BofA: China software AI monetization is accelerating, with Kingdee and Meitu as top Buy picks

Overview

The report summarizes BofA’s discussions during the 2026 China Conference with management teams from more than 10 software companies and with agentic AI experts. The core conclusion is that AI monetization across China software is accelerating overall, with AI driving a shift from SaaS to RaaS and from seat-based subscriptions to usage/token-based pricing; however, demand divergence across subsectors is clear.

Core views

Revenue contribution from AI, ARPU uplift, and efficiency improvement are the main themes. ERP and image/design tools have relatively high growth visibility; AI cloud, real-time interaction, and real-time data infrastructure also offer incremental opportunities. In contrast, IT outsourcing, cybersecurity, and real estate software remain affected by downstream budgets, the real estate cycle, or government-client demand.

Analysis framework

The report is mainly based on cross-company comparisons using management meetings, expert meetings, FY26E operating targets, AI product revenue/contract value, ARPU and gross margin trends, as well as each company’s valuation method and target price.

Methodology notes

• Research frameworkManagement meetings and expert sessions

  Conference notes

  During the China Conference, BofA engaged with management teams from more than 10 software companies and invited agentic AI and AI model experts to discuss application progress, distilling industry demand, business model, and profitability trends from first-hand feedback.

• Business modelMigration from SaaS to RaaS

  AI-driven change in software service models

  The report argues that AI is pushing software from traditional SaaS subscriptions toward results-as-a-service, while reinforcing usage- and token-based pricing, potentially driving ARPU higher.

• Demand breakdownAssessing growth visibility by product and customer type

  Growth visibility

  Demand is stronger for ERP and image/design tools; SOEs and large private enterprises have relatively stable demand, while government-related clients, IT outsourcing, cybersecurity, and real estate software demand remain weak.

• Valuation frameworkP/S, P/E, and SOTP

  Target price methodology

  Different companies are valued using 12-month forward P/S, P/E, or SOTP, with premiums or discounts assigned based on profitability, growth, real estate exposure, or cash flow pressure.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Kingdee (268 HK)

  Top Buy-rated name and beneficiary of ERP and AI-native products.

  Strengths

  Revenue growth accelerated in 1Q26, AI-native contract value reached RMB230mn, FY26 AI revenue target exceeds RMB1bn, and OCF and margin targets are constructive.

  Weaknesses

  Large enterprise NDR fell to 103% in 1Q26, and recovery to above 110% by 4Q26 needs to be validated.

  Comparison

  Compared with other software companies, ERP demand and recurring revenue are more visible.

  Risks

  Intensifying SaaS competition, slower-than-expected growth in cloud services or subscription revenue, and valuation pressure from market sentiment.

• Meitu Inc. (1357 HK)

  Top Buy-rated name and beneficiary of AI monetization in image and design tools.

  Strengths

  ARPU upside comes from overseas subscriptions and AI-driven productivity features, the long-term GPM target is around 70%, and business-model resilience alleviates market concerns.

  Weaknesses

  Greater use of AI features will increase AI credits consumption, requiring continued control of third-party API costs.

  Comparison

  Compared with most China SaaS peers, Meitu has a higher share of overseas revenue and stronger profitability.

  Risks

  Slower-than-expected paid-user conversion, competitive pressure, regulatory and geopolitical risks, and valuation compression.

• Agora (API US)

  Buy-rated name with incremental revenue opportunities in real-time interaction and Conversational AI.

  Strengths

  Ranked first in China’s RTE market by revenue, at about 1.5x Tencent, the second place, in 2025; Conversational AI is expected to contribute about 5% of revenue in 4Q26.

  Weaknesses

  AI revenue is still ramping up, and scenario deployment and revenue mix improvement need validation.

  Comparison

  Management expects the competitive landscape in China to ease, as the third-largest player may slow customer acquisition due to financial constraints.

  Risks

  Regulation, listing status, antitrust, data and cybersecurity, political tensions, customer in-house development, and intensifying competition.

• Chinasoft International (354 HK)

  Neutral-rated name facing both AI service growth and pressure in traditional IT outsourcing.

  Strengths

  AI products and services target 70% YoY growth in FY26, and the AI business GPM can reach 40%.

  Weaknesses

  Weak traditional IT outsourcing demand, customer price cuts, industry competition, and workforce optimization weigh on margins.

  Comparison

  AI business growth is partially offset by the decline in traditional outsourcing, leaving overall FY26 revenue growth only in the low single digits.

  Risks

  Slower-than-expected revenue recovery from key customers, customer concentration risk, and intensifying competition in the IT outsourcing industry.

• Ming Yuan Cloud (909 HK)

  Neutral-rated name transitioning from real estate CRM SaaS to AI usage-based monetization.

  Strengths

  CRM SaaS covers the full process from online lead generation to offline conversion, AI product adoption is lifting average revenue per project, and cost control plus AI efficiency gains support profitability improvement.

  Weaknesses

  Weak real estate sales continue to pressure revenue, and new project activity remains soft.

  Comparison

  Compared with general SaaS, the company has higher real estate exposure, so the growth discount is more pronounced.

  Risks

  Weaker-than-expected recovery in China residential real estate, further downturn in real estate regulation or the industry, and slower-than-expected expansion into industrial and infrastructure clients.

• Qi An Xin (688561 CH)

  Underperform-rated name, with cybersecurity demand recovery still slow.

  Strengths

  FY26 targets faster revenue growth and OCF breakeven; it has established an AI-related business subsidiary, with FY25 revenue of about RMB200-300mn.

  Weaknesses

  Large enterprise IT budgets for 2026 are already set, leaving limited incremental budget for AI cybersecurity; accounts receivable are about RMB5bn, and credit impairment pressure remains.

  Comparison

  Compared with other AI software applications, cybersecurity AI budget release is more likely to be delayed until 2027.

  Risks

  Weaker-than-expected recovery in China cybersecurity demand, insufficient terminal budgets, intensifying competition, continued OCF pressure, or delayed breakeven.

• Xunce (3317 HK)

  Not covered, provider of AI real-time data infrastructure and analytics solutions.

  Strengths

  The solution emphasizes speed, accuracy, and scalability; token-based ARR rose rapidly from RMB60mn to more than RMB200mn.

  Weaknesses

  Entering new verticals usually requires a 1-3 year investment period, with relatively low early-stage GPM.

  Comparison

  The business is expanding from asset management into verticals such as smart energy, telecom, electric power, city operations, robotics data platforms, and commercial aviation.

  Risks

  Long expansion cycles in new verticals, complexity of customized delivery, and the sustainability of token revenue growth needs validation.

• UCloud

  Not covered, beneficiary of AI cloud computing demand.

  Strengths

  Management is optimistic about AI inference demand, GPU-based AI cloud services have seen price increases of 20%-25%, profitability has improved over the past few quarters, and the company has been breakeven since 4Q25.

  Weaknesses

  AI server costs are rising, and expanding data centers and AI cloud services requires capital and resource investment.

  Comparison

  It benefits more directly from AI inference compute demand rather than traditional software subscriptions.

  Risks

  GPU costs, cloud-service price competition, data-center expansion execution, and demand volatility.

Key data

• Industry overviewAI monetization is accelerating; pricing is shifting from seat-based to usage/token-basedThe report says software AI monetization is accelerating overall, with ARPU upside coming from token-based revenue models.
• Kingdee1Q26 AI-native product contract value reached RMB230mn; FY26 AI-native product revenue target exceeds RMB1bnManagement also reiterated FY26E targets of double-digit revenue growth, adjusted net margin above 7%, and OCF growth of more than 20% YoY.
• Kingdee NDR1Q26 large enterprise NDR was 103%, with a 4Q26 target above 110%Short-term NDR fell from 110% in 4Q25, but management expects recovery within the year.
• MeituLong-term GPM around 70%AI credits consumption is rising, but third-party API costs are manageable, with some high-frequency features migrating to in-house or fine-tuned vertical models.
• Meitu product catalystImage Festival in mid-June 2026More AI features with agent teams characteristics are expected to be launched for productivity tools such as DesignKit and Kaipai.
• AgoraConversational AI business is expected to contribute about 5% of total revenue in 4Q26Management is seeing gradual usage growth in scenarios such as customer service, companion toys, and robots.
• Agora profitability targetOperating profit breakeven in 4Q26; OPM target of 15% over the next 2-3 yearsThe drivers are revenue growth and operating leverage.
• ChinasoftFY26 revenue to grow low single digits YoY; AI products and services target 70% YoY growthWeak traditional IT outsourcing demand offsets AI business growth, and management expects headcount to decline by about 1k people per month.
• Ming Yuan CloudCRM SaaS remains the core, with AI driving hybrid monetization through subscriptions and usageNew real estate projects remain weak, but average revenue per project is improving due to higher adoption of AI products.
• Qi An XinFY26 targets faster revenue growth and OCF breakevenAI-related cybersecurity budgets are more likely to materialize in 2027; credit impairment of RMB400-500mn may still occur annually in FY26-27.
• XunceToken-based ARR rose from RMB60mn in January 2026 to more than RMB200mn in AprilManagement expects token-based revenue to account for 20% of total FY26E revenue.
• UCloudAverage price increase of 20%-25% for GPU-based AI cloud servicesManagement is optimistic about AI inference demand, and the company has achieved breakeven since 4Q25.

Impact & implications

For investors, the focus of the AI theme is shifting from conceptual disruption to quantifiable improvements in revenue, ARPU, gross margin, and cash flow. Companies with recurring revenue, visible AI monetization paths, and cost control capabilities are more likely to see valuation re-rating, while companies with slower budget recovery or higher exposure to real estate and government clients still need observation for a demand inflection point.

Risks

• AI monetization may progress more slowly than expected, and token- or usage-based pricing may fail to sustainably increase ARPU.
• Third-party API, computing power, and AI credits costs may rise, compressing gross margins.
• Demand recovery from government-related clients may be slow, while downstream budgets for IT outsourcing, cybersecurity, and real estate software remain weak.
• Multi-model strategies among large models reduce switching costs and may intensify competition among LLM suppliers and software application layers.
• A continued downturn in the real estate cycle may keep pressuring real estate software companies such as Ming Yuan Cloud.
• Qi An Xin’s accounts receivable and credit impairment pressure may weigh on profits and cash flow.
• Regulatory, data security, cybersecurity, listing-status, and geopolitical risks may affect valuations of related stocks.

What to watch

• Whether Kingdee’s large enterprise NDR can recover to above 110% in 4Q26.
• Whether Kingdee’s FY26 AI-native product revenue can exceed RMB1bn.
• New AI features at Meitu’s June 2026 Image Festival and changes in ARPU for productivity tools.
• Whether Agora’s Conversational AI can reach about 5% revenue contribution in 4Q26 and achieve operating profit breakeven.
• Whether Chinasoft’s 70% YoY growth target for AI products and services can offset the decline in traditional IT outsourcing.
• Whether the decline in real estate demand for Ming Yuan Cloud continues to narrow and whether AI usage-based monetization can raise revenue per project.
• Whether AI cybersecurity budgets for Qi An Xin will be released in 2027 as management expects.
• The sustainability of Xunce’s token-based ARR and its FY26 target of 20% revenue contribution from token revenue.
• UCloud’s AI inference demand, GPU cloud pricing, and progress in data center expansion.