The rapid iteration in the tech industry keeps us on our toes, and the cloud services sector is now entering a new phase of transformation.

From last year's 'Battle of Large Models' to this year's 'Competition of AI Applications,' the cloud services industry has entered a new stage, initiating a new era of technological innovation that has persisted in cloud computing for years. The cloud service giants have embraced the charm of new technologies, found new development paths, and begun to wonder about the endless possibilities of 'Cloud + AI.'

In the eyes of the industry, 'Cloud + AI' signifies new suspense: How will cloud computing evolve, and where is it headed? What changes will occur in the market landscape of large models? What roles will cloud service giants play, and what impact will they have? Can large models and AI applications become the second growth curve for cloud service providers?

With a string of questions lingering, people eagerly await the answers.

01

New Stage: AI Applications Take Center Stage

In the blink of an eye, the AI wave represented by large models has surged in China for nearly two years. Public data shows that, to date, over 190 large models have been registered and launched in China, providing services to the public, with over 600 million registered users.

The latest research from Accenture, a renowned consulting firm, reveals that 59% of Chinese enterprises plan to increase their investment in digital transformation within the next year, a 6% increase from last year. Simultaneously, a significant number of Chinese enterprises aspire to harness AI technology to continuously innovate and accelerate their transformation. This not only underscores the potential demand for AI services across various industries but also highlights the enormous commercial value of large models.

For cloud service giants, key players in the large model arena, these insights and figures signify technological innovation and significant opportunities: The continuous iteration of large model capabilities enhances their commercial value and expands user bases, fostering new business logics and growth engines.

The implementation of large models has indeed transformed the cloud industry. They not only facilitate AI-powered business innovation and upgrades for industry clients but also drive infrastructure reconstruction and upper-layer application transformations, unlike previous technological iterations.

A review of the history of information technology reveals that this scenario is unprecedented. Prior to the emergence of large models, the IT industry was dominated by the combination of 'CPU + OS + Software.' However, the virtualization of computing, networking, and storage has transformed computing power into a mass resource, altering the capabilities and forms of cloud infrastructure.

In this new era, the combination of 'GPU + Cloud + AI' has emerged as the new norm. With GPUs providing computational support for large models, cloud service providers must reconstruct their service architectures from the bottom to the application layer, while large models necessitate reconstruction from research and development to application.

This underscores the current transformation and trends in cloud services: The future of large models and cloud computing are inseparable. The training infrastructure of large models necessitates computational power, storage, and networking, and their upgrades and iterations place even higher demands on computational power.

Simultaneously, MaaS (Model as a Service) has become a crucial component of the new cloud architecture, with cloud service providers vying to enhance their capabilities. This will significantly alter the technical systems of major cloud service providers. As cloud services and large models integrate deeply, cloud service providers will promote the servitization of large models and the scale-up of their applications through the cloud model.

As a result, cloud service giants such as Alibaba Cloud, Tencent Cloud, and Baidu Intelligent Cloud, after enduring downturns in recent years, have prioritized AI as their future development focus, Refactoring one's own technical business ， thereby driving performance growth. Their financial reports also indicate a gradual increase in AI-related revenue.

In August this year, Alibaba Group announced its quarterly results, revealing a 6% year-on-year growth in revenue for its Cloud Intelligence Group to RMB 26.549 billion in the latest quarter. Notably, AI-related product revenue achieved triple-digit year-on-year growth.

Baidu Group's 2024 Q2 financial report showed that intelligent cloud revenue reached RMB 5.1 billion, up 14% year-on-year. AI-related revenue accounted for 9% of total revenue, up from 6.9% in the previous quarter.

In discussions with the author, several industry insiders noted that a new generation of cloud computing is emerging, ushering in a new round of transformation in the cloud computing industry. The resulting technological iterations, industrial applications, and commercial conversions are expected to persist for years.

Meanwhile, the deep involvement and substantial investments of cloud service giants in AI have gradually dissipated the smoke from last year's 'Battle of Large Models.' AI applications driven by large models have now taken center stage.

This signifies the emergence of a new cloud service ecosystem centered on MaaS platforms and AI-native applications. AI-transformed application software will occupy a more prominent position within this ecosystem.

Due to the 'Matthew Effect' in the cloud industry, underlying resources such as data, computational power, and storage will be dominated by a handful of cloud service giants. Commercial opportunities in cloud services are gradually shifting upwards, congregating at the PaaS, MaaS, and SaaS layers, particularly SaaS, which will undergo fundamental changes in both research and development and business models.

Currently, cloud service giants are comprehensively upgrading and transforming their products using large models while also delivering them to various industries, assisting clients and partners in rapidly developing AI applications.

For instance, hundreds of Tencent products have integrated the Tencent Hunyuan large model, expanding their paying user base through intelligent upgrades. According to reports, Tencent Meeting revenue doubled year-on-year in the first quarter, while Tencent Work revenue surged 200% year-on-year. Additionally, Tencent has aided numerous enterprise clients in efficiently and simply developing AI-native applications using large models, enabling their rapid integration into production scenarios.

Thus, driven by the AI wave represented by large models, China's cloud industry has entered a new stage – 'New Cloud Service Providers' have initiated a 'Large Model and Application Elimination Round.' Against this backdrop, cloud service giants are gearing up for another round of intense competition.

02

Competition Among Giants: Strengths and Weaknesses

For cloud service giants in China's transforming cloud industry, the frequent mention of 'computational power,' 'large models,' and 'AI applications' stems from their eagerness for continuous new opportunities.

As cloud service giants' technologies and services continue to improve, they all aspire to find their direction and strategies in the increasingly competitive 'Cloud Services + Large Models + AI Applications' landscape in 2024 through AI.

AI industry experts state that as their technical systems mature, cloud service giants will compete across different dimensions centered on AI. The driving force and participation of numerous cloud service giants will not only rapidly manifest the effects of AI-related business models but also profoundly impact AI research and development, capabilities, applications, and model iterations across various industry scenarios, including the construction of the AI ecosystem.

Today, the AI layouts of domestic cloud service giants are increasingly clear. Tencent Cloud excels in fostering SaaS services empowered by generative AI, while Huawei Cloud continually expands its 'circle of friends' with its Ascend AI chips. Telecommunication operators' clouds persist in bolstering AI computational power.

According to several industry insiders, among the fiercely competitive cloud service giants, Alibaba Cloud and Baidu Intelligent Cloud are representative, constituting the two poles of China's current cloud and AI development, each with their strengths and weaknesses.

In 2023, Alibaba underwent significant business and organizational adjustments, with cloud services becoming the strategic focus of the group. Alibaba Cloud has a clear approach to constructing its technical service system, comprehensively upgrading cloud services through full-stack technological innovation, spanning from underlying computational power to AI platforms and model services. Built upon the IaaS layer, supported by PaaS layer components like containers, middleware, databases, big data computing, data warehouses, and data lakes, it underpins the foundational and industry-specific large models of the MaaS layer.

Simultaneously, Alibaba hopes to foster business synergies with numerous AI enterprises. Since 2023, Alibaba has invested in numerous AI companies, including leading startups such as Zhipu AI, Baichuan AI, ZeroOne, MiniMax, and Dark Side of the Moon. This investment model saves time and enhances the capabilities of Alibaba Cloud's generative AI business deployment.

However, issues arise as Alibaba Cloud's recent downtime incidents (including the recent bug in Aliyun Disk) repeatedly test its brand reputation. As a large-scale infrastructure service, the costs of establishing and rebuilding a service trust system are exorbitant.

Moreover, while Alibaba Cloud's revenue scale has grown compared to previous years, its latest quarterly revenue growth was still only 6%. The challenge remains daunting in breaking through the scale bottleneck of the 'AI + Public Cloud' combination, maintaining a leading position, distancing itself from competitors, and catching up with top global cloud service providers.

Unlike Alibaba Cloud, Baidu Intelligent Cloud has consistently leveraged AI to drive its cloud services. Since officially announcing 'ERNIE Bot,' Baidu Intelligent Cloud has launched cloud computing products related to large models, encompassing underlying infrastructure, large model development and application, and end-to-end AI-native application development. This continually upgrades Baidu Intelligent Cloud's 'Cloud + AI Integration' architecture, accelerating the restructuring of Baidu's business and revenue structure.

As the 'Cloud + AI Integration' architecture matures, Baidu has established a layout in AI chips, frameworks, models, and applications, forming an 'intelligent closed-loop path.' However, AI-native applications require more than simply replicating traditional software and mobile internet apps; they must solve issues that were previously unsolvable or inadequately addressed for clients. This poses significant challenges to Baidu's 'Cloud + AI Integration' strategy, large model industry ecosystem, and AI-native application ecosystem. Providing more effective empowerment tools for enterprise clients is the next challenge for Baidu Intelligent Cloud.

Several industry observers and research experts believe that Alibaba Cloud and Baidu Intelligent Cloud are two typical representatives of domestic cloud service providers. While their respective cloud and AI technology systems are well-established and their performance is rebounding, they face numerous tests and challenges.

This indicates that China's cloud service market is currently in a structural adjustment phase. Before the true dividends of China's AI large models arrive, the new round of cloud computing technological transformation will persist for years due to factors such as upgrades to technical and service architectures, application product deployments, and the establishment of a new AI-centric ecosystem, all of which require substantial time.

During this process, an 'Elimination Round' will commence, gradually eliminating or acquiring weaker 'players' in terms of funding, technology, and service capabilities by mainstream players.

03

Elimination Round: How to Break Through?

Industry observers suggest that the 'Elimination Round' in the cloud industry often unfolds through price reductions. The cloud industry is a capital-intensive sector, and with enterprise clients' fixed annual IT expenditures, the market is in a state of Stock competition . The expansion of client bases and the acquisition of larger contracts by one cloud service provider inevitably impacts the performance of others.

Cloud giants believe that price reductions can enhance cloud computing's penetration across industries, expand user bases and stickiness, unleash technological dividends by popularizing computational power, and drive non-internet industries from simple cloud adoption to deeper utilization, thereby fostering economies of scale, reducing marginal costs, and generating higher profits. These profits facilitate new resource procurement and R&D investments, consolidating competitive advantages.

The same applies to AI large models. In May this year, cloud service giants drastically reduced large model invocation prices:

For instance, Doubao's primary model under ByteDance was priced at RMB 0.0008 per thousand tokens in the enterprise market, 99.3% cheaper than the industry average. Alibaba's Tongyi Q&A GPT-4-level primary model, Qwen-Long, had an input price of RMB 0.0005 per thousand tokens, a 97% reduction, and an output price of RMB 0.002 per thousand tokens, a 90% reduction. Such drastic price cuts made it challenging for large model startups to compete on price.

'It's evident that cloud service giants aim to expand their markets and user bases through price reductions, which will be their long-term strategy,' estimated the industry observers. They anticipate that the 'Elimination Round' of large models, which began this year, will last approximately three years. Cloud service giants hope to stimulate significant growth in invocation counts by reducing large model invocation prices, thereby spreading computational costs and gradually generating profits.

Moreover, repeated price reductions can attract numerous developers and ecological partners. During this process, other large model service providers may collapse due to pricing and cost pressures, leaving only a handful of major players in the market – precisely the goal of cloud service providers in driving the 'Elimination Round.'

However, experts believe that this also reflects room for improvement in cloud service giants' large model strategies. In contrast, OpenAI's ChatGPT incurs annual research, training, and labor costs amounting to billions of dollars. This underscores the exceptionally high technological and financial thresholds of large models, necessitating continuous upgrades and substantial investments, along with numerous risks.

China's tech giants currently exhibit a trend of similar large model platforms, highlighting the homogeneity of large model product services. The crucial question is how much further growth large models can drive for cloud service giants. As investments intensify, it remains uncertain whether AI will deliver the abundant returns these giants anticipate.

The expert noted that, beyond one or two sufficiently robust foundational large models, other giants should focus on intermediary services such as computational power, data, and large model training or on truly value-creating industry applications derived from large models. 'Current cloud service giants are fraught with suspense, and the true test of AI large models lies ahead.'

04

Conclusion

Indeed, large models and AI applications have brought new changes and expectations to cloud service giants, accompanied by new challenges. In comparison, foreign tech giants have been more solid in developing large models.

Today, giants like Amazon, Microsoft, and Google have launched large model products through investments or in-house research and development. Notably, Microsoft has carved out a clear path in cloud, software, and AI large models, with these three businesses complementing and enhancing each other. Centered on these three businesses, Microsoft has gradually formed a comprehensive business layout spanning the IaaS, PaaS, and SaaS layers while actively engaging in investments and mergers and acquisitions to expand its business scope.

"From Microsoft's example, they have established a good business cycle," the above industry expert believes that first, the application of large models consumes a lot of computing power, which can drive the growth of cloud service revenue; second, large models will also optimize the functionality and experience of software applications, driving the growth of software business revenue, discovering new user needs, and in turn, further promoting the maturity of large models.

However, these practices, ideas, and experiences do not provide much benefit to domestic cloud service providers and large model service providers. Due to many differences in the Chinese and American cloud service markets, customer demands, technical services, etc., domestic cloud service providers do not have many examples to refer to when faced with large models. In terms of large models, the differences between Chinese and American cloud service providers will increase, and domestic large model service providers need to find their own pace, characteristics, and advantages.

Currently, domestic large models and AI application services are still in their infancy, with market size and industrial maturity yet to be improved, leaving vast room for growth. Competition among various parties is fierce. In this situation, how can we cultivate a rigid demand user group for large models? How to define the business boundaries of large models? How can large models become a stable and enduring business growth engine? How can privacy and data security be protected when large models are implemented in various industries?

Only by continuously building a healthy and autonomously controlled large model ecosystem and attracting more and more entrepreneurs, developers, partners, and customers to join, can large models and AI applications truly become the "rigid demand" for intelligent upgrades in various industries, forming new qualities of productivity, thereby truly revealing new momentum and advantages for innovative development.