China software conference takeaways: AI monetization accelerates, and growth visibility becomes the core stock-picking theme

Overview

This report summarizes Bank of America’s software company management meetings and expert exchanges during the China conference in Shenzhen in May 2026, as well as its Shanghai AI/technology research trip. Covered themes include AI applications, LLMs, AI cloud, ERP, photo and design tools, real-time engagement, IT outsourcing, property SaaS, cybersecurity and real-time data infrastructure. The core conclusion is that AI monetization in China’s software sector is accelerating, but demand recovery and earnings recovery are structurally diverging.

Core views

First, AI is driving software service models from SaaS toward RaaS, with pricing moving from seat-based subscriptions to usage and token-based billing. Second, ERP and photo/design tools have stronger growth visibility, supported by AI product revenue, overseas subscriptions and recurring revenue. Third, IT outsourcing, cybersecurity and property software remain pressured by client budgets, property-sector weakness and competitive pressure. Fourth, demand from large SOEs, key infrastructure industries and large private enterprises is relatively stable, while government-related demand has not yet recovered. Fifth, AI-driven efficiency gains, prudent headcount expansion and cost control should support FY26E cash flow and margin improvement.

Analysis framework

The report compares demand, business models, profitability, competitive landscape and valuation risk-reward across subsectors based on management discussions with more than 10 software companies during Bank of America’s China conference, as well as meetings with AI agent and AI model experts and a meeting with UCloud management.

Methodology notes

• Business model analysisSaaS to RaaS migration

  Shift from seat-based subscriptions to outcome-, usage- and token-based billing

  The report views AI-driven changes in software service models as the core monetization upgrade, arguing that token-based and usage-based billing can raise ARPU, but may also lower switching costs for customers and intensify competition among model vendors.

• Valuation methodP/S, P/E and SOTP

  Choose valuation multiples based on each company’s profitability stage and business structure

  Agora, Kingdee, Ming Yuan Cloud and Qi An Xin are valued using P/S or SOTP frameworks; Chinasoft and Meitu rely more on P/E frameworks to reflect profitability and growth prospects.

• Demand segmentationCustomer type and vertical industry split

  Assess demand by SOE, private enterprise, government-related customers and key infrastructure industries

  The report argues that demand is more stable in SOEs and key infrastructure industries, that government-related customers remain weak, and that property-related demand is still under pressure.

Asset mapping & comparison

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

• Kingdee International Software

  Preferred Buy name and beneficiary of ERP and AI-native products

  Strengths

  1Q26 revenue growth accelerated; AI-native product contract value was RMB230mn; FY26 AI-native revenue target exceeds RMB1bn; cash flow and margin targets are improving.

  Weaknesses

  Large-enterprise NDR fell from 110% in 4Q25 to 103% in 1Q26, and recovery to above 110% in 4Q26 still needs to be validated.

  Comparison

  Compared with traditional IT outsourcing and property software, ERP demand and AI product visibility are stronger.

  Risks

  Intensifying SaaS competition, cloud service or subscription revenue growth slower than expected, and weak market sentiment weighing on valuation.

• Meitu Inc.

  Preferred Buy name and beneficiary of AI monetization in photo and design tools

  Strengths

  Overseas subscriptions and AI features in productivity tools are driving ARPU improvement; the long-term GPM target is about 70%; 1Q26 product revenue and subscription user growth are easing concerns about AI disruption.

  Weaknesses

  AI credits consumption is increasing, requiring continued control of third-party API costs.

  Comparison

  Compared with most Chinese SaaS companies, Meitu has higher profitability and a larger share of overseas revenue, making it better suited to global SaaS profitability comparisons.

  Risks

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

• Agora

  Buy-rated name and beneficiary of real-time engagement and conversational AI revenue

  Strengths

  No. 1 in China RTE market revenue, with 2025 revenue about 1.5x that of Tencent, the second-largest player; conversational AI is expected to contribute about 5% of total revenue in 4Q26; the competitive landscape may ease.

  Weaknesses

  Profitability is still recovering, with operating profit breakeven targeted only in 4Q26.

  Comparison

  It has a scale lead in the RTE market, but AI revenue contribution is still in the ramp-up phase.

  Risks

  Regulation, antitrust, data and cybersecurity, political tensions, intensifying competition, customer self-development and weak macro demand.

• Chinasoft International

  Neutral-rated name; AI service growth is dragged down by traditional IT outsourcing

  Strengths

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

  Weaknesses

  Weak traditional IT outsourcing demand, customer price cuts, workforce optimization and competition are pressuring margins, and FY26 revenue is expected to grow in the low single digits.

  Comparison

  Compared with ERP and design tools, the company has weaker growth visibility and relies more on efficiency improvement and new business contributions.

  Risks

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

• Ming Yuan Cloud

  Neutral-rated name; property SaaS pressure is easing but recovery is still early

  Strengths

  CRM SaaS remains the core business, while AI-related products are raising revenue per project; cost control and AI-driven efficiency gains support profitability improvement.

  Weaknesses

  Weak property sales continue to pressure revenue, new project activity remains soft, and outcome-based monetization and overseas expansion are still at an early stage.

  Comparison

  Compared with general SaaS companies, it has greater exposure to the property sector, so it deserves a valuation discount.

  Risks

  Residential property recovery weaker than expected, adverse regulation or further property downturn, and industrial/infrastructure expansion slower than expected.

• Qi An Xin

  Underperform-rated name; cybersecurity demand recovery is lagging

  Strengths

  FY26 target includes faster revenue growth and OCF breakeven; the AI-related business already has a separate subsidiary, with FY25 revenue of about RMB200-300mn.

  Weaknesses

  Incremental AI cybersecurity budgets for key infrastructure enterprises in 2026 are limited, and credit impairment losses of RMB400-500mn per year may still occur in FY26-27.

  Comparison

  Compared with AI applications and ERP, AI budget deployment in cybersecurity is more likely to materialize in 2027, leaving near-term upside weaker.

  Risks

  China cybersecurity demand recovery weaker than expected, intensifying competition, end-customer budgets below expectations, continued operating cash flow pressure or delayed breakeven.

• UCloud

  Not covered; beneficiary of AI cloud and inference demand

  Strengths

  AI inference demand is strengthening, GPU AI cloud service prices have risen by an average of 20-25%, and the business has achieved breakeven since 4Q25.

  Weaknesses

  AI server costs are rising, and expanding AI cloud and data center businesses requires capital and execution capabilities.

  Comparison

  It is more exposed to AI infrastructure demand rather than application-software ARPU expansion.

  Risks

  Server costs, supply-demand volatility, data center expansion progress and price competition.

• Xunce

  Not covered; beneficiary of AI real-time data infrastructure

  Strengths

  Its real-time data infrastructure offers speed, accuracy and scalability; token-based ARR rose from RMB60mn in January 2026 to more than RMB200mn in April.

  Weaknesses

  Entering new verticals usually requires an investment period of 1-3 years, and early-stage GPM is low.

  Comparison

  Compared with traditional project-based software, token-based revenue is growing faster, but the business still needs to prove cross-vertical replication capability.

  Risks

  Longer new-vertical expansion cycles, customer customization complexity, and gross margin pressure before scale is achieved.

Key data

• Conference coverage10+ software companiesBank of America held management meetings with multiple software companies during the week of May 11, 2026, Shenzhen China conference, and also held meetings with AI agent and AI model experts.
• Kingdee 1Q26 AI-native product contract valueRMB230mnManagement expects FY26 AI-native product revenue to exceed RMB1bn.
• Kingdee FY26E targetsdouble-digit revenue growth, adjusted net margin above 7%, OCF growth above 20% YoYLarge-enterprise NDR in 1Q26 was 103%, and management expects it to exceed 110% in 4Q26.
• Meitu long-term gross margin targetabout 70%Management believes third-party API costs are controllable and continues migrating high-usage functions to self-developed or fine-tuned vertical models.
• Agora conversational AI revenue contributionabout 5% of total revenue in 4Q26Management expects use cases such as customer service, companion toys and robots to drive higher usage.
• Agora medium-term margin targetoperating profit breakeven in 4Q26, with OPM of 15% achievable in the next 2-3 yearsGrowth and operating leverage are the main drivers.
• Chinasoft FY26 AI products and services growth targetabout 70% YoY growthHowever, weak traditional IT outsourcing demand offsets part of the growth.
• Qi An Xin FY25 accounts receivableabout RMB5bnManagement warned that annual credit impairment losses of RMB400-500mn may still occur in FY26-27.
• UCloud GPU AI cloud service price increaseaverage increase of 20-25%Driven by stronger AI inference demand and rising AI server costs.
• Xunce token-based ARRfrom RMB60mn in January 2026 to more than RMB200mn in April 2026Management expects token-based revenue to account for 20% of FY26E total revenue.

Impact & implications

For investors, the sector’s main theme is shifting from a pure AI narrative to AI monetization, subscription resilience, ARPU expansion, cost optimization and profit delivery. Kingdee and Meitu are favored because of stronger growth visibility and better valuation risk-reward, while IT outsourcing, cybersecurity and property software names that rely more heavily on demand recovery, face budget uncertainty or operate in highly competitive markets will need further order growth, cash flow improvement and credit-risk improvement to re-rate.

Risks

• AI monetization proceeds more slowly than expected, or customers are not sufficiently accepting of token/usage-based billing.
• Government-related customer demand does not recover as expected, leaving IT outsourcing, cybersecurity and property software under continued pressure.
• Multi-model AI strategies lower switching costs and intensify competition among LLM and software vendors.
• Rising third-party API, AI credits and AI server costs compress gross margins.
• Macro demand, regulation, data security, geopolitics and valuation sentiment may affect related stock performance.
• Accounts receivable and credit impairment risk may weigh on earnings and cash flow for some companies.

What to watch

• Kingdee’s AI agent product launch on May 20, 2026 and FY26 AI-native revenue realization.
• Whether Kingdee’s large-enterprise NDR can recover to above 110% in 4Q26.
• Meitu’s launch of new AI productivity tools and agent team features at the June 2026 Image Festival.
• Whether Agora conversational AI usage ramps up and contributes about 5% of revenue in 4Q26.
• Whether Chinasoft’s 70% growth in AI products and services can offset the decline in traditional IT outsourcing.
• Whether Qi An Xin’s customer POCs can convert into 2027 AI cybersecurity budgets.
• Demand elasticity after UCloud’s AI cloud price increases and progress in data center expansion.
• Whether Xunce’s token-based ARR growth can continue and increase its share of FY26 revenue.