As artificial intelligence sweeps through finance, many firms treat it as a time‑saving add‑on. Zhiyu Zhishan Investment’s general manager He Li has chosen a different route: to recast AI as part of the investment infrastructure itself. His team has built an integrated system they call “AI Cybertan,” merging research, trading, risk control and backtesting so that models, rules and human judgment evolve together rather than sitting side‑by‑side.
Zhiyu Zhishan is not pitching a lofty narrative. The firm points to track records to make its case: its core value products have outperformed major benchmarks across multiple market regimes since the launch of a representative fund in 2022. That performance underpins He’s argument that AI’s competitive value comes not from isolated reports or chat‑style tools, but from becoming a replicable, trainable part of the investment process that raises consistency and decision speed.
The AI Cybertan architecture is deliberately holistic. In research it digests vast unstructured flows and extracts structured signals; in trading it co‑designs execution frameworks and constraints; in risk control and backtesting it continuously monitors portfolio behaviour and historical outcomes. The firm stresses that AI models systematize a decade‑plus of human research rather than displacing employees, turning tacit investment reasoning into code that can be validated and scaled across markets and asset classes.
That engineering approach informs the house’s global, multi‑asset positioning. Rather than using macro calls to allocate capital top‑down, Zhiyu Zhishan treats macro analysis as a “water‑temperature test” — useful for context but not for dictating precise market weights. Actual allocations are the outcome of bottom‑up, stock‑level research and a “quality‑price” assessment framework; the result today is roughly 70% exposure to overseas markets (notably the US, Korea and Japan) and 30% to A‑shares and Hong Kong stocks, an allocation that has accumulated gradually from individual stock convictions.
Sectorally, the firm tilts to AI and AI enablers, new consumption and resource plays, roughly a 6:2:1 split. He sees AI as the largest investment opportunity of his career, with new consumption and blockchain also attractive on a quality‑price basis. By contrast, he judges parts of biotech and robotics to be rich on current metrics and prefers patience until valuations better match fundamentals. Resources such as copper, aluminum and gold play a stabilising role as hedges against concentrated thematic volatility.
He Li frames the current environment as an “ice and fire” episode: unprecedented structural opportunity on one hand, unprecedented uncertainty on the other. The biggest single market risk, he says, is not a specific asset class but investors’ failure to appreciate how AI will reshape industries, business models and the volatility profile of earnings. For that reason Zhiyu Zhishan treats its AI system as a tool for observing and managing uncertainty rather than a panacea; the firm still emphasises human oversight and explicit risk boundaries.
The firm’s approach speaks to a broader industry cleavage. Many asset managers today deploy AI for transcription, summarisation and efficiency gains; a smaller cohort is integrating machine learning and large language models into core decision pipelines and trading automation. Those that move the fastest could gain lasting advantages in decision consistency and execution, but they also inherit model risk, data governance challenges and governance obligations that are attracting regulatory scrutiny in both China and abroad.
Whether systems like AI Cybertan will consistently beat markets over longer horizons remains an empirical question. The next stage of differentiation is likely to come from who can assemble clean training datasets, sustain high‑quality human feedback loops, and embed robust model governance — plus the ability to manage valuation mismatches and client expectations as winners of the AI era emerge and re‑rate. For investors, the immediate imperative is to understand whether such firms’ claimed advantages are structural or transient, and to set risk limits that reflect the new regime’s potential for both rapid disruption and sudden reversals.