As a rule of thumb, I personally like to have 3 times more historical data relative to my forecast horizon (the distance into the future I want to forecast). However, this amount of data is not necessarily a hard stop when you use simpler forecasting methods such as ARIMA and exponential smoothing. The reason I prefer this amount of data is because it allows for creating lags (information that helps create the forecasts) further from the past, which can help the model understand seasonality better.
Another factor that can allow for successful forecasting with less data is the smoothness of what you are forecasting. If it isn’t very volatile less data may be able to accurately capture the trends and patterns.