Citi: Widening Divergence in A-Share Software & AI Application Sectors — Commercialization Capability to Determine Valuation Trajectory
nashnova research
Citi compiled data from 186 software companies and concluded: China's software industry has hit an AI monetization inflection point, but only companies converting AI into recurring revenue are pulling ahead — the rising-tide trade is over.
What's happening to the sector? Revenue up, profits flat
In the first seven months of 2026, China's software industry grew revenue 9.2% year-on-year, down from 12.3% in the same period of 2025.
The sharper number: net profit growth collapsed from 12.4% to 1.3%. This means → companies are front-loading AI R&D spend, but haven't earned it back yet.
In plain terms = the industry is still in a "spend now, monetize later" phase. Citi sees this as the reason stock-picking in H2 must be far more targeted.
Four sub-sectors — which is growing fastest?
AI sector revenue rose 16% YoY, the fastest of the four. Operating expenses were flat, setting up the conditions for margin improvement.
IT services revenue grew 4% YoY, ending a streak of declines. The main driver: AI-integrated service revenue from leaders like Chinasoft International.
ERP SaaS revenue grew 6% YoY, a clear rebound from the 1% decline in H2 2025. The core driver is a rising share of subscription revenue.
Cybersecurity revenue fell 8% YoY, worse than the 6% drop in H1 2025. This reflects continued tightening of enterprise and government IT budgets, plus fierce price competition.
AI revenue — who has delivered, who hasn't?
Chinasoft International reported H1 full-stack AI revenue of RMB 1.53 billion, nearly 20% of total revenue, up 133.7% YoY. Management guided full-year AI revenue of RMB 4 billion (doubling YoY) and net profit growth above 50%.
Kingdee International posted H1 revenue of RMB 3.625 billion, up 14% YoY. Operating profit swung from a loss of RMB 154 million to a gain of RMB 29 million; cash flow also turned positive. The Cosmic AI platform is expected to generate over RMB 1 billion in full-year revenue.
Yonyou Network returned to revenue growth, but below expectations. High operating expenses widened its loss. This means → Citi has given Yonyou a 12-month to-do list: narrow losses, push cloud migration among mid-to-large enterprises, and make AI revenue visible.
Auto and property software — how wide is the gap?
ThunderSoft grew Q2 revenue 39.6% YoY with net profit up 117.3%. Its AIoT business — smart devices and connected systems — now accounts for roughly 43% of revenue. Citi upgraded the stock to Buy, but cut the target price from RMB 78 to RMB 73 on gross-margin pressure from product-mix shifts.
NavInfo saw Q2 revenue fall 9.4% YoY; net loss widened to RMB 344 million. Average selling prices for autonomous-driving and HD-mapping products keep declining. Citi maintains a Sell rating.
In property software, both Mingyuan Cloud and Glodon reported H1 revenue declines. Glodon's AI-powered quantity estimation is up to 10× faster in some workflows, but weak domestic construction demand is dragging on the core business. Citi cut estimates and keeps a Sell rating.
Cybersecurity: who turned around, who didn't?
Sangfor Technologies grew Q2 revenue 35.7% YoY; net profit surged over 12×, returning to single-quarter profitability. Cloud and AI infrastructure revenue now exceeds 55% of total. Citi upgraded the stock from Sell to Neutral, raising the target from RMB 104 to RMB 148.
Venustech grew Q2 revenue 15.3% YoY, but rising storage-component costs and industry price wars cut gross margin by 7.6 percentage points. Citi expects losses to continue in 2026 and 2027, maintaining Sell and a RMB 11.5 target.
Citi's bottom line: what's the variable that separates winners from losers?
Of 11 covered companies, six beat or met expectations in Q2 or H1. Net profit improvement is relatively broad-based, but revenue improvement is concentrated in the few names delivering real AI revenue.
Citi's top picks: Chinasoft International, ThunderSoft, and Kingdee International. Most cautious on: Yonyou Network and NavInfo.
In plain terms = the question is no longer "who is doing AI" but "who can turn AI into recurring revenue and cash flow." That single line is the core variable driving valuation divergence from here.
市场有风险，内容仅供研究参考，不构成投资建议。