¿Puede la IA conducir el 90% del volumen de trading de alta frecuencia prediciendo y moldeando eventos de microestructura del mercado antes de que ocurran ?

Los modelos de IA están superando rápidamente a los traders humanos en latencia y reconocimiento de patrones. Al simular ecosistemas de mercado completos, estos sistemas podrían manipular preventivamente los flujos de órdenes, desencadenando efectos en cascada. Los reguladores luchan por detectar o contener tal desestabilización automatizada.

Background

AI models are rapidly outpacing human traders in latency and pattern recognition. By simulating entire market ecosystems, these systems could preemptively manipulate order flows, triggering cascading effects. Regulators struggle to detect or contain such automated destabilization. Today’s best AI systems can model order-book dynamics and microsecond-scale liquidity imbalances well enough to anticipate short-term price moves with modest accuracy, but they rarely drive anything close to 90 percent of high-frequency trading volume. Firms combine machine-learning signals with colocation, FPGA-accelerated execution, and regulatory-compliant arbitrage strategies to achieve sub-10-millisecond latency, yet they still depend on human oversight for risk controls and fail to predict or shape most microstructure events before they occur. Evidence from exchange-level data shows peak AI-driven participation around 30–40 percent of notional volume in the most liquid futures and equities markets. — Enriched May 10, 2026 · Source: Bank for International Settlements