Microsoft recently announced that Microsoft 365 Copilot will be launched on the B side on November 1. based on application scenarios, efficiency improvement, function evolution, commercial space and other aspects, we reiterate our view of attaching importance to office software for AI application landing core scenarios. Jinshan Office held a 2023 Technology Open Day on September 20, announcing that WPS AI has been connected to all products and invited users to experience it. The company explained the technical path of WPS AI landing. Based on the consideration of high efficiency, low cost and realization of subdivided scene functions, it self-developed 7B and 13B small models on the open source base to support formula generation in table calculation and reading comprehension in text search. In addition, it also announced the combination of WPS AI and enterprise knowledge base to facilitate digital asset activation and form a closed loop. Since the beginning of this year, the company has anchored the three strategic directions of AIGC, Copilot and Insight on WPS AI, combined with the Interim Measures for the Management of Generative Artificial Intelligence Services of the Network Information Office, and we believe that the landing of WPS AI will start and the commercialization process is expected to accelerate. Maintain a "buy" rating.

Microsoft 365 Copilot announced the launch date, reiterating the importance of office software for AI application core scenarios.

according to Microsoft's official blog, on September 21, Microsoft announced that Microsoft 365 Copilot will be available to B- end customers on November 1, with the new launch of Microsoft 365 Chat and the introduction of Copilot new features in Outlook, Word, Excel, etc. We reiterate that office software is one of the core scenarios for AI application landing: 1) solid application scenarios: office scenarios have many connections, high information content and strong timeliness; 2) Clear efficiency improvement: AI + office software will bring efficiency improvement in AIGC (content creation), Copilot (intelligent assistant), Insight (knowledge insight) and other directions; 3) Clear evolution function: with the expansion of the testing scope and the formal push to customers, the function is expected to evolve rapidly in combination with feedback. 4) Commercial space is expected: Microsoft365 Copilot will be provided at US $30/month, and E3, E5, commercial standard version and commercial premium version will be charged at US $36/57/12.5/22/month, with a higher-than-expected increase.

WPS AI access to the full line of products, open to the user full component AI experience. On September 20, the company announced on the 2023 Technology Open Day that WPS AI, an intelligent office assistant based on a large model, has been connected to all products and users can experience all component AI functions. Since this year, the company has anchored three strategic directions of AIGC, Copilot and Insight: AIGC, WPS AI solves the problem of creating from scratch for users and provides an iterable document or first draft of demonstration; In terms of Copilot, WPS AI is expected to change the 28th Law through software interaction, simplify the WPS product interface and lower the user's use threshold. In terms of Insight, WPS AI will make full use of the structured and unstructured data accumulated by enterprises to realize knowledge reuse in the form of dialogue. Looking ahead, we are optimistic that WPS AI commercialization will accelerate, product use threshold exploration, function enhancement will drive penetration, AIGC led the growth of payment rate, with Microsoft 365 Copilot pricing as a reference, WPSARPU improvement space is broad.

Based on the open source base self-developed small model, high efficiency, low cost to achieve WPS AI subdivision function. The company demonstrated the progress of its own model for the first time on the technology open day. In view of the high cost of large model reasoning, operation and maintenance deployment, and the lack of extensibility in some niche scenarios, the company's research and development, algorithm engineering team based on the industry open source model, training 7B, 13B and other small models. In terms of engineering methods, the total corpus of 7B model reaches TB level, and the training efficiency on dozens of GPU clusters can reach 1600token/sec/GPU, which is better than other open source models in the industry in natural language understanding tasks. 13B model training data reached 1 trillion token, and the training utilization rate on GPU cluster was higher than 65%. In landing applications, such as WPS formula generation in table calculation scenarios and reading comprehension of unstructured data in document search scenarios, small models achieve fast response speed and controllable cost, providing further support for WPS AI landing in Copilot directions.

WPS AI combined with the new enterprise knowledge base to achieve digital asset-based content intelligence. The traditional knowledge base has high construction cost, slow response to new knowledge in the common domain, and low efficiency of traditional retrieval mechanism. In recent years, users have created a large number of documents in WPS, which are stored in the cloud to form cloud documents that can be dynamically updated and collaborated in real time. Based on this, the company combines WPS AI with an enterprise knowledge base to implement intelligent Q & A directly based on cloud documents. Relying on RAG (Retrieval Enhanced Generation) technology, the company combines private data with the general knowledge of large models to expand knowledge boundaries and solve the illusion of large models. In addition, the document library question and answer can also be integrated into other business systems of the enterprise, such as building a customer service robot in Jinshan collaboration, binding the document library question and answer, and assisting customer service work. We believe that the new knowledge base is expected to help users fully activate unstructured documents and data assets in the cloud, forming a closed loop of enterprise digital assets in the Insight direction of WPS AI.

risk factors: the company's new product promotion is not as expected risk; C- end user payment conversion is too slow risk; trust and government and enterprise demand is not as expected risk; market competition is increasing risk; information security risk.

earnings forecast, valuation and rating: with the upgrading of the company's membership system and the expansion of new products such as WPS AI and WPS 365, the company will further open up the future growth space through the product-cloud-collaborative closed loop. We maintain the company's 2023-2025 revenue forecast of 48.75/61.90/8.037 billion yuan, maintain the net profit forecast of 14.17/18.31/2.433 billion yuan, and the corresponding EPS forecast of 3.07/3.96/5.27 yuan. Since the beginning of this year, generative AI has accelerated its development under the blessing of a large language model, and changes in the external environment have further established the development trend of Xinchuang industry. We have selected Huada Jiutian, General Lun Electronics, Wealth Trend and Cambrian as comparable companies, with an average PS valuation level of 30 times compared with Wind's consensus expectation in 2024, considering the company's rare industry position as a leader in the field of domestic office software and the high certainty of AI and cloud collaboration development trend, we give a certain valuation premium, we give the company a target market value of 204.3 billion yuan (corresponding to 33 times PS in 2024), give the target price of 442 yuan, maintain the "buy" rating.