Ken Griffin, founder of Citadel, told Bloomberg that generative artificial intelligence has not meaningfully helped hedge funds beat the market, a blunt assessment that undercuts a prevailing narrative of an imminent AI-driven trading revolution. "Generative AI isn’t helping hedge funds produce market-beating returns and isn’t meaningfully impacting the industry so far," he said, reflecting skepticism from one of Wall Street’s most influential technologists.

Griffin’s comments come as asset managers, quant funds and market makers have invested heavily in machine learning talent and compute infrastructure amid a broader surge in interest in large language models and other generative systems. But finance has long placed a premium on incremental advantages: faster signals, cleaner alternative data, and lower trading costs. In highly efficient markets, predictable gains from a new technology are often arbitraged away quickly, and Griffin emphasized that reality.

Industry participants and academics point to several reasons why generative AI has not yet been a silver bullet for trading performance. Financial markets are noisy and nonstationary, meaning patterns that exist in historical data frequently break down in real time. Generative models excel at pattern recognition and synthetic output — useful for document generation, research summarization and client communications — but they do not automatically translate into stable forecasting power for price movements. Moreover, much of the early value from AI in finance appears to be operational: automating research workflows, speeding due diligence and improving risk monitoring rather than generating pure alpha.

The implications are practical. Firms that expected immediate improvements in returns may be reconsidering budgets, redirecting funds toward proprietary data acquisition, low-latency execution systems and model-validation teams. The democratization of large models — with many firms able to license or replicate baseline capabilities — also erodes exclusive advantages, forcing a shift toward more expensive and idiosyncratic inputs. Griffin, whose firms combine market-making operations with asset management, has argued for a focus on engineering and data rather than hype-driven investments.

Regulators are watching the transition as well. While the Securities and Exchange Commission and the Commodity Futures Trading Commission have not issued sweeping AI-specific rules for trading, both agencies have highlighted risks tied to model governance, market structure and operational resilience. The limited payoff to date from generative systems may lower immediate systemic risks but raises policy questions about transparency, model risk management and whether AI will concentrate advantages among the largest players who can afford bespoke models and vast computing budgets.

Looking ahead, the market’s view of AI in finance is likely to bifurcate. Generative models will continue to reshape back-office functions, client-facing services and research productivity, delivering measurable cost and time savings. But the elusive quest for persistent, model-driven alpha will remain constrained by competition, data limits and fundamental market efficiency. Griffin’s verdict is a reminder that the most disruptive financial technologies often mature slowly, producing incremental structural change rather than instant windfalls for investors.