Chinese enterprise generative AI is moving from pilot to scale, with cloud vendors and model providers set to benefit most
AI summary card
Chinese enterprise generative AI is moving from pilot to scale, with cloud vendors and model providers set to benefit most
UBS Evidence Lab's survey of 102 Chinese IT executives and data engineers shows that 78% of surveyed companies have already deployed GenAI live, and 2026 GenAI budgets are expected to grow 21.0% YoY, significantly faster than overall IT budgets.
- Chinese enterprise GenAI deployment has moved beyond the proof-of-concept stage, with 51% of respondents live in at least one function and 27% already deployed at production scale across multiple business units.
- AI Agent is seen as the next stage, with respondents expecting it to reach scale in an average of 1.7 years and to deliver about 20% productivity gains over the next two years.
- Enterprise AI spending intent is clear: average 2025 GenAI budgets are about Rmb66m, or roughly 12% of IT budgets, and 2026 GenAI budgets are expected to grow 21.0%, faster than the 5.0% growth expected for overall IT budgets.
- In cloud and model selection, performance, security, and ecosystem support matter more than cost, underpinning the pricing power of leading cloud vendors amid strong demand.
- The report sees Alibaba, Tencent, Kingsoft Cloud, and MiniMax as likely beneficiaries; the SaaS segment still faces pressure from AI substitution and the need to rebuild investor confidence.
Report interpretation
Overview
This report is based on UBS Evidence Lab's first China AI enterprise survey conducted in December 2025. The sample includes 102 Chinese IT executives and data engineers across multiple industries and is skewed toward mid-sized to large enterprises. The report's core conclusion is that Chinese companies' investment in and deployment of generative AI has moved from the pilot stage into production expansion, with enterprise budgets, management support, and localization progress jointly driving sustained AI demand growth.
Core views
The report argues that GenAI adoption among Chinese enterprises is already high, with 78% of respondents having gone live and 27% having scaled across multiple business units. AI Agent will become the next adoption wave; although the timeline to scale is expected to be longer than for GenAI overall, open-source Agent frameworks such as OpenClaw may lower the entry barrier and accelerate experimentation. Current AI use cases in China lean more toward customer experience and visual analytics, while the US is more focused on copilots, AI assistants, productivity gains, and developer tools. From an investment perspective, cloud vendors and model providers are the most direct beneficiaries, while SaaS companies still need to prove earnings resilience and valuation support.
Analysis framework
The report uses a combination of survey research and China-US comparison: on one hand, it cites UBS Evidence Lab's China AI enterprise survey to measure GenAI deployment stage, budget growth, use cases, challenges, localization, and vendor preferences; on the other hand, it cross-compares with US AI enterprise surveys, consumer AI surveys, and QuestMobile active user data to assess enterprise and consumer AI application trends.
Methodology notes
Survey of enterprise AI adoption and budgets
The survey was conducted in December 2025 and covered 102 Chinese respondents, mainly senior technology decision-makers, to assess enterprise GenAI deployment, AI Agent expectations, budget allocation, localization, and vendor preferences.
Comparison of enterprise AI use cases in China and the US
The report compares differences in AI adoption between China and the US: China leans more toward customer experience and visual analytics, while the US leans more toward productivity tools, workflow automation, and developer tools.
Valuation methods for the relevant covered companies
The report notes that Tencent and Alibaba are valued using SOTP, MiniMax using Price/ARR, and Kingsoft Cloud using EV/sales, but no specific target prices are disclosed in the summary content.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- AlibabaBeneficiary
- Strengths
- A leading domestic cloud vendor with full-stack AI capabilities; Qwen ranks highly in enterprise model usage.
- Weaknesses
- Faces macro, competition, regulation, platform execution complexity, and short-term earnings pressure from long-term investment.
- Comparison
- Alongside Tencent Cloud, it leads enterprise cloud adoption, and AliCloud has stronger penetration among SMEs.
- Risks
- Data and content regulation, macro pressure, offline retail competition, IT system disruptions, corporate governance, and key-person risks.
- TencentBeneficiary
- Strengths
- Tencent Cloud and Tencent Hunyuan have relatively high awareness and usage among surveyed companies, and cloud adoption is more balanced across company sizes.
- Weaknesses
- Execution on new businesses and integration of invested companies remain uncertain.
- Comparison
- It leads enterprise cloud adoption together with AliCloud, and Hunyuan's model-layer usage rate of 37% is higher than DeepSeek and Qwen.
- Risks
- Intensifying competition, execution on new businesses, rising traffic and promotion costs, IT system maintenance, overseas expansion, intellectual property, and regulatory risks.
- Kingsoft CloudBeneficiary
- Strengths
- AI workload revenue accounts for more than 40% of total revenue, giving it relatively high leverage to rising AI demand.
- Weaknesses
- Revenue growth, market share, and profitability balance still need to be proven.
- Comparison
- Compared with large integrated internet cloud vendors, Kingsoft Cloud has more direct exposure to AI workloads.
- Risks
- Revenue growth below the industry, continued market share loss, failure to reach net profit break-even, weak cloud demand from internet clients, support from Xiaomi and Kingsoft Group falling short of expectations, and GenAI development and enterprise adoption slower than expected.
- MiniMaxBeneficiary
- Strengths
- Rising popularity among domestic and overseas enterprises and developers gives it exposure to the model commercialization theme.
- Weaknesses
- Commercialization and margins are still at an early validation stage.
- Comparison
- Compared with models from large internet companies, MiniMax represents a more pure-play model company opportunity.
- Risks
- Macro and geopolitical uncertainty, competition and operational challenges, data-related model training risks, user-generated content governance risks, and disruption at key service providers.
- SaaSPressured segment
- Strengths
- If it can prove earnings resilience and AI-enhancement capabilities, investor confidence could still recover.
- Weaknesses
- Pressure from AI substitution and business model re-rating continues to persist.
- Comparison
- Compared with cloud and model infrastructure, SaaS is viewed in this report as more exposed to AI disruption.
- Risks
- Substitution by AI features, bottlenecks in enterprise software adoption, insufficient valuation support, and unclear earnings resilience.
- Doubao / Qwen / Yuanbao and other consumer AI chatbotsTheme to watch
- Strengths
- Doubao leads among consumer AI products in China, while Qwen and Yuanbao saw active users rebound after the Lunar New Year promotion.
- Weaknesses
- China's consumer AI chatbot market remains relatively fragmented, and growth has slowed recently.
- Comparison
- The US consumer AI market is closer to a duopoly led by ChatGPT and Gemini, while China's market is more fragmented.
- Risks
- Intensifying competition, higher acquisition and promotion costs, weaker-than-expected product upgrades, and uncertainty over user retention and monetization.
Key data
- Share of Chinese enterprises with GenAI live deployment78%51% are live in at least one function, and 27% are deployed at production scale across multiple business units.
- Enterprises still at PoC or earlier stage22%These respondents expect to reach production scale in an average of 9.3 months.
- Expected time for AI Agent to scale1.7 yearsLonger than the 9.3-month expectation for GenAI overall to move from early stage to scale.
- Expected productivity lift from AI Agent20%Respondents expect it to deliver about 20% productivity gains over the next two years.
- Average 2025 GenAI budget for Chinese enterprisesRmb66mThis is about 12% of overall IT budgets.
- Expected 2026 growth in China GenAI budgets21.0% YoYFaster than the expected 5.0% growth in overall IT budgets and also faster than the expected 17.7% growth in US GenAI budgets.
- AI budget mixPersonnel 23%, infrastructure 21%, models 21%, software 18%, project-related 17%Shows that Chinese enterprises are investing across multiple fronts during the early AI adoption stage.
- Localization level of the AI model layer82%Higher than AI infrastructure software at 60% and AI hardware at 29%.
- Expected localization in 2026AI hardware 51%, AI infrastructure software 61%, AI models 81%The report expects localization across the AI value chain to continue advancing.
- Top factor when choosing cloud vendorsCompute performance 60%Followed by security/compliance at 48-52%, with cost at 43% overall; among large enterprises, only 30% list cost as a key factor.
- Most commonly used enterprise modelsTencent Hunyuan 37%, DeepSeek 36%, Alibaba Qwen 28%The survey was conducted at the end of 2025, and the report notes that the model landscape is changing quickly, so enterprise preferences may continue to shift.
- Biggest challenge to GenAI adoptionUncertain ROI: China 35%, US 59%Additional challenges in China include system integration complexity, a shortage of qualified talent, and data constraints.
Impact & implications
The investment implication of the report is that enterprise AI budget growth is significantly outpacing overall IT budget growth, and cloud and model selection place greater emphasis on performance, security, and ecosystem support than on cost. This benefits leading cloud vendors and model companies with infrastructure, model, and ecosystem capabilities. Alibaba and Tencent are viewed favorably because of their cloud and model capabilities, Kingsoft Cloud benefits from its high exposure to AI workload revenue, and MiniMax is drawing attention as enterprise and developer adoption rises domestically and overseas. By contrast, SaaS companies still face AI substitution risk, the need to prove earnings resilience, and valuation recovery pressure.
Risks
- The competitive landscape in AI is changing rapidly and may intensify further.
- Technology trends, internet user demand, and preferences are changing quickly.
- AI commercialization and ROI remain uncertain.
- Traffic acquisition, content, and brand promotion costs may rise.
- Enterprise IT system maintenance, system integration, and data infrastructure remain bottlenecks.
- Overseas expansion, geopolitics, and export restrictions may affect access to advanced hardware and model upgrades.
- Regulatory risks, especially around data use, online content, and user-generated content governance.
- SaaS and enterprise software may be pressured by AI substitution risk.
What to watch
- Whether 2026 GenAI budgets for Chinese enterprises grow by 21.0% YoY as expected.
- The pace of AI Agent moving from experimentation to scale, especially adoption of open-source Agent frameworks such as OpenClaw.
- Whether enterprise AI use cases expand beyond customer experience and visual analytics into productivity and workflow automation.
- Changes in AI workload share at AliCloud, Tencent Cloud, Huawei Cloud, and Baidu Cloud.
- Changes in enterprise usage of models such as Tencent Hunyuan, DeepSeek, and Qwen.
- Whether localization ratios in AI hardware, infrastructure software, and the model layer continue to rise.
- Active user and monetization trends for consumer AI products such as Doubao, Qwen, and Yuanbao.
- Whether SaaS companies can provide clear evidence of earnings resilience, AI productization capability, and valuation support.