Source: Wall Street See
Author: Liu Baodan

In November last year,$PDD Holdings (PDD.US)$when the market cap surpassed,$Alibaba (BABA.US)$/$BABA-SW (09988.HK)$Jack Ma encouraged internally, saying that the AI e-commerce era has just begun and it is an opportunity for everyone.

In the past year, Alibaba's e-commerce and Aliyun businesses have been reshaped with AI. In terms of external investments, Alibaba spent billions and invested in China's "AI Five Dragons" all around.

Compared to external investments, Aliyun and DingTalk are the key players in Alibaba's AI layout. As the underlying technology of AI, Aliyun dominates the market, while DingTalk is heading towards super applications.

On June 26th, DingTalk announced at the 2024 Ecological Conference that it would open to all large model vendors, and companies such as 'The Dark Side of the Moon' are now on board with DingTalk. In addition, DingTalk has launched a new AI search product and upgraded many related features.

This is a complete big model landing chain, from the model to the product; DingTalk is attempting to connect the big model technology and office scenes, exploring more usage scenarios, and becoming a super entrance to the era of artificial intelligence.

The AI era has already arrived, and Alibaba is quickly shifting towards the business of big AI models.

Accelerate.

It is quite rare for the 'AI Five Dragons' of China to gather together to support one company.

At the press conference, DingTalk announced that in addition to Tongyi's big models, six other large model vendors, MiniMax, The Dark Side of the Moon, Orion Star, Baichuan, Zhipu, and 01 Wanneng, will collaborate with DingTalk. Thus, DingTalk's ecological open strategy fully covers the AI field.

With the introduction of large model vendors, DingTalk’s ambition is to become a collaborative office platform and application development platform that links big models with business needs, which is also the inevitable path for DingTalk to become a 'super application.'

Currently, DingTalk and large models have three types of cooperation models. DingTalk's AI capabilities for products such as IM, documents, and audio and video rely mainly on Tongyi's big model. Based on this, DingTalk will explore the application of different model capabilities in various products and scenarios according to the characteristics of each big model vendor.

For example, DingTalk is working with 'The Dark Side of the Moon' to explore educational application scenarios based on the large model's long-text understanding and output capabilities.

In terms of AI Agent development, DingTalk has opened its AI assistant development platform to big model ecological partners. Developers can create AI assistants on DingTalk and choose different big models in addition to the default Tongyi big model.

In addition, DingTalk and big model vendors will customize corresponding intelligent solutions for customers, and provide services such as model training and tuning, AI solution building, and customized application development.

With the introduction of many large model vendors, DingTalk can explore more usage scenarios in the office scene, which is currently the most important way to explore the commercialization of big models.

Wang Ziyang, director of the Business Model Innovation Research Center of Tsinghua University School of Economics and Management, told Wall Street News that the core factor of AI big model commercialization exploration is scenario mining, but this is not easy.

"The opportunity of the AI application ecosystem lies in combining it with scenarios, and after the application value of scenarioization is discovered, it can be commercialized." In the view of DingTalk President Ye Jun, selling resources is the most basic business model of big models, selling services is slightly better, and selling models and brands is the most powerful.

Regarding business models, Ye Jun has a clear understanding. Currently, the global industry has two main pricing methods for AI applications: one is subscription-based charging, such as Microsoft Copilot for Microsoft 365, and the other is billing based on usage. Currently, DingTalk is considering both methods.

Since the announcement of commercialization in 2022, DingTalk has been trying to connect its business model. Now, this demand is becoming stronger and stronger. Ye Jun stated earlier this year that the company will achieve profitability in the fiscal year 2025.

DingTalk has the advantage of scene and data, and its competitive barrier lies in constructing rich office and business digital scenes. By the end of 2023, DingTalk has had seven hundred million users.

In addition, the collaborative office field where DingTalk is located is considered to be the first scene to undertake the artificial intelligence trend, and the enterprise payment channels are relatively smooth, with potential for payment, which is the premise of commercial realization.

Therefore, Alibaba has also invested more resources in DingTalk. DingTalk is the first product of Alibaba Group to access the Thousand Questions Model. Ye Jun also stated that after the "Four Little Dragons" strategy, the group's help and support for DingTalk will be greater than before, and the space will be larger.

Ambition

In the strategic layout of the AI model, Alibaba is one of the fastest-responding technology giants.

While many people are still amazed by the technological breakthroughs brought by OPEN AI, Alibaba has already taken action. In April 2023, Alibaba officially released its self-developed Thousand Questions Model, which is less than half a year since ChatGPT became popular.

This is not accidental. From the introduction of Min Wanli, former chief scientist of Alibaba Cloud Machine Intelligence, in 2013, to the establishment of Alibaba Artificial Intelligence Laboratory and the DAMO Academy, Alibaba's layout for artificial intelligence has enabled it to enter the top tier in terms of technical capabilities. In early May, Alibaba Cloud's Thousand Questions 2.5 scored the same in GPT-4 Turbo.

By superimposing computing power on the basic model, Alibaba Cloud's business model is similar to the combination of Microsoft+OPEN AI. This is Alibaba Cloud's biggest trump card for the AI era. Moreover, with scarce computing resources in its hands, Alibaba has invested in almost all mainstream large model start-ups.

According to IT Orange data, Alibaba has invested in national mainstream large model start-ups such as Minimax, Moon's dark side, ZeroOne Everything, Zhupu AI, Bai Chuan Intelligence, with a total investment of over 13 billion yuan. According to Alibaba's disclosure, the company will invest about 800 million US dollars in Moon's dark side in the 2024 fiscal year, and purchase about 36% of its equity.

Alibaba is becoming a high-AI-content technology company. Wu Yongming proposed that the driving force for promoting industry development in the future will be technology-driven forces represented by AI, and Alibaba will have three important priority directions towards the future: technology-driven internet platform business, AI-driven technology business, and global business network.

Over the past year, the big model has undergone unprecedented technological iteration and closely related investment, but the unresolved problem facing all big model companies is how to achieve profitability, which is the key to verifying whether the big model can move from technology to application.

For Alibaba, the cloud business, as infrastructure, has achieved significant growth under the drive of AI. According to the quarterly financial report for March 2024, product-related revenue from AI has achieved triple-digit year-on-year growth, and the company expects overall commercial revenue to return to double-digit growth in the second half of the 2025 fiscal year.

But obviously, this is far from enough. For Alibaba, whether it is a self-developed model or an invested model, neither has run a business model so far.

Although it has a huge e-commerce application scenario, Alibaba also firmly believes that all products are worth redoing with big models, but the impact of big models on many mature businesses is still in the stage of reducing costs and increasing efficiency.

How to find the appropriate application scenarios for big model capabilities and run a business model has become a consensus among today's technology companies, which is especially the case for Alibaba that has reinvested in big models.

Last year, DingTalk ended the situation previously managed by Alibaba Cloud and was confirmed by Alibaba Group as one of the first batch of strategic innovation enterprises. DingTalk's strategic position within the company has been significantly improved, and it can face the broadest market with a more independent strategy, and its responsibilities and pressures have increased.

Over the past year, DingTalk has continuously increased its strategic layout in AI, from AI products to open AI PaaS platforms, and then to the release of AI assistant products, moving towards a super application.

From computing power to models, and then to scenarios. Alibaba is playing a big game, and its huge AI map has begun to show synergies. For example, DingTalk's ecological strategy can leverage Alibaba Cloud, and invested AI start-ups can also empower Alibaba's business development by using Alibaba Cloud computing resources.

In contrast to the world's leading technology companies, Microsoft, Google, and Meta all have their own strengths, but there are shortcomings in AI development capabilities, technological advantages, and cloud computing services, respectively. Companies that can have both large models and cloud computing service capabilities are few and far between.

Alibaba, which has deployed AI comprehensively, urgently needs to find a breakthrough in businesses such as DingTalk. Alibaba can also refresh the outside world's stereotype that it is no longer just an e-commerce company, but a technology company with the ability of the entire AI industry chain.

This is bound to be a difficult road.

Editor / jayden