI, Zero-One Everything, and MiniMax, which Alibaba has invested in, all train their large models on Alibaba Cloud.

Investing in these companies brings new momentum. Firstly, by investing in various startups, Alibaba can broaden and solidify its position in the large model sector, gaining more leverage. Secondly, partnerships with large model startups offer enterprise customers more choices. Thirdly, as investees train their large models on Alibaba Cloud, it drives growth in Alibaba Cloud's computing power revenue.

However, the aforementioned investment professionals noted that large models are still in their infancy, lacking a clear development path. Large model startups require significant investments and burn through funds quickly, with uncertain prospects. Alibaba's heavy investments in these startups remain uncertain in terms of returns at present.

The fourth challenge lies in building a more mature and reasonable To B service system.

After the previous "land grab" mentality and the "one-stop-shop" model of "general contracting," internet cloud service giants have shifted towards improving growth quality and profitability. In recent years, giants like Alibaba Cloud have adopted healthy and sustainable strategies, aiming for high-quality growth and revenue and profit enhancements, which necessitate a mature and comprehensive To B approach, mindset, and business system.

"Internet cloud service providers have a To C background and have long lacked a genuine understanding of To B," remarked the aforementioned industry observers. "Cultivating a To B mindset within internet giants with a To C heritage is extremely challenging."

Taking Alibaba Cloud as an example, while the top-level design is in place, the challenge lies in establishing a genuine To B corporate culture, management system, and process norms to enhance internal and external organizational efficiency. "Whether Alibaba Cloud can stay the course and resist new temptations and pressure for growth is indeed a test for management," they added.

03

The Significant Gap with "International Clouds"

The Cloud Town Conference showcased Alibaba Cloud's ambition to become a global leader in cloud computing. However, it cannot be overlooked that Alibaba Cloud lags significantly behind international top cloud service providers in terms of performance.

To a large extent, this also reflects the gap between "Chinese clouds" and "international clouds."

Microsoft's fiscal year 2024 full-year and fourth-quarter financial results, released in July, revealed that its Intelligent Cloud business segment generated revenue of $28.5 billion, a year-on-year increase of 19%. In contrast, Alibaba's fiscal year 2025 Q1 financial results, released in August, showed that Alibaba Cloud's revenue grew by 6% to 26.549 billion yuan. Microsoft Intelligent Cloud's revenue scale is nearly ten times that of Alibaba Cloud.

This disparity underscores not only the difference in revenue scales but also in technological innovation capabilities between Chinese and American cloud service providers. For instance, Microsoft's AI assistant Copilot, directly targeting enterprise customers, is deeply integrated with Microsoft's suite of applications. The combination of Copilot and Microsoft 365 drives continuous growth in PaaS services like Dynamic ERP, CRM, and Power Platform. The driving force of Copilot has enabled Microsoft's software to continually increase its paid space.

Currently, Microsoft has established a relatively mature "cloud + SaaS + AI" model, with all three business lines growing rapidly, contributing to high revenue and profits. With cloud services at its core, every product and service Microsoft sells globally drives cloud growth, generating high growth and revenue.

This approach facilitates further product and service innovation, spreading out infrastructure costs. The resulting profits can fuel new waves of technological innovation, thereby creating new technologies, ecosystems, and market spaces.

Several factors contribute to the significant gap between domestic and international cloud service providers. Firstly, domestic cloud service providers have long been trapped in homogenized competition in the IaaS layer, failing to fully tap into customers' deep-seated business needs, which has weakened the revenue-generating capabilities of the more valuable PaaS and SaaS layers.

Secondly, the ecosystem is a crucial driver for cloud services. However, in China, the dominance of cloud service giants and their ever-expanding business boundaries have long impacted vendors in consulting, software, and integration, resulting in an incomplete and immature cloud ecosystem.

Thirdly, the coexistence of various cloud service forms, including public, private, hybrid, and dedicated clouds, has fragmented market demand. Coupled with clients' often customized service requirements, this has lowered cloud standardization, hindered the scalability and reusability of technologies, products, and services, increased operating costs, and made profitability challenging.

Fourthly, due to US restrictions on the export of high-end AI chips, the computing power costs for domestic AI large models remain high, posing significant obstacles to their research, development, and popularization. At present, domestic cloud service providers' AI chip capabilities need improvement, and it takes time for AI chips to evolve from research and development to mass procurement and, ultimately, the formation of a software ecosystem.

In fact, after strategic adjustments in recent years, domestic cloud service providers have reduced integrated projects and shifted their focus to self-developed products, leading to improved profits. However, it will take time to fully eliminate high investments, low growth, low profits, and the need for enhanced ecosystem, software, and AI computing capabilities.

Analysts suggest that domestic cloud service providers must abandon low-end competition tactics like price wars, business battles, and integration services. Instead, they should focus on the research and development of public clouds, high-end AI chips, large models, and AI applications, fundamentally enhancing the competitiveness of Chinese cloud services to compete with international cloud giants.",

04

Conclusion

The Cloud Town Conference brings two inspirations:

First, in the next few years, the domestic public cloud market will maintain a pattern of "one superpower and multiple powers", and Alibaba Cloud will still have an obvious advantage, but it will also face enormous pressure. How can Alibaba Cloud properly position itself, find new growth points, gain recognition from the capital market, and bring Alibaba's share price back to its peak? How can Alibaba Cloud achieve a global layout of cloud computing and AI and become a giant in international cloud services?

The biggest opponent in solving these challenges is actually Alibaba Cloud itself. For Alibaba Cloud, the most important thing is mindset. It must have the courage and perseverance to invest in and innovate technology. Without a commitment to technological innovation, there will be no future.

Second, the domestic cloud service market has gone through three stages. The first stage was the "land grab" phase, the second stage was improving technological strength, and now it has entered the third stage, where cloud computing is deeply integrated with large models, and cloud service providers seek their own differentiated competitiveness.

Overall, the competitive landscape of domestic cloud services is still in its early stages. As cloud service giants improve their businesses and ecosystems, and large models and related applications are implemented, giants like Alibaba Cloud will narrow the gap with international giants.

However, domestic cloud service providers need to reflect on years of product service homogenization and low-level competition. Indicators such as ROI, LTV, and CAC still require close attention, or else even if AI changes the cloud industry, cloud service providers will struggle to capture the market dividends brought by large AI models.