Citadel founder Ken Griffin declares that "generative AI is useless": Alpha cannot be discovered, and Wall Street still has to use its hands to beat the market

Ken Griffin bluntly stated at the JPMorgan Robin Hood Conference that GenAI has not yet mined Alpha for hedge funds, triggering the market to re-examine the practicality of generative AI for trading, risks and human-machine collaboration
(Preliminary summary: ChatGPT analyzes the moment of the crypto market correction: Is it the bull market "halftime" now?)
(Background supplement: Testing the AI of three major exchanges with the "1011 crash" Assistants: Hexagon Player, Riddler, and one is "Android Thinking")

Contents of this article

The JPMorgan Robin Hood Investor Conference held in New York yesterday (15th) attracted heavyweights from Wall Street and Silicon Valley. Ken Griffin, founder of Citadel, threw cold water on generative artificial intelligence (GenAI) in his speech, pointing out that although GenAI can currently improve productivity, "for the discovery of Alpha Still unable to do so."

"GenAI can indeed improve productivity, but it still falls short of discovering Alpha (excess returns)."

At a time when AI concept stocks are hot, Griffin's attitude has made the outside world reflect on the true value of this technology wave.

Citadel’s practices behind the cautious attitude

Griffin’s point of view does not mean that Citadel is resistant to new tools. The company has already introduced GPT-like models into the research process to assist with document summarization, instant insights and chatbot-assisted analysis, but key decisions are still led by senior investors.

In the second quarter, Citadel reduced its long positions in Broadcom by about 82% and Palantir by about 48%, but quadrupled its holdings in Nvidia to more than 8 million shares. These operations reflect that Citadel only bets heavily when it determines competitive barriers and hardware leadership, while keeping a distance from highly valued software stocks. Citadel focuses on maintaining a balance between leveraging market momentum and risk control.

The market is optimistic about GenAI: Will profits increase?

In contrast to Griffin’s reservations, some research institutions continue to issue positive assessments, with annualized returns increasing by 3–5% for hedge funds that are the first to adopt GenAI, which is particularly beneficial for quantitative and equity strategies. GenAI provides faster market signal extraction and portfolio stress testing by integrating synthetic data, large-scale language models and modular workflows.

"Moody’s CreditView Blog" pointed out that Agentic AI can monitor transactions 24 hours a day, flag abnormalities, and strengthen risk management. AIMA research also shows that 95% of hedge fund managers have used GenAI tools, and 90% of investors expect to see positive contributions within three years.

Wall Street Adoption Challenges

However, there are still challenges that cannot be ignored under the optimistic data. CFA Institute analysis pointed out that the GenAI model does not respond well to highly complex scenarios such as geopolitical shocks and is prone to being inaccurate during large fluctuations. In terms of supervision, the problem of model black box and bias still needs to be solved. Once the basis for judgment cannot be traced, it will put pressure on the compliance of financial institutions.

For GenAI to fully realize its potential in 2030, the hedge fund field still needs 140,000 skilled professionals, which poses problems in terms of infrastructure and talent training. Popularization may also strengthen market homogeneity and increase concerns about systemic risks and the weakening of human analytical capabilities.

Hybrid human-machine collaboration: a more realistic blueprint

Under the pressure of efficiency, risk and supervision, many practitioners focus on the "human-machine collaboration" model. This method is to let AI conduct preliminary screening and pattern search on massive data, and then investors add experience and situational judgment to finally confirm the strategy. This model not only takes advantage of AI's computing advantages, but also retains human sensitivity to extreme events and unstructured information.

For Griffin, prudence does not mean denying technology, but ensuring that every investment can bring returns with controllable risks. After understanding this, the market may be able to evaluate GenAI's position more calmly: it is only an assistant, not a savior.

In summary, GenAI is undoubtedly reshaping data processing and operating processes. The short-term profit improvement may only be an improvement in efficiency, but there are still variables for the stable generation of Alpha at the hedge fund level and the path to beating the market. As Wall Street and Silicon Valley debate the next wave of innovation, Griffin's words remind investors that true competitive advantage comes from precise application of tools, managed risk, and clear judgment not blinded by popular narratives.