Magnetar Capital, a hedge fund managing $18 billion in assets, plans to launch a new fund later this year that completely eliminates human analysts from the research process. Instead, hundreds of AI agents will scour for investment opportunities, analyze individual stocks, and predict trends. Final trading decisions, however, will still be made by humans.
The new fund won't need a large team of researchers for bottom-up fundamental analysis. Magnetar employees' roles will shift to developing and maintaining the AI infrastructure. The system architecture was designed by Trevor Mottl, the firm's head of AI quantitative research. Mottl previously served as a long-short equity risk manager at Balyasny and led AI investments at Man Group and Walleye. The fund's strategy leans toward long positions with long holding periods, allocating only a tiny fraction of capital to capture millisecond-level trading signals.
The system's reasoning layer acts as a command center, orchestrating the coordination of hundreds of AI agents at different points in time. The entire setup runs on multiple Nvidia servers, using high-density signal processing to filter out market noise and identify potential pricing patterns.
Similar experiments are already underway in the quant hedge fund world. Earlier this year, former Coatue investment manager Rahul Kishore launched a fund managed by three humans and an AI agent named Eve. But the real-world effectiveness of AI in investing remains controversial. In a recent simulated test involving eight cutting-edge large language models, most systems ended up losing money.
Magnetar was founded in 2005 and primarily invests in alternative credit. The new fund is Magnetar's first AI-driven investment vehicle in public equities. In 2024, the firm also launched a venture capital fund focused on generative AI.