Ant Sustainability Report: Focus on AI and data in the next decade

From 2021 to 2023, Ant Group's technology investments were 18.8 billion yuan, 20.46 billion yuan, and 21.19 billion yuan, respectively, with a compound annual growth rate exceeding 6%.

What is Ant Group's business model?

"Using technological innovation to provide more development opportunities for small and medium-sized enterprises and ordinary people. This is our business model and also the social significance of our existence." In the preface to the 2023 Sustainable Development Report released on June 13, Ant Group Chairman and CEO Eric Jing stated.

A few days prior, at a communication meeting, Ant Group's Chief Sustainability Officer Peng Yijie also said, "What we have seemingly been doing over the past 20 years is payment and financial services, but behind it all is the use of technology to solve a social problem."

The year 2024 marks the 20th anniversary of Ant Group's establishment. At the beginning of the year, Ant announced a new round of organizational structure upgrades, involving its three major business sectors: technology, Alipay, and globalization, corresponding to its "AI First," "Alipay Dual Flywheel," and "Accelerate Globalization" strategies, respectively. The company has also appointed several executives with technology and globalization backgrounds, while the previously much-anticipated digital finance business has been given less emphasis.

In the aforementioned report, Jing stated that development is a prerequisite for having a future, and technological innovation is the most fundamental driving force for development. According to Alibaba's financial report estimates, Ant Group's net profit for 2023 was 28.32 billion yuan, a decrease of over 20% from the previous year. During the same period, Ant Group continued to increase its technology investment. In 2023, the technology investment grew by 3.6% to 21.19 billion yuan.

Among this research and development investment of over 21.1 billion yuan, AI (Artificial Intelligence) technology and data element technology are the core investment directions for Ant. By the end of 2023, Ant had established 16 joint laboratories with nine prestigious schools, covering fields such as data security and privacy protection, intelligent computing, distributed databases, and remote sensing.

"Large models and data elements are two key elements that technology companies must have for future AI layout, and the two complement each other. Large models rely on massive data for training and optimization, thereby continuously improving the algorithmic capabilities of the models; while data elements need to rely on the learning and processing of large models to continuously release value," said Liu Chen, a researcher at the Bank of China Research Institute, to Caijing.

"In the next decade, Ant Group is focusing on the AI track and data elements, which we believe are two very key technological areas. We hope to start a brand new future for Ant Technology Group," said Peng Yijie.Net profit for 2023 was 23.82 billion yuan.

Since the basic completion of the rectification of financial services in July 2023, Ant Group has been under high external scrutiny regarding its subsequent development path. At that time, Ant Group announced that it would continue to enhance its technological research and development capabilities, striving to serve the real economy, especially small and micro groups, and create greater value.

At a recent exchange meeting, Peng Yijie stated, "We are still grassroots internet entrepreneurs."

The sustainability report indicates that Ant Group is continuously increasing its investment in technology. Since 2022, Ant has disclosed the scale of its annual technology investment through its sustainability reports for three consecutive years. From 2021 to 2023, Ant Group's technology investments were 18.8 billion yuan, 20.46 billion yuan, and 21.19 billion yuan, respectively, with a compound annual growth rate of over 6%.

"Ant's technology investment in 2023 reached 21.19 billion yuan, a historical high, which is a major highlight of the 2023 sustainability report," said Wang Pengbo, a senior analyst in the financial industry at Broadband Consulting.

At the same time, the annual report data released by Alibaba recently shows that Ant Group contributed 7.86 billion yuan (equivalent to 1.088 billion USD) in profit for the whole year of 2023, a decrease of about 23.6% compared to 2022. Based on Alibaba's 33% shareholding, Ant Group's net profit for 2023 was 23.82 billion yuan.

Alibaba's annual report stated that the decline in Ant Group's profit contribution year-on-year reflects the fine of 7.07 billion yuan imposed on Ant Group by Chinese regulatory authorities in July 2023.

In fact, Ant Group's net profit has been declining consecutively. From 2021 to 2023, Ant Group's net profits were 72.98 billion yuan, 31.19 billion yuan, and 23.82 billion yuan, respectively.

"Over the past few years, Ant has faced many challenges, but the investment in scientific research has increased instead," Peng Yijie said, noting that in 2023, Ant Group's investment in technological research and development accounted for more than 11%, ranking seventh in the national federation of industry and commerce's private enterprise technology investment rankings.The report states that these research funds are primarily directed towards the development of cutting-edge technologies such as AI and data element technologies.

"Technologies related to artificial intelligence and data elements are also the focus of development for all industries in the future," said Wang Pengbo, "Looking at Ant Group's investment in the technology sector, it should also be based on its own characteristics and needs for layout."

As of 2023, Ant Group has established 16 joint laboratories with 9 renowned institutions both domestically and internationally, covering cutting-edge topics such as privacy computing and AI, and has set up a scientific research fund to establish an organizational platform and support system for the exploration and development of cutting-edge technologies. Globally, Ant Group holds a total of 22,102 authorized patents, with 95.7% being invention patents, and ranks first in the world in the number of authorized blockchain patents.

Research and development funds are mainly invested in AI and data elements

AI large models are one of the key focuses of Ant's research and development investment.

According to relevant personnel from Ant Group, Ant's large model approach is different from companies like Alibaba and Baidu. It neither sells the underlying large models nor sells computing power, but instead applies large model technology to products and serves users through products.

"We hope that artificial intelligence can really serve every ordinary person in the future, just like QR code services," said Peng Yijie. Currently, Ant has focused on "three butlers" in artificial intelligence, namely the life smart butler, medical butler, and financial butler, corresponding to the three AI applications "Alipay Smart Assistant," medical assistant "An Zhen Er," and financial butler "Zhi Xiao Bao." At present, the underlying technology of these "three butlers" is the Bailin large model developed by Ant.

Taking Zhi Xiao Bao as an example, as early as 2018, the Ant Smart Financial Assistant project had been initiated, with the internal code name "Anna." Zhi Xiao Bao 1.0 used a small model, which had many limitations in semantic understanding, financial problem analysis and reply, and the provision of professional services.

After applying large model technology, Zhi Xiao Bao 2.0's financial knowledge data storage capacity reached the level of billions, and it can serve hundreds of millions of users at the same time. Lu Xin, the technical head of Zhi Xiao Bao, said that if the Zhi Xiao Bao of the small model era is compared to an elementary school student, then the Zhi Xiao Bao 2.0 of the large model era is equivalent to a college graduate, or even the average level of an industry expert.

"Ant has long been clear about the three main directions for launching large model commercial applications, emphasizing the degree of combination with scenarios. The advantage of doing so is that it can not only effectively reduce operational costs but also accumulate a large amount of AI usage experience and a large amount of related data early on, cultivate user habits, and seize the industry's opportunity," said Wang Pengbo.However, Peng Yijie frankly stated that the current "three butlers" are still in the early stages and will require continuous optimization and iteration. In addition, according to the aforementioned Ant Group insider, apart from the "three butlers," Ant Group is also brewing AI to B services, and has currently established a company named "AI Force" in Beijing to advance related businesses.

In this regard, Liu Chen warned that the development of AI large models is still facing two major pain points.

On one hand, the issue of computational constraints on model training needs to be urgently addressed. Under the intense market competition, deploying and maintaining large models requires significant technical investment and resource costs. Taking ChatGPT as an example, GPT-5 is expected to be released between the end of 2024 and 2025, requiring a computational load 200 to 400 times that of GPT-3.

On the other hand, the model capabilities and data quality still need to be improved. Currently, general large models for vertical industry applications mostly remain at the margin of improvement, without fundamentally solving the core pain points that traditional AI technology has not addressed, such as insufficient accuracy and limited professional capabilities.

Furthermore, Liu Chen indicated that the effectiveness of large models is highly dependent on data quality. How to connect various data circulation channels within the industry and accumulate more professional training corpora will continue to be a significant issue that needs to be addressed in the coming period.

Data element technology is another key focus of Ant's research and development investment.

Jing Xiandong believes that for data elements to be "used well," the key is to "flow well." The circulation and development of data elements can be divided into three stages: the "well" style of self-harvesting and self-use, the "bottled water" style of point-to-point circulation, and the "urban water supply network" style of trustworthy circulation between industries and regions.

"In the third stage of data circulation, we have made some interesting attempts," Peng Yijie introduced, saying that Ant has used blockchain and intelligent IoT technology to encrypt and chain the main assets of a new energy battery swapping service provider—such as the charging and discharging, estimated lifespan, and other related data of new energy batteries. These data, once chained, have helped the company reduce the comprehensive financing interest rate for operational lease loans from the previous 10% to 6%.

However, the circulation of data may have strong negative externalities. Industry insiders have indicated that with various types of data not yet circulating and integrating on a large scale, the negative externalities of data integration and use have already begun to emerge. Events such as big data price discrimination and delivery personnel trapped in systems are warning of the potential risks that data integration and use may bring.

In recent years, the development of privacy computing technology has provided a feasible solution to the aforementioned challenges. Through privacy computing, it is theoretically possible to achieve data usability without visibility, with controllable and measurable usage. "Data usability without visibility" can prevent information leaks during data circulation, and combined with "controllable and measurable usage," it can ensure that data is used according to pre-agreed methods and purposes, preventing misuse.Ant Group has stated that it is exploring next-generation privacy computing technologies, striving for efficiency and ease of use akin to plaintext computing, allowing the flow of data value to be as readily available as turning on a tap.

It is noteworthy that a privacy computing expert has cautioned that merely applying privacy computing does not mean it can solve all circulation issues. Alongside technological development, it is also necessary to complement with systems, regulations, standards, and so on, to truly establish a sound circulation order.

Aiming for Carbon Neutrality by 2030

Since launching its ESG sustainability strategy in 2022, Ant Group has consistently emphasized the concept of "dual value creation" that integrates business value with social value.

"Public welfare projects such as Ant Forest have been ongoing for 5 to 10 years, and from a practical standpoint, they align with the nation's expectations for the social value of internet companies," said Wang Pengbo.

The sustainability report indicates that by the end of 2023, Ant Forest users, with funding from Ant Group, have planted a cumulative total of 475 million trees. Over 690 million Alipay users have participated in Ant Forest, adopting green and low-carbon lifestyles in their daily lives, accumulating over 31 million tons of "green energy" through carbon emissions reductions.

Beyond Ant Forest, Ant Group is also advancing its own carbon neutrality process.

The report shows that Ant Group has achieved carbon neutrality for operational emissions for three consecutive years and plans to achieve corporate carbon neutrality by 2030.

In 2023, Ant Group's absolute emissions for Scope 1 (direct greenhouse gas emissions) and Scope 2 (indirect greenhouse gas emissions from electricity generation) decreased by 43.19% compared to the base year of 2020, surpassing the reduction target of 30% from 2020.

Additionally, Ant Group conducted a comprehensive inventory of Scope 3 (other indirect greenhouse gas emissions) for the first time, covering 12 categories related to activities upstream and downstream of Ant Group's value chain, providing a scientific basis for promoting carbon emission reductions across the value chain. Data indicates that in 2023, the emissions from 11 categories within Ant's Scope 3 amounted to 764,179.17 metric tons of carbon dioxide equivalent.The statistical difficulty of Scope 3 carbon emissions is widely recognized as a challenging issue in the industry. Zhao Lijian, President of Carbon Trust China, stated, "The scientific quantification and public disclosure of Scope 3 emissions are a tough 'hard bone' for the vast majority of companies to tackle. The difficulty lies in the lack of local factors, calculation models, and other tools, as well as the availability and accuracy of carbon emission data from numerous other enterprises upstream and downstream in their value chain."

Guo Jun, Deputy Director of the Listed Company Supervision Department of the China Securities Regulatory Commission, once disclosed a piece of data showing that nearly 80% of large listed companies reported an inability to account for carbon emissions from their upstream and downstream industrial chains, as well as from joint ventures and cooperative enterprises.

Ant Group also frankly admitted in its report that there are two major challenges in the emission accounting process regarding data accuracy: First, the availability and accuracy of raw data from activities upstream and downstream in the value chain are limiting factors; second, there are limitations in methodology, such as the applicability of industry factors. Yan Meng from Ant Group's ESG office summarized the Scope 3 carbon audit as "unmanageable, unobtainable, and imprecise."