India's agriculture is a significant component of the Gross National Product and is largely dependent on the Indian summer monsoon. This makes an accurate prediction of rainfall a key factor in improving agricultural production. A statistical cycle, widely known to have a strong 2.85-year period and will therefore be called a triennial oscillation here, plays a vital role in such predictions. In the present work, we examine the spatial distribution of this short-period oscillation at 1° × 1° (lat./long.) resolution using 1,260-month data (1901–2005). A power spectral analysis of the observed data set shows that the statistically significant triennial oscillation (at 95{\%} confidence level) is present over 54{\%} of the total number of 354 grids across India, which covers 69.03{\%} area. Projections of selected data from models used in Coupled-Model Inter-comparison Project Phase 5 predict weakening of this oscillation (in amplitude) by 2100. The projections indicate reduced confidence levels of this oscillation at 80{\%}, 85{\%} and 90{\%}, respectively, in 3{\%}, 20{\%} and 62{\%} of the selected 54{\%} grids, which covers 41.3{\%} area. A weakened triennial monsoon cycle will have a severe impact on agriculture and water resource management, particularly over the southwest-coastal, northern and northeast-central parts of India.