Khi LP dùng Doubao dạy tôi đầu tư: Câu chuyện chuyển nghề của một GP quỹ私募

Original｜Odaily Planet Daily（@OdailyChina）

Author｜Golem（@web3_golem）

When LPs start using AI, life is getting tougher for small private fund managers.

Ergou (@ryansoon777) was a general partner (GP) at a small offshore USD-denominated private equity fund focusing on US stocks before the Chinese New Year, but left shortly after to join an AI startup.

"It was already difficult for small private funds to raise capital, and with the rise of AI, many limited partners (LPs) would rather use AI-assisted stock trading tools than allocate funds to us."

Ergou says his decision to change careers was largely due to the subtle impact of AI on the GP-LP relationship. Information and analytical capabilities are seemingly leveled by AI, making it easier for LPs to question professional judgment, potentially increasing friction and leading to capital withdrawal or fund liquidation in severe cases.

The Already Struggling Small USD Private Funds

Ergou's previous fund wasn't performing badly, managing tens of millions of dollars invested mainly in highly liquid US stocks with a small allocation to crypto assets. Its annualized returns over the past three years significantly outperformed the Nasdaq.

Logically, strong performance combined with growing investor demand for overseas wealth management should make fundraising easy. However, Ergou reveals that, in reality, small USD funds like theirs have almost no chance of attracting institutional LPs.

Currently, top-tier Chinese USD private funds with billions in assets (like Jingu, Hillhouse, and Boyu) typically use an "offshore + onshore" structure. The fund entity is domiciled in the Cayman Islands, often as an exempted company or SPC, while the management entity is based in Hong Kong or Singapore.

Recently, however, due to regulatory and fundraising changes, more USD private funds are adopting purely onshore structures like Hong Kong LPF or Singapore VCC.

Ergou's former small USD fund still uses the most "traditional" structure: Cayman SPC + BVI fund manager.

A common saying in the fund industry is "LP determines the structure." One reason top Chinese USD funds stick with the Cayman structure is that their overseas LPs include US university endowments, Middle Eastern sovereign wealth funds, and large European family offices. These elite "old money" investors have been familiar with Cayman structures for decades, and maintaining this tradition reduces communication and trust costs.

But small Chinese USD funds also domiciled in the Caymans cannot attract this top-tier international capital. Their LP base remains primarily in Asia, putting them in an awkward position.

From an Asian perspective, the capital behind USD private funds comes mainly from private banks, mainland China (capital outflows), Hong Kong family offices, and Southeast Asian tycoons.

Even for small funds of similar size, these circles have a natural affinity and trust for Hong Kong or Singapore, preferring to invest in Hong Kong LPFs or Singapore VCCs over Cayman SPCs.

Beyond the structural and scale limitations, differing investment strategies also make fundraising difficult for such small USD private funds.

Fund strategies broadly fall into discretionary and quantitative categories. Discretionary strategies rely on the GP's research, experience, and judgment for decisions, with profitability hinging on the fund manager's market insight. Quantitative strategies involve codifying investment logic into mathematical models and programs for automated or semi-automated high-frequency trading, with profitability relying on statistical patterns within the model.

"Currently, quantitative funds find it easier to raise capital than discretionary funds, especially with AI empowerment, making LPs more confident in quantitative strategies." Ergou notes that after DeepSeek (Odaily note: incubated by quantitative fund firm High-Flyer) went viral last year, market enthusiasm for quantitative strategies surged.

Furthermore, quantitative funds can build trust by showing data and algorithms to LPs. Profits or drawdowns are within a controllable range, and excellent quant strategies can even function like fixed-income products. Discretionary strategies are more abstract; gaining LP trust requires more communication cost, especially during significant drawdowns when LPs easily question the GP's ability.

Therefore, in China, the survival space for small USD private funds like Ergou's former employer has been compressed by the macro environment, making fundraising increasingly difficult. The remaining large LPs in these funds are also questioning whether AI's "investment ability" surpasses the GP's.

The "Diverse" LPs

"In the past, LPs trusted us because we were formally trained. Now, they feed our reports to AI to translate into plain language, then try to 'teach' us how to invest," Ergou says. Since AI became widespread, LPs who only cared about final results have shown significantly more "interest" in his trading decisions.

Ergou once had to remove an LP due to this. This was a 50-year-old real estate entrepreneur, "very cynical" in manner, who invested about $1 million. However, he didn't stay hands-off. He frequently argued with Ergou using fragmented market information and AI-generated conclusions. "His attitude was terrible; he thought I was a clueless young man. We couldn't build trust, so eventually we coordinated his removal."

"Honestly, our LPs are incredibly successful in their fields; they are authorities there. But with AI as a tool, they think they've become authorities in investing too," Ergou sighed.

Small USD private funds, with limited fundraising channels, often get LPs through founder friends or referrals, resulting in a "diverse" group. According to Ergou, their LPs included high-net-worth individuals, real estate entrepreneurs, and funds of funds (FOFs). "Our LPs ranged from Shanxi coal bosses to Forbes-listed billionaires, and some were even second-generation friends who introduced their fathers."

Their relationship with LPs was nuanced. They sometimes didn't charge the 2% management fee, only taking 20% of performance profits. The main characteristic of this LP base was enthusiasm for financial markets and capital outflows, but lacking time and energy for rapid learning and market research.

Thus, in a sense, the GP's core value lies in handling information gathering, market research, opportunity screening, and investment judgment for LPs, using professional skills to compensate for their lack of time, energy, and knowledge, bridging the gap from information to decision-making.

However, with the proliferation of AI tools, this past reliance on professional institutions for information processing and research is being rapidly democratized. Except for final capital allocation and trade execution, a large part of the GP's traditional role is being replaced by AI at lower cost and higher efficiency.

"Our LPs can easily open an IBKR brokerage account. With AI assistance, they can buy whatever industry or stock they like by themselves." Ergou believes the impact of AI is particularly severe on discretionary funds, as investing is always results-oriented. If an LP rides a trend and outperforms the fund, they naturally question the fund's ability.

In contrast, the "information democratization" brought by AI has less impact on quantitative funds and may even widen the gap between them.

Quantitative strategies involve parameters and algorithms that constantly iterate. AI accelerates this iteration, creating a field comparing efficiency and intelligence. Ordinary people without specialized math or finance knowledge cannot build quantitative strategies via AI that rival those of large quant funds.

"Quantitative strategy is essentially about staying ahead of market peers to generate alpha. If you think your basic AI has built a good strategy, chances are many smart people have already found and iterated on it," Ergou says, highlighting the advantage of top-tier quant funds.

Will AI Replace GPs?

However, Ergou isn't worried that AI will completely replace GPs or analysts. AI is neutral and accessible to everyone; it's a lever. GPs can use AI to enhance their knowledge and investment strategies, generating more returns for LPs. What truly frustrates Ergou is that AI increases friction between GPs and LPs.

"Some LPs even question why we didn't invest in hot current targets, and they analyze it with great confidence. They don't understand that GPs don't just buy whatever is trendy," Ergou finds this phenomenon a bit exasperating, especially after US AI and semiconductor stocks became hot themes this year, allowing retail investors to achieve excess returns by betting on sector leaders.

In a bull market, retail returns can easily surpass funds. First, personal investing is more flexible, fault-tolerant, and concentrated. Second, with AI-assisted research, retail investors' research efficiency is greatly enhanced, akin to having a versatile expert on call 24/7.

Especially in this year's US stock market, retail investors betting on popular storage stocks like SanDisk, Micron, and SK Hynix could achieve returns exceeding most funds. "At this point, LPs might either increase their personal accounts and reduce their fund allocation, or directly withdraw from the discretionary fund," Ergou says. In bull markets, everyone often thinks they are "investing gods."

But all this depends on retail investors using AI correctly. Using poor-quality AI can be counterproductive, and Ergou says this is the biggest source of friction with his LPs. "The high-net-worth individuals in China mostly use companion-style AI like Doubao, while more analytical tools like ChatGPT and Claude aren't widespread. This companion-style AI, to provide emotional value, is highly prone to hallucinations in professional fields."

Essentially, the problem isn't AI's capability, but that most people don't truly understand how to use it. AI can integrate vast information in seconds, building a logically coherent analytical framework. But logical coherence doesn't equal factual accuracy. LPs lacking professional backgrounds struggle to distinguish conclusions based on real data from probabilistic inferences generated by models.

Therefore, most investors are seeking validation from AI rather than analysis. AI's ultimate goal isn't to help investors discern truth from falsehood but to complete a conversation.

So, will AI replace GPs? AI can generate ten thousand logically sound investment research reports at low cost, but the essence of asset management is an "ancient service industry" based on trust and mental delegation. The GP-LP relationship is also a mutual selection process.

However, in a future where all "tasks" will eventually be executed by AI to maximize "results," "human private funds" also need to learn from AI and practice providing more emotional value.