For months, the sharpest question in AI investing has been whether large language models can do more than replace lawyers and coders — can they replace hedge fund managers? Nof1's Alpha Arena is turning that debate into a live experiment, pitting frontier models from OpenAI, Anthropic, Google and xAI against each other in autonomous trading competitions using real capital. SUI Group and Karatage co-led Nof1's $15 million raise, and Wall Street is now paying attention.
The early results are instructive precisely because they're ugly: most models lost money, overtraded, or failed at basic risk management. That failure is the signal. Language models were trained to predict text, not allocate capital under adversarial market conditions — and Alpha Arena is designed to close that gap by rewarding or punishing models with real outcomes.
The same investor pair also backed Recursive…
CoinDesk