11 AI Models Forecast Bitcoin Up to $105,000 by Year-End

Eleven AI models run by independent researchers and quantitative teams produced year-end Bitcoin price forecasts that reached as high as $105,000. The forecasts targeted Bitcoin’s price at the end of the calendar year and produced a wide spread of outcomes.

Models diverged because teams weighted different indicators and used different training data and architectures. Some models emphasized on-chain metrics such as active addresses and exchange flows. Others prioritized macroeconomic variables including interest rates and inflation. A subset relied on market microstructure data and derivatives positioning.

Approaches varied across teams. Several used ensemble techniques that combined multiple machine-learning algorithms. Others used single-model architectures focused on pattern recognition in historical price series. Methods included supervised learning on historical price and indicator pairs, reinforcement learning aimed at trading objectives, and unsupervised methods to detect regime shifts.

Many teams produced probabilistic distributions rather than single-point estimates. Where distributional outputs were available, tails extended both above and below central estimates. Teams ran simulations to account for different volatility and liquidity scenarios and adjusted outputs to reflect model uncertainty.

Forecast inputs also included potential drivers such as changes in monetary policy, institutional inflows, regulatory developments affecting cryptocurrencies, miner activity and token issuance. Several models added sentiment indicators from social platforms and derivatives signals such as futures funding rates and options skew to adjust short-term risk assessments.

Teams published backtests showing mixed historical performance for short-term cryptocurrency forecasts. Developers noted that extreme market events can reduce model reliability. Retraining schedules varied: some teams updated models weekly, while others used longer retraining intervals to limit sensitivity to recent noise.

The forecasts reflected differing choices on features, time horizons and data handling, producing a broad range of possible year-end prices for Bitcoin.