BlackRock’s AI crypto forecast aims to quantify risks

BlackRock released an AI-based forecast for cryptocurrencies this month that generates probability distributions for prices, volatility and correlations. The research team said the tool is designed for institutional clients and portfolio managers who want quantitative views of crypto risks and outcomes.

The model combines historical market data, on-chain metrics and selected macroeconomic indicators to produce short- and medium-term scenarios for major digital assets. BlackRock described the approach as a probabilistic forecasting system that delivers ranges of possible outcomes rather than a single price target.

BlackRock wrote that the system uses ensemble machine-learning techniques to blend signals from price histories, transaction-level blockchain data and broader financial indicators. Outputs include probability-weighted price ranges, volatility projections and scenario-based stress tests that model asset behavior under conditions such as rising interest rates, liquidity shocks and equity market stress.

The research note included sample use cases showing how the forecast can inform position-sizing, hedging and stress testing. Portfolio managers can apply the probability distributions to calculate expected shortfall and value-at-risk for crypto holdings or estimate the likelihood of extreme price moves over a specific horizon. The note stated the forecast is meant to complement existing fundamental analysis and due diligence on protocol risk, custody arrangements and counterparty exposures.

The forecast follows BlackRock’s recent expansion of crypto offerings, including involvement in spot Bitcoin exchange-traded funds and increased hiring across digital-asset teams. The firm positioned the model as a way to produce institutional-grade analytics that can be integrated into conventional risk frameworks and portfolio construction processes.

BlackRock provided scenario outputs that show higher correlation between crypto and other risk assets during severe global market stress and different drawdown patterns connected to liquidity and regulatory developments. The firm highlighted the use of multiple data sources-on-chain flows, exchange order books and macro indicators-to capture drivers that can accelerate price moves in digital-asset markets.

The research team warned of limitations common to machine-learning approaches. They noted model performance can deteriorate during regime changes, rare events are difficult to predict, and forecasts are sensitive to the choice of training windows and input features. The note recommended combining the forecast with fundamental assessments and treating model outputs as dependent on input quality and model uncertainty.