Compared to five or six years ago, the current AI industry is undergoing fundamental changes, and the most obvious trend is that the widespread application of generative AI has made business departments an important driving force for technological innovation. Customers hope to use AI technology to solve practical problems and improve business efficiency, "said Dai Wen, General Manager of Amazon Web Services Greater China Solution Architecture Department, in an interview with media including Daily Economic News on December 19th.
In Daiwen's view, the competition among cloud service providers has extended from the technical dimension to the actual business needs of customers, which has also become one of the key focuses of Amazon Web Services' layout in AI.
On December 19th, Amazon Web Services released the Amazon Nova series of basic models in Shanghai and strengthened core services such as Amazon SageMaker, Amazon Bedrock, and Amazon Q. Amazon Web Services is one of the earliest leading cloud providers to lay out its AI business. According to international market research firm Gartner, in 2023, Amazon Web Services will have the highest market share in the global public cloud IaaS (infrastructure) market, at 39.0%.
But Amazon Web Services' competitive advantage is gradually shrinking. Since 2023, the competition between Amazon Cloud and Microsoft, the two giants in the Internet cloud computing field, has become increasingly fierce. In this war without gunpowder, Microsoft was the first to taste the sweetness of big model technology with its early precise investment in AI rising star OpenAI.
Data shows that by 2023, Microsoft Azure's global market share has climbed to 23.0%, an increase of 1.5 percentage points compared to the previous year. Under the pursuit of competitors, Amazon Web Services is also further strengthening its layout in the AI field.
How to use innovative methods to solve needs is a pain point
The IT industry undergoes a technological revolution every decade.
Ten years ago, "cloud transformation" became the mainstream trend, and Amazon Web Services successfully stood at the forefront of technological change with its relatively forward-looking layout. Since its first revenue announcement in 2014, Amazon Web Services' revenue scale has grown from $4.64 billion to $90.76 billion by 2023, an increase of nearly 20 times, and is expected to exceed $100 billion by 2024.
Nowadays, "AI transformation" is gradually replacing "cloud transformation" as a new trend in the IT industry, and AI has also become the most imaginative concept in Silicon Valley.
However, with the popularization and deep application of AI technology, Amazon Web Services is also facing more and more challenges. On the one hand, enterprises need more customized and differentiated AI solutions to meet their unique business needs; On the other hand, the complexity and uncertainty of AI technology also pose many risks for enterprises in the application process.
And this has also driven cloud vendors to undergo a series of new changes in their business operations. Dai Wen stated that in the era of AIGC, how to actively respond to user needs and provide out of the box data analysis capabilities to help enterprises quickly integrate AI technology has become a difficult problem that cloud vendors need to consider.
In terms of model development, cloud vendors need to consider how to enhance the factory capability of model implementation, reduce hardware costs, improve hardware performance, and provide enterprises with a more stable and efficient AI operating environment, "said Dai Wen.
Among them, grasping and identifying customer needs is a major challenge. Dai Wen said that customers often only raise intuitive requirements such as needing faster and cheaper GPUs, but do not disclose deeper specific requirements. How to use innovative methods to solve corresponding needs is a major pain point.
Daiwen stated that due to the diverse needs of users, Amazon Web Services will meet their needs comprehensively based on different user identities, from the underlying hardware and platform software, to the large model released this time, and to the related services launched.
Recently, Amazon Web Services also announced the release of its latest technologies in three areas: models, applications, and infrastructure. In terms of models, Amazon Web Services has released six basic models. In terms of applications, Amazon Q Developer has added three new agents for unit testing and integration with GitLab. In terms of infrastructure, Amazon Web Services has launched four innovations for Amazon SageMaker AI.
The technological trends and current situation have shown that there is no one big model that can handle all tasks. Dai Wen said that different scenarios have differentiated requirements for model capabilities. For example, when engaging in pure text interaction, only cost-effective pure text processing capabilities are required, without the need for image processing capabilities; If it involves computer monitor related operations and the system does not have a specific interface, it may be necessary to perform display barrier recognition, simulated clicking, and other operations in order to access large model capabilities. Therefore, it is necessary to "adapt to local conditions" and introduce diverse models to solve different specific problems.
Giants are vying for dominance, and the competitive landscape is beginning to emerge
Recently, the Amazon Web Services re: Invent conference ended in the United States, attracting high market attention. On the one hand, the market has high expectations for Amazon Web Services; On the other hand, Silicon Valley tech giants and Wall Street tycoons are also eager to know who can become the biggest winner in the surging wave of generative AI?
With the development of the industry, the answer is gradually becoming clear. Microsoft, Amazon, and Google have engaged in fierce competition in the field of AI based on their respective advantages.
In October of this year, OpenAI, the leader in generative AI, completed a $6.6 billion financing with a valuation of $157 billion, becoming a focus of the industry. Star startups such as xAI, Perplexity, and Anthropic have also started a new round of large-scale financing, with xAI's valuation reaching as high as $45 billion. After three years, the trend of generative AI remains the hottest investment field today.
Listed companies are also deepening their layout of AI business. In the first quarter of 2024, Microsoft's intelligent cloud division's revenue was $24.1 billion, a year-on-year increase of 20%. Among them, Azure and other cloud service revenue increased by 33%, higher than the previously expected growth rate of 28% to 29%.
The growth of cloud services mainly comes from the contribution of AI. The financial report shows that about 12 percentage points of Azure's growth comes from AI services. During the earnings conference call, Amy Hood, Executive Vice President and Chief Financial Officer of Microsoft, said, "The increased demand for the Microsoft Cloud platform and the growth in long-term commitments have driven our performance. In addition, we have seen an increase in the number of contracts for Azure exceeding $100 million
The expenditure of enterprise customers on AI business is gradually increasing. Specifically, there are two main ways for enterprises to purchase generative AI technology: one is to directly purchase products related to AI big models; The second is to call AI big models, which can be called from model developers such as OpenAI, Google, Microsoft, or the Microsoft Azure platform and Google Cloud platform.
In the generative AI ranking competition, the top AI model startups mainly generate revenue from selling AI model enterprise level products, their own AI model services, and collaborating with cloud platforms to sell related AI model capabilities, OpenAI、Anthropic、 Google's modeling capabilities are recognized by enterprises.
Meanwhile, cloud giants such as Microsoft, Google, and Amazon generate substantial profits by partnering with powerful model companies and selling enterprise services.
Looking ahead to the future, although the improvement of basic large-scale models has propelled AI application development from the "last mile" to the "last 100 meters", there are still many challenges in developing AI applications. This is precisely the significance of cloud computing platforms transitioning from technology to services. In the future, it is worth paying continuous attention to how this tripartite pattern will evolve in the continuous development process of generative AI.