🎙 Listen to a summary of this story
Artificial intelligence is moving deeper into China's financial markets, transforming technology once used mainly to automate routine work into systems capable of analysing companies, identifying investment opportunities and helping determine when trades should be made.
The shift is beginning to alter the daily work of analysts, fund managers and securities firms as China's finance industry experiments with increasingly powerful generative AI and machine-learning systems. Instead of simply examining what happened to a company or market, investors are deploying AI to interpret enormous quantities of information and generate forecasts about what could happen next.
The change comes at an unusually active moment for China's capital markets. The country's 150 securities firms reported average net-profit growth of 23.5% in the first half of 2026, according to the Securities Association of China, while operating revenue rose 31%. Revenue from investment advisory services jumped more than 57%, reflecting growing demand for research and wealth-management services.
AI is increasingly being inserted into that expanding research business. Natural-language processing systems can scan company filings, earnings statements, government announcements, research reports and news stories in seconds, extracting information that previously required analysts to spend hours reading and comparing documents.
The technology can also analyse less structured information. News reports, online discussions and corporate statements can be examined for changes in sentiment, while models can compare those signals with share prices, trading volumes, economic indicators and historical market behaviour.
From reading reports to predicting markets
The result is potentially a fundamental change in investment research. Traditional analysis largely begins with historical information — revenue, profits, balance sheets and economic data — before analysts construct forecasts. AI systems can perform much of that preliminary work automatically while continuously incorporating new information.
That does not mean the machines have become reliable substitutes for experienced analysts. Research published in 2026 evaluating AI systems performing professional financial research found that AI-generated reports continued to fall short of human professionals in areas including qualitative analysis, forecasting and valuation accuracy, and the credibility and verifiability of claims.
But even imperfect systems can dramatically accelerate research. An analyst studying semiconductor companies, for example, can ask an AI system to compare years of earnings reports, identify changes in capital spending and margins and cross-reference those findings with industry data. The analyst can then concentrate on determining whether the patterns identified by the machine actually matter.
China's enormous retail investment market also gives securities companies an incentive to push such technology into consumer products. AI assistants can potentially explain financial statements, summarise market developments and generate personalised research for investors who previously had limited access to institutional-style analysis.
That development coincides with renewed activity in Chinese equities. Foreign institutional investors increased their holdings of yuan-denominated shares during the second quarter, with the value of stocks held through the Qualified Foreign Institutional Investor programme rising to 272.8 billion yuan ($40.6 billion), according to Wind Information data. Technology and AI-related companies were among the beneficiaries.
AI itself has become an important investment theme. Profits at companies listed on Shanghai's technology-heavy Star Market increased more than fourfold in the first half of 2026, while earnings among companies on Shenzhen's ChiNext board rose 33%, according to the China Association for Public Companies.
AI enters the trading engine
The more consequential development may be occurring after research is completed. Machine learning is increasingly intersecting with quantitative and algorithmic trading, where computer programs already generate and execute orders according to mathematical models.
AI potentially makes those systems more adaptive. Instead of following only fixed instructions, machine-learning models can search for relationships among prices, trading volumes, economic indicators and other signals, continuously adjusting as market conditions change.
China has an unusual connection between the two worlds. DeepSeek, one of the country's most prominent AI developers, emerged from High-Flyer, the quantitative investment firm founded by Liang Wenfeng. The relationship illustrates how similar the technological foundations of advanced AI and quantitative finance have become: both depend on large datasets, computing infrastructure, mathematical modelling and systems designed to discover patterns beyond the practical capacity of human researchers.
The rise of AI-assisted investing, however, also creates the danger that sophisticated-looking technology can encourage investors to place too much confidence in predictions that remain uncertain. China's securities regulators have already warned consumers about companies marketing supposed "AI stock-picking" and quantitative trading systems with exaggerated claims of accuracy.
In January, the Jiangsu branch of the China Securities Regulatory Commission warned that operators were using terms including "AI stock selection," "big-data stock diagnosis" and "quantitative trading" to promote unlicensed financial services. Shenzhen regulators issued a similar warning in April about businesses selling purported AI quantitative stock-picking software and manipulating historical back-testing results to create the appearance of successful performance.
Regulators confront the black box
For regulators, the challenge extends beyond fraud. Greater automation raises questions about whether large numbers of computer-driven strategies could react to the same signal simultaneously, potentially amplifying volatility rather than reducing it.
China has already tightened supervision of program trading. Rules introduced by the China Securities Regulatory Commission require program traders to report relevant information and stipulate that automated activity must not threaten exchange systems or disrupt normal trading. Exchanges are also required to supervise such activity.
AI adds another complication: explainability. Some advanced models can produce surprisingly accurate predictions without providing a straightforward explanation of how they reached them. That "black box" problem is particularly important in finance, where institutions must understand their exposure to risk and regulators need to determine why consequential decisions were made.
The emerging model is therefore less likely to be one in which machines simply replace analysts and fund managers than one in which responsibilities are redistributed. AI can read more documents, process more data and test more relationships than a human research team could reasonably manage. Humans remain responsible for deciding whether those relationships make economic sense — and whether acting on them is prudent.
That distinction may become increasingly important as AI systems evolve from tools that answer questions into agents capable of carrying out sequences of tasks. In financial markets, an AI agent could theoretically identify a development, research affected companies, revise valuations and recommend or eventually execute portfolio changes.
Such capabilities promise a market that responds to information faster than ever. They also introduce a new source of uncertainty: when thousands of investors employ machines trained to discover opportunities in the same ocean of data, the competitive advantage may no longer come simply from possessing AI. It may come from knowing when not to trust it.
