AI-enabled productivity tools

YXT.com has used AI to evolve its corporate learning management software suite and now offers four distinct products: TalentNova for talent management, NeoLearning for training and knowledge management, SaleSmart for sales force training and AI BOX, the engine that powers the three solutions and serves as a platform to build AI agents. These AI-enhanced products provide scope for YXT.com to improve net revenue retention (NRR) and attract new customers. The company believes it is more efficient to serve the small and medium-sized enterprise market via a channel distribution model and is growing its partner network accordingly.

Targeting operating cash flow breakeven in FY26

The company has been loss-making over FY22–25, albeit with losses reducing every year as the use of AI boosts automation and efficiency, driving gross margin expansion and lower operating costs. Now in a small net debt position, a key target for management is to achieve operating cash flow breakeven, which we forecast in H226 through a combination of revenue growth, higher gross margins and continued cost reductions.

Valuation: Significant upside potential

Our DCF analysis, which uses a WACC of 11.6% and a long-term growth rate of 2%, generates a base case valuation of $1.40 per ADS. This uses our forecasts to FY28, revenue growth trending down to the long-term growth rate by FY35, and EBITDA margin expansion to 18% by FY35. This equates to an EV/sales multiple of 1.7x FY26e and 1.5x FY27e, slightly above Hong Kong-listed software peers and at a significant discount to Shanghai STAR-listed software peers. Key triggers to reduce the discount include positive momentum in subscriber numbers, growing NRR and substantial cost reductions in FY26, supporting progress towards cash breakeven.

Investment summary

YXT.com was established in China in 2011 and has since become the leading provider of corporate learning and development software in China. Its software is used by companies that want a system to create and provide the necessary training to ensure their staff are proficient at their jobs, to provide a process to manage the careers of staff from initial hiring through to progression through the company, and to manage the base of internal knowledge. Since 2022, the company has been through a process of improving internal efficiency through the use of AI tools while redesigning its product suite to make optimum use of AI. This has resulted in a lower cost to develop and support software while providing new functionality that improves personalisation, increases interactivity and enables companies to develop training content at a much lower cost. It has applied this learning to a new module, SaleSmart, which is designed to help train and manage sales teams, and in the longer term, we believe the company is planning to develop new products for other business functions as part of its intelligent productivity suite. The company’s current strategy for growth is threefold: 1) deepen relationships with large enterprises to improve retention and cross-selling potential, 2) continue to scale AI solutions and 3) manage costs, balancing targeted investment in high-return areas with optimisation of the existing cost structure.

Financials

YXT.com’s recent results highlight the company’s focus on large enterprises. Pro forma revenue has grown for the last two years despite a decline in subscriber numbers, indicating growth in revenue per subscriber. The use of AI to automate multiple processes throughout the business has resulted in gross margin expansion and a reduction in operating costs, and we expect this trend to continue. Our forecasts assume growth in subscriber numbers and revenue per subscriber over the forecast period. We expect a substantial decrease in operating costs in FY26 followed by minimal growth in FY27 and FY28. Management is targeting operating cash flow breakeven in FY26; our forecasts are slightly more conservative assuming that this is achieved in H226.

Valuation

Our discounted cash flow (DCF) analysis, which uses a WACC of 11.6% and a long-term growth rate of 2%, generates a base case valuation of $1.40 per ADS. This uses our forecasts to FY28, revenue growth trending down to the long-term growth rate by FY35, and EBITDA margin expansion to 18% by FY25 (average 13.9% FY29-25). This equates to an EV/Sales multiple of 1.7x FY26e and 1.5x FY27e, slightly above Hong Kong-listed software peers and at a significant discount to Shanghai STAR-listed software peers. Key triggers to reduce the discount include positive momentum in subscriber numbers, growing NRR and substantial cost reductions in FY26, supporting solid progress towards cash breakeven. As our forecasts to FY28 assume a comprehensive turnaround of the business, we have also performed a reverse DCF that attempts to estimate the growth and margin assumptions factored into the current ADS over FY26-35. Revenue growth of 3% per annum and the EBITDA margin expansion from -22% in FY26 to 12% in FY35 result in the current ADS price of $0.35; these assumptions are substantially lower than our forecasts.

Sensitivities

Company-specific sensitivities include the pace of adoption of AI solutions, the reversal of subscriber declines, the timing to reach cash breakeven, the ability of the company to grow outside China, the dual-class share structure, the risk of delisting and the corporate governance for the Cayman Islands-registered listed company. Risks related to the company being headquartered and operating in China include the variable interest entity (VIE) structure, Chinese licencing requirements, Chinese currency controls, exposure to RMB/USD exchange rates, US/China geopolitics and possibility of Chinese state intervention.

Company description: AI-enabled corporate learning software

YXT.com is a Nasdaq-listed provider of digital corporate learning solutions. Under the name Radnova, it predominantly operates in China where it has a leading position.

The overall goal of the company is to help companies unleash productivity through the use of AI. Using data it has accumulated over many years of providing corporate learning and development software, it understands how companies turn inexperienced new joiners into proficient and productive employees. Its AI-enhanced solutions help companies build more resilient workforces able to adapt to rapidly changing market conditions and technological advances. Its software enables companies to:

• Manage the talent within their workforces better;
• Manage private domain knowledge;
• Generate training courses and individual learning paths; and
• Provide a work co-pilot.

Background

YXT.com was founded in 2011 as Yunxuetang Network. In 2013, it released corporate learning software products on a trial basis, and in 2015 it launched commercial software products. From 2017, it built a content ecosystem including collaborations with content partners, and from 2019, it started developing content in house. By 2023, the company had become the largest digital corporate learning solution provider in China by total revenue, subscription revenue and number of subscription customers, according to market research company Frost & Sullivan.

YXT.com listed on Nasdaq in August 2024, issuing 2.273m American depositary shares (ADS) (equivalent to 6.819m Class A ordinary shares) at a price of $11.00 per ADS, raising gross proceeds of $25.0m/RMB178.1m.

The chart below shows the group structure. The entity at the top of the chart, which is listed on Nasdaq, and the two entities below it are incorporated outside China and are deemed offshore companies. The Hong Kong-incorporated entity, YXT.com (HK), owns 100% of Yungxuetang Information Technology (the wholly foreign-owned entity or WFOE), which was incorporated in China and is deemed an onshore company. As foreign-owned entities are not permitted to own Chinese companies, the VIE structure is used. A series of contractual arrangements between Yungxuetang Information Technology and Jiangsu Radnova Intelligence Technology (Radnova) mean that YXT.com has 100% effective ownership of Radnova. Radnova is the main operating company in the group. Radnova in turn owns two subsidiaries that cover operations in Beijing and Suzhou.

Strategy

The company has a three-year strategy in place to drive the growth of the business, covering the following:

• Define, design and implement the Intelligent Productivity strategy: starting with sales productivity enablement before expanding to other functions.
• International expansion: proactively explore international business, enter the mainstream global AI application market and compete internationally.
• Expand global human resources (HR) technology reach and ecosystem: through partnerships, investments and M&A.
• Reform the training service ecosystem with AI: create new ways to generate, curate, orchestrate and distribute training content.
• Continue to transform internal processes and organisation with AI to boost productivity.

In the video below, we speak with the company’s chief growth officer, Alan Wang, to discuss the company’s background, market positioning and growth strategy.

Executive interview with Alan Wang, chief growth officer at YXT.com

Edison Investment Research

Management

The company was founded by Xiaoyan (Peter) Lu, Teng Zu and Jie Ding. Peter Lu is the chair and CEO and holds 67.5% of the company’s voting power. In addition to Peter, the board consists of Jie Ding, Guodian Huang (independent director) and Yunjian Ling (independent director). The company’s management team consists of chair and CEO Peter Lu, Jie Ding, CTO Haihua Huang and CFO Shen Cao, and each of the major product lines is headed up by a manager. We provide more detailed biographies at the back of the report.

Share structure

YXT.com has two classes of shares: Class A and Class B.

• Class A shares: 163,294,773 in issue with 1,769,610 held in treasury. Each share entitles holders to one vote.
• Class B shares: 16,931,824 in issue. Each share entitles holders to 20 votes per share. These are all held by the company’s chair/CEO.

AI-enabled product suite

YXT.com operates in the corporate training and development market. It has developed a digital learning management system, with associated knowledge management and talent management systems, which it sells in conjunction with training content. It has also developed a specialised sales training solution. Over the last four years, it has evolved its product suite to make use of AI tools, which support better personalisation and more efficient production of training content.

Core product suite

The diagram below shows the company’s product suite as at the time of its IPO in August 2024, which consisted of a ‘SaaS software plus content’ offering. This is essentially a platform that integrates with other IT systems, such as human capital management (HCM), enterprise resource planning (ERP) or customer relationship management (CRM), and supports applications for knowledge management, learning management, talent management, learning experience and training tools. On top of this, YXT.com provides content, either developed in house or provided by third parties, and in online or in-person format.

Exhibit 2: Software plus content offering

Source: YXT.com

At the end of FY25, the company had 2,382 customers, of which 2,301 were subscribers to YXT.com’s software, and estimates it has a user base of more than 20 million. FY25 revenue was made up of RMB337.7m/$48.3m from corporate learning solutions (94% subscription revenue, 6% from offline courses and courseware recording services) and RMB2.5m from Other, which consists of customised software and related maintenance services.

Evolving the suite to exploit AI

In 2022, the company put in place its ‘AI First’ strategy and started transforming the business to become a supplier of intelligent productivity solutions. It has incorporated the use of AI tools across six functions and 22 processes (for example, creating assistants to help with content marketing, account management or legal contracts).

The company has provided examples of how the use of AI internally is improving efficiency.

• Customer service: in 2023, the company started to develop service bots, which resulted in customer service headcount reducing from 80 to six.
• Customer acquisition: in 2024, the company switched from SEO (search engine optimisation) to GEO (generative engine optimisation) to improve its customer marketing. This resulted in a 40% increase in enquiries and sales leads while reducing spend by 70%.
• Sales training: in 2025, the company adopted the use of SaleSmart internally, improving the sales win rate by 150%.
• Custom software: this year, the company has started to use AI to help with coding for customer-specific projects.

The company is also redesigning its software to make use of AI tools; we highlight below some examples of how AI has been incorporated into the solutions:

• TalentNova: uses AI to screen CVs, generate job descriptions, create and undertake role plays and create learning maps.
• NeoLearning: uses AI to personalise learning and develop training content.
• SalesSmart: uses AI to undertake administrative processes, extract examples of successful behaviour and create personalised guidance.
• AI BOX: includes an agent platform that enables an enterprise to design and commission AI agents across the business.

The company makes use of nearly all large language models (LLMs) that are authorised for legal use in China, selecting the model depending on the specific task. Exhibit 3 below shows the new Intelligent Productivity offering, which currently comprises four solutions but could in time be expanded to incorporate new business function-specific solutions.

AI BOX: Powering AI agents

At the heart of the company’s product suite is AI BOX. Based on years of accumulated data on how staff perform essential and critical tasks, this knowledge has been used to develop AI agents for a variety of tasks.

The diagram below shows the functionality of AI BOX. AI BOX has been designed to enable domain experts to build AI agents and to achieve human/AI collaboration. Areas in which AI BOX could be used include sales, customer service, knowledge management, decision support, office automation and employee services.

TalentNova: Talent management

TalentNova is an AI-native talent management system. The bottom layer of the diagram shows the data that feeds TalentNova via integration with the customer’s HCM software. The next layer up shows the four key areas of the talent management process: talent acquisition, development, deployment and retention. Within each area, TalentNova enhances the original software functionality with AI agents where appropriate. For example, within hiring, AI agents have been created to generate job descriptions, review résumés and assess interviews. In development, agents help with extracting knowledge, developing training courses and creating learning maps. The top layer shows how the portal is designed to be accessed on a per role-basis: for management, employees and HR staff.

Exhibit 5: TalentNova functionality

Source: YXT.com

NeoLearning: Training and knowledge management

NeoLearning provides AI-enhanced training and knowledge services. At the heart of the system is the corporate’s knowledge base, which is enhanced with AI to make it easier to extract relevant information. NeoLearning includes a ’multilateral trading platform’ providing access to external training resources and lecturers as well as using AI to help enterprises develop their own training resources. This is all used to create an AI-native learning experience that provides personalised and adaptive intelligent training.

Exhibit 6: NeoLearning functionality

Source: YXT.com

SaleSmart: Specialist training and development for the sales function

YXT.com has developed SaleSmart as a specific training application for the sales industry. SaleSmart integrates with CRM and other software used by sales teams. The application offers training for both individual team members (SalesCoach) and the team manager (SalesCommander). SalesCoach is able to analyse a team member’s communications with customers and offer personalised guidance for improvement. It analyses the behaviour of the individual members of the sales team and identifies examples of successful behaviour. It also offers an AI assistant to undertake sales-related administrative tasks. SalesCommander enables a team manager to assess staff competency, monitor the sales pipeline and manage team key performance indicators (KPIs) and objectives and key results (OKRs).

Exhibit 7: SaleSmart functionality

Source: YXT.com

Of the four products, the company is seeing the most demand for TalentNova and expects it will be the main contributor to revenue over the medium term. The next most popular product is NeoLearning, followed by SaleSmart.

While the company is currently focusing its resources on the larger end of the enterprise market, it sees opportunities to serve the small and medium enterprise markets more efficiently than before. The company is considering developing a ’lite’ version of TalentNova. For NeoLearning, the use of AI has made it cheaper and simpler to implement and support the software. This opens up the longer tail of smaller and medium-sized companies as potential customers for both products via the partner channel.

Corporate learning and development market

Market research firm Frost & Sullivan estimated that the Chinese corporate learning market was worth RMB641.6bn (c $90.7bn) in 2023. The market includes the products, services and content required to develop employees’ skills, knowledge and behaviours to meet business goals. This includes a combination of learning content, such as course libraries and learning services, such as in-person or online training. Digital platforms have been developed to enable training to be conducted through digital methods, including online training and online merge offline (OMO). Frost & Sullivan estimated that just under 20% of the corporate learning market was digital in 2023, generating revenue of RMB126bn (c $17.8bn).

YXT.com’s digital corporate learning infrastructure incorporates three main systems: learning management, knowledge management and talent management.

Learning management systems

Learning management systems (LMSs) cover the corporate and academic markets; YXT.com is solely focused on the corporate market. A corporate learning management system ensures that the right employees get the right training at the right time, and provides proof that it happened. The digital transformation of corporate learning enables consistent, scalable training and helps employers meet their compliance and risk requirements. A lower reliance on in-person training reduces the cost of delivery and increases flexibility. As AI tools are incorporated into LMSs, employees can access more personalised training at an even lower cost of delivery.

An LMS would typically include the following functionality: user and organisational management; learning content and programmes; learning delivery and experience; assessments, certifications and compliance; tracking, analytics and business reporting; manager and talent enablement; communication and engagement; administration, security and governance; and integrations with enterprise systems.

Knowledge management systems

A knowledge management system (KMS) is designed to make it easier to find and reuse internal knowledge, providing a single source of truth. As well as being used to provide information for training, such as corporate policies, it also manages information for other purposes, for example product documentation and standard operating procedures.

Talent management systems

A talent management system (TMS) is designed to manage employee lifecycle and development decisions. It usually includes functionality such as performance reviews, continuous feedback, goal setting and OKRs, career paths, succession planning and talent analytics. Depending on the functionality provided by HCM systems, it may also include recruiting and onboarding workflows and compensation planning.

There are overlaps between all three systems. For example, an LMS will often access data from a KMS. And a TMS can link with an LMS so that an employee can undertake the training suggested in their career development plan.

Competition mainly from domestic Chinese companies

There is an active digital corporate learning market in the west, including companies such as Docebo, Workday (including recently acquired AI-native Sana), SAP (SAP SuccessFactors), Oracle (Oracle Learning) and Cornerstone OnDemand. However, these companies tend not to design software specifically for the Chinese market and are not cost competitive with domestic suppliers.

In China, domestic technology companies have developed software solutions specifically for the Chinese market. At the large end of the scale, Tencent, Alibaba and ByteDance have created enterprise workplace platforms (WeCom, DingTalk and Lark, respectively) that combine chat, meetings, documents, workflows and administration in one place. These would be roughly equivalent to Microsoft Teams with Microsoft 365 or Google Workspace. Domestic SaaS software companies have developed software in areas such as HCM, ERP, customer service and finance, and these are usually designed to integrate with the large enterprise workplace platforms.

Within China, there are a number of domestic players who have developed SaaS software tailored specifically to the Chinese corporate learning market. As most of them are privately owned, there is limited publicly available information.

• SME-focused: 1) Beisen (HK: 9669) is primarily an HCM software provider, but acquired Chinese e-learning provider Cool College (Kuxueyuan) in January 2025. Cool College is more focused on the SME market with a lower-cost solution, relying on integrations with DingTalk, Lark and WeCom. When acquired, Cool College had almost 4,000 paying clients, 20 million users and a top ranking in DingTalk Mall’s training product category. 2) Moxueyuan (privately owned).
• Tool-centric: UMU (privately owned) was founded in 2015. It now has more than 1,000 paying customers and provides basic functionality.
• Content-focused providers: Liangzi and Shidai Guanghua (both privately owned).
• Custom-built solutions: Zhixueyan (privately owned) has developed a low-code platform-as-a-service. It claims more than 2,500 clients and 30 million users.

To a more limited extent, Radnova will also compete with digital content providers, software developers and traditional offline service providers expanding into digital training. Radnova believes that it has the most comprehensive solution and is the only provider built to serve large enterprises.

Market size and growth potential

China has been an engine of growth for several decades; GDP growth has moderated in recent years to mid-single digits and market commentators forecast it to be in the 4.0–4.5% range in 2026. At the same time, after years of growth, the size of the population has started to decline (Exhibit 9), and the ratio of workers to pensioners is decreasing, with a shrinking pool of working age people supporting a growing pensioner cohort. The UN has forecast the population will decline to 1.265bn by 2050 (down 11% from 2024) and 639m by 2100 (down 55% from 2024), and although the precise numbers may be contested, there is broad agreement in the decline. In response, the government enacted pension reforms from 1 January 2025, raising the pension age for men from 60 to 63 and for women from 55/50 to 58/55 over a 15-year period.

We see the following drivers of demand for digital corporate learning:

• The shrinking pool of workers makes it important to maximise the skills and productivity of each employee.
• A later retirement age means some manual workers will need to move away from physical work and will need retraining.
• The pace of AI and robotics adoption means employees will need to upskill and potentially shift roles.

The charts below show Frost & Sullivan market forecasts for corporate learning and digital corporate learning. The second chart splits out revenue by size of enterprise, with YXT.com focused on the large and mega enterprises.

In terms of the number of companies that YXT.com can target, Frost & Sullivan estimated in mid-2024 that there were 34.4m enterprises in China, with 120k classed as large or mega, 730k as medium-sized and 33.6m as small. Of the large enterprises, around 70% had already adopted some form of corporate learning, whereas adoption was at only 36% for medium-sized enterprises and 14% for small enterprises. More recent data sourced from the Chinese State Administration for Market Regulation (SAMR) estimated that the total number of registered enterprises was 61.2m at the end of 2024. We would assume a similar split by size, which would imply a market of 213k large enterprises to target. The company estimates the following target markets for its products:

• TalentNova: the company estimates that around 6,500 large enterprises have well-established HCM systems that could benefit from AI-enabled talent development.
• SalesSmart: the company estimates that at least 100,000 large enterprise customers could benefit from sales enablement training. It estimates that around 1,500 of its subscribers have digital sales management that could be targeted with this solution.
• AI BOX: the company estimates that there is demand from around 200,000 enterprises for intelligent transformation, of which c 38,000 are likely to engage in external procurement of products and services. The company is currently targeting c 800 of its subscribers with this solution.

YXT.com’s sales and marketing approach

YXT.com’s customer base consists of Chinese domestic companies and the Chinese subsidiaries of global multinational companies. The charts below show a selection of customers.

Exhibit 12: Selected customers by industry

Source: YXT.com

Exhibit 13: Sample of non-Chinese multinational customers

Source: YXT.com

Increased focus on large enterprises

The company used to sell to the entire enterprise market, including large enterprises (1,000–10,000 employees), medium-sized enterprises (300 to 1,000 employees) and small enterprises (less than 300 employees), but in 2024, the company decided to focus on the large enterprise market as these customers are more likely to have the budget to invest consistently in corporate learning solutions. This has partly contributed to the decline in the number of subscribers and has weighed on net revenue retention, but this process is largely complete.

The focus to date has been on selling in China, and, through a follow-my-customer approach, providing its software outside of China to international subsidiaries of Chinese customers. Substantially all revenue in FY25 was generated in China. YXT.com’s software is available in 19 languages to cater to its multinational customer base.

Regional direct sales approach plus channel partners

The company has a mainly direct sales approach but also has partnerships with regional Chinese IT service providers and software companies in adjacent areas. The direct sales force consists of 190 employees regionally focused within China. YXT.com’s software can be accessed via third-party platforms such as WeCom, Ali DingTalk and Lark. The company expects to use the direct sales force to target large enterprise customers and to use channel partners to access small and medium enterprises.

The four products can be sold independently. AI Box is only likely to be sold standalone if a customer is looking to build their own AI applications, otherwise it would be included with the other applications and would serve as the AI engine supporting each product’s capabilities. The sales force targets HR and training departments when selling TalentNova and NeoLearning, whereas SaleSmart is sold to corporate sales departments. Standalone AI Box sales are targeted at IT and digital transformation departments. To encourage adoption of its AI-enabled products, the company offers bundled solutions (typically TalentNova and Neo Learning) as well as discounts and volume rebates.

The company’s account managers and customer success managers identify opportunities to upsell and cross-sell to the existing customer base. This includes selling to additional subsidiaries and selling additional learning content. In some cases, customers act as referral partners to their vendors or customers.

Sensitivities

The company’s financial performance and share price are sensitive to a number of factors. We divide these into two groups: company-specific and those risks associated with investing in a Chinese-based company.

Company-specific risks

• Adoption of AI technology: the pace at which existing and new customers will adopt Radnova’s AI-enabled technology is uncertain.
• Reversal of subscriber declines: our forecasts assume that subscriber declines reverse in FY26 and that subscriber numbers continue to grow thereafter. The more frequent provision of subscriber data as a key performance indicator would provide investors with confidence that progress is being made.
• Timing of cash breakeven: YXT.com has been loss-making for several years and in FY25, moved from a net cash to a net debt position. While management is targeting operating cash flow breakeven in FY26 and expects to use debt funding until it is cash generative, there is the risk that the break-even point may take longer to reach than planned.
• Geographic expansion: the company has had success selling its technology to the domestic Chinese market. Its ambitions to grow outside of China may prove more difficult to execute than expected.
• Dual-class share structure: as explained in the corporate governance section, while holding only 9.4% of shares outstanding (Class A and Class B combined), the chair and CEO holds 67.5% of voting rights and therefore has the ability to control any shareholder vote that requires a majority.
• Holding company not incorporated in the US: the holding company that is listed is incorporated in the Cayman Islands, which has different corporate governance rules compared to Nasdaq. This includes a) no requirement that the majority of the board is independent, b) no requirement for compensation or nominations committees to consist entirely of independent directors, and c) no requirement to have regularly scheduled executive sessions with only independent directors each year.
• Delisting risk: YXT.com has been trading below $1 since early January and on 28 January, received a letter from Nasdaq confirming that it has until 27 July to regain compliance with the minimum bid requirement of $1. If not compliant at this date, the stock could be delisted.

China-specific risks

• VIE structure: the operating company, Radnova, is not owned directly by YXT.com, the listed entity. Instead, it uses the VIE structure, which relies on contractual arrangements between Radnova and YXT.com (HK). There is the risk that the VIE may not meet its obligations per the contracts or there could be conflicts of interest between VIE shareholders and YXT.com shareholders.
• Licence requirements: the Chinese government requires licences for a variety of reasons. Where relevant, Radnova has obtained licences. In other cases, where it is not obvious that a licence is required, Radnova has not obtained a licence. There is a risk that the Chinese government may decide that Radnova is operating in breach of licence requirements.
• Currency controls: the Chinese government has restrictions on the amount of currency that can be transferred out of the country. This is likely to prevent Radnova from paying a dividend to YXT.com (HK), which in turn would not be able to pay a dividend to YXT.com.
• Access to Chinese auditors: for the stock to remain listed in the US, the SEC requires access to audit papers from the audit of the operating company in China. As these can currently be made available in Hong Kong, YXT.com meets this requirement.
• US/China geopolitics: the recent tariffs imposed by the US government are an example of trade friction between the two countries.
• Currency risk: YXT.com generates and reports its financial results in renminbi, while it has a US dollar listing, which creates currency translation risk.
• State intervention: if the Chinese government decides an industry or technology is deemed to be strategic, it may lead to forced divestitures or involvement in the business that is not in the interests of shareholders.

Financials

Business model

The company sells access to its cloud-based software on a subscription basis. Contracts run from five months to three years, with one-year contracts the most common. The subscription fee is based on the number of users and the functionality selected. The company reports revenue across three lines as ‘Corporate learning solution: subscription’, ‘Corporate learning solution: non-subscription’ and ‘Other’.

Cost of sales includes the cost of producing and delivering content as well as the costs of cloud platforms to host the software and tokens for using third-party AI models. The majority of cloud hosting is in China across Tencent Cloud, Alibaba Cloud, Huawei Cloud and Baidu Cloud, with AWS used outside of China. As the company increases its focus on AI-enabled content production, we would expect the cost of content production to decline.

The company reports operating costs across R&D, sales & marketing (S&M), and general & administrative (G&A). The company does not capitalise the costs of developing its software.

Performance and forecasts

The company has published a relatively limited history of financial performance. In the table above we summarise financial performance since FY22 and our forecasts to FY28.

Historical performance masked by CEIBS PG accounting

Results for FY24 reflect the de-consolidation of China Europe International Business School Publishing Group (CEIBS PG). In 2020, the company acquired 100% of two businesses, CEIBS Management Ltd (ManCo) and Digital B-school China Limited (Digital B), each of which held a stake in CEIBS PG. ManCo held 21% of the common shares of CEIBS PG and Digital B held 39% of the preferred shares. As a result of effectively holding 60% of CEIBS PG, YXT.com consolidated the business from the date of acquisition. After a period of litigation brought by CEIBS, the holder of the remaining 40% of CEIBS PG, an arbitration court held that the transfer of the 21% held by ManCo was invalid. From mid-January 2024, CEIBS PG was de-consolidated and the 39% of CEIBS PG was recorded as a debt investment worth RMB4.976m. The company reported a gain on de-consolidation of RMB78.76m in FY24. In late 2025, CEIBS was successful in its bid to have CEIBS PG wound up. We understand that the company is pursuing litigation to recoup the losses incurred from these transactions.

The reported results therefore include CEIBS PG for FY22, FY23 and one month of FY24. In the table above, we also show pro forma data where available (only provided for revenue and subscribers), which treats the business as never having consolidated CEIBS PG.

Revenue growth despite declining subscriber numbers

Pro forma subscriber numbers declined by 4% in FY24 and FY25, reflecting the transition away from small and medium-sized companies. Despite this, pro forma revenue increased 1% in FY24 and 4% in FY25, which implies growing average revenue per subscriber.

Subscription revenue for FY25 of RMB317.4m was 5% higher y-o-y, with RMB144.7m in H125 and RMB172.7m in H225. Subscription revenue made up 93% of total revenue, up from 91% in FY24. On lower average subscribers in H225 the company reported higher subscription revenue. We estimate that revenue/average subscriber in H225 was RMB74k, up 21% from RMB61k in H125 and H224. NRR for the year was 101.4%, compared to 100.9% in FY24 and 100.3% in H125.

The company noted that monthly recurring revenue (MRR) from AI-related product was RMB1.1m in December 2025 compared to RMB0.3m in December 2024 and RMB0.5m in June 2025. Based on H225 subscription revenue divided by six, AI-related MRR made up nearly 4% of subscription revenue exiting the year. To date, AI-related product sales have mostly come from the existing customer base.

Gross margin expansion

YXT.com reported a gross margin of 68.3% in FY25 (H1 65.1%, H2 71.0%), up from 61.8% in FY24. The main reasons for this were the use of AI internally to reduce costs, lower headcount and third-party infrastructure costs, and lower instructor costs due the reduction in offline solutions.

Lower operating costs

Overall, operating costs (net of other operating income) reduced 4.6% y-o-y to RMB373.6m.

• R&D: costs reduced 4% y-o-y, as headcount was reduced due to higher AI-related automation.
• S&M: costs were essentially flat y-o-y. Costs were materially higher in H225 to support the launch of the new AI-enabled productivity suite.
• G&A: costs reduced 12% y-o-y, mainly because FY24 included IPO-related costs.

The combination of slightly higher revenue, higher gross margins, and lower operating expenses reduced the operating loss from RMB187.1m in FY24 to RMB141.0m in FY25.

Net finance costs

Net finance costs of RMB17.3m comprised interest income of RMB3.6m, interest expense of RMB6.6m, investment losses of RMB14.8m (on the available for sale debt securities), and an fx gain of RMB0.4m. The 25% investment in Kangshengji made in H125 generated an equity method investee loss of RMB0.5m.

The standard corporate tax rate is 25% in China, reduced to 15% if a company qualifies as a ‘High and New Technology Enterprise (HNTE)’. Both Yungxuetang Information Technology (Jiangsu) and Jiangsu Radnova Intelligence Technology have HNTE status so are taxed at the 15% rate. The Cayman-registered top company, YXT.com, has a tax rate of 0%. The company has been loss-making in recent years and therefore has not paid any tax. It has tax losses carried forward worth RMB2,258m, but these will not be recognised until the company has reasonable visibility of profitability.

Balance sheet and cash flow

At the end of FY25, the company had cash and cash equivalents of RMB115.0m, short-term investments of RMB19.7m, short-term debt of RMB139.5m and long-term debt of RMB8.0m, resulting in a net debt position of RMB12.8m. After leases of RMB22.9m, net debt was RMB35.7m.

Debt at the end of FY25 consisted of secured and unsecured loans with interest rates ranging from 3.0% to 3.55%. In FY25, the company made use of short-term discounted bank acceptance notes, which incur lower interest at rates (0.96–1.26% per year); these are included in ‘Other payable and accrued liabilities’ on the balance sheet.

The company holds minority stakes in a number of companies in the form of preference shares, which are treated as available-for-sale debt securities and included as long-term investments on the balance sheet. We understand that these investments are in companies in related areas of HCM software, and the company aims to use partners’ platforms as routes to market for YXT.com’s technology. At the end of FY25, long-term investments had a carrying value of RMB104.3m. The remaining stake in CEIBS PG is included in this balance.

Relating to CEIBS PG, the company carries a provision for litigation costs of RMB17.756m and still owes RMB14.775m for the original acquisition. The company is currently pursuing litigation to recover these losses.

Capex is minimal; there is no capitalisation of development costs and the company spends a small amount on IT equipment. All premises are leased. For working capital, as the subscription fee is typically received annually, there is usually some element of deferred revenue.

We note that the company has not yet published a cash flow statement or detailed notes to the accounts (the 20F is expected to be filed in April), so our FY25 cash flow statement is still estimated as are depreciation and amortisation charges.

Forecasts assume subscriber growth and uplift from AI

We construct our revenue model primarily by forecasting subscriber numbers and average revenue per subscriber. We assume a 5.2% increase in subscribers in FY26, 7.2% in FY27 and 6.0% in FY28. Reflecting the potential for upselling and cross-selling, we assume that the average revenue per subscriber increases by 2.9% in FY26, 2.4% in FY27 and 1.4% in FY28. We assume that the gross margin continues to improve, with all the improvement from subscription revenue, helped by scale and the increasing use of AI. The company already started to reduce operating costs and expects to continue to do this in FY26, mainly through increased automation due to AI. In particular, we factor in a large decrease in G&A costs reflecting reduced headcount and the non-recurrence of professional advisor fees incurred in FY25. In FY27 and FY28, we factor in minimal growth in the cost base. We forecast the company to reach positive EBITDA and operating profit in FY28.

Management is aiming to reach operating cash flow breakeven in FY26. We assume this is through a combination of reduced operating losses and a higher working capital contribution (mostly due to increased deferred revenue as a proportion of subscription revenue is paid in advance). Our operating cash flow forecast is slightly more conservative, although does assume positive operating cash flow in H226. We forecast minimal capex over FY26–28.

We note that as the company only reports twice a year, we have limited visibility for our forecasts.

Valuation

During the forecast period, we expect the company to be loss-making at the net income level on a reported basis . We therefore focus on EV/Sales and EV/EBITDA multiples. From a peer group perspective, we look at valuation multiples for four groups of peers:

1. US-listed learning and development software companies operating in Europe and North America.
2. US-listed Chinese software and IT companies.
3. Software companies listed on the Shanghai stock exchange (SSE) STAR segment.
4. Software companies listed on the Hong Kong stock exchange. This includes Beisen, who recently acquired Cool College.

The first group is well-established, with solid revenue growth boosted by acquisitions, and profitable. Docebo is a pure-play peer and is expected to grow its EBITDA margins from 18% in FY25 to 20% in FY26e. This would be a more suitable comparator for YXT.com than Workday, which has a large HCM software business and is already generating margins above 30%.

The second group is very limited, as most Nasdaq-listed Chinese software companies do not have analyst coverage, and therefore provides less meaningful comparison.

The third and fourth groups contain many software companies that are well-established in China. The SSE STAR-listed companies are growing revenue at a median rate of 18.2% in FY26 and 14.4% in FY27, with median EBITDA margins of 15.0% and 14.9% respectively (this is less reliable as fewer forecasts are available for profitability). The Hong Kong-listed companies are growing revenue at a median rate of 14.5% in FY26 and 13.3% in FY27, with median EBITDA margins of 9.7% and 11.6% respectively. The SSE STAR-listed multiples are significantly higher than for the Hong Kong listed stocks, with median EV/Sales multiples of 6.0x FY26 and 5.2x FY27, compared to 1.3x and 1.0x for Hong Kong-listed stocks. Similarly, median EV/EBITDA multiples are 41.6x for FY26 and 36.3x for FY27 compared to 13.4x and 10.9x for Hong Kong-listed stocks.

We would expect YXT.com to trade at a discount to higher growth, higher margin peers while it is still loss-making. As operating losses are reduced and the company approaches cash breakeven, we would expect this discount to reduce.

We also have performed a discounted cash flow analysis. We model the cash flows in RMB and use the Chinese risk-free rate of 1.8%, a country risk premium of 5.1% (source: Damodaran, Jan 2026) and a long-term growth rate of 2%. To incorporate several risks, including the voting power of the CEO, China investment risks and size of the company relative to peers, we use a beta of 2. This compares to an average beta of 1.5 for SSE STAR and HKEX-listed software companies. After factoring in the small net debt position, this results in a WACC of 11.6%. We use our forecasts to FY28 and thereafter revenue growth trending down to the long-term growth rate (average 5% FY29-35), and an EBITDA margin of 8% in FY29 growing to 18% by FY35. This generates a per ADS valuation of $1.40, 301% above the current ADS price. We have included the Class B shares in the total share count when calculating the value per ADS.

In Exhibit 17, we show the results of varying the WACC and long-term growth rate. In Exhibit 18, we show the results of varying the revenue growth rate and EBITDA margin (the table references FY29 assumptions).

The base case valuation equates to an EV/Sales multiple of 1.7x FY26, 1.5x FY27 and 1.4x FY28.

We have also performed a reverse DCF that attempts to calculate the assumptions inherent in the current ADS price. Factoring in revenue growth of 3.5% per annum from FY26-35 and EBITDA margin expansion from -22% in FY26 to 12% in FY35 (average 1.5% FY26-35 vs 9.4% in the base case scenario) results in an ADS price of $0.35; these assumptions are considerably more conservative than our forecasts.

Recent M&A

• January 2025: Beisen acquired Cool College for RMB180m/$24.6m. According to press comments at the time of the acquisition, Cool College had been valued at a peak of RMB1bn in previous funding rounds.
• November 2025: Workday acquired Sana, a provider of AI-based learning and agents, for $1.1bn. Sana generated revenue of $52.1m in 2024, more than doubling y-o-y. The price paid equates to a trailing price/sales multiple of c 21x.
• January 2026: Docebo acquired 365Talents, an AI-powered skills intelligence and workforce analytics company based in France. Docebo paid cash consideration of $54.6m and expects 365Talents to contribute revenue of $9m from the date of completion to the end of 2026. This equates to a forward price/sales multiple of c 5.7x.

The Docebo and Workday acquisitions show the high value placed on AI functionality.