This study employs Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models to forecast tea production, cultivated area, and yield in India using historical data from 1918 to 2023. The dataset was preprocessed through normalization, exploratory analysis, and division into training and testing subsets. Performance metrics, including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE), were used to evaluate the models. The CNN model demonstrated superior predictive accuracy, achieving an MSE of 1,585,503.7, RMSE of 1,259.1, MAE of 1,086.4, and MAPE of 0.36, outperforming the LSTM model. Forecasts for 2024–2030 indicate an initial increase in cultivated area and production until 2027, followed by a decline, with yield peaking in 2025 before decreasing. These trends suggest constraints due to land scarcity, resource depletion, and technological stagnation. The findings highlight the necessity for sustainable agricultural practices, technological innovation, and policy interventions to mitigate the impacts of climate change and urbanization on agricultural productivity. The results demonstrate the efficacy of deep learning models in time-series forecasting, providing valuable insights for policymakers in optimizing resource management and enhancing agricultural resilience.