The objective of this study was to investigate the effect of carbohydrate source and cottonseed meal level in the concentrate on feed intake, nutrient digestibility, rumen fermentation and microbial protein synthesis in swamp buffaloes. Four, 4-yr old rumen fistulated swamp buffaloes were randomly assigned to receive four dietary treatments according to a $2{ imes}2$ factorial arrangement in a $4{ imes}4$ Latin square design. Factor A was carbohydrate source; cassava chip (CC) and CC+rice bran at a ratio 3:1 (CR3:1), and factor B was level of cottonseed meal (CM); 109 g CP/kg (LCM) and 328 g CP/kg (HCM) in isonitrogenous diets (490 g CP/kg). Buffaloes received urea-treated rice straw ad libitum and supplemented with 5 g concentrate/kg BW. It was found that carbohydrate source did not affect feed intake, nutrient intake, digested nutrients, nutrient digestibility, ammonia nitrogen concentration, fungi and bacterial populations, or microbial protein synthesis (p>0.05). Ruminal pH at 6 h after feeding and the population of protozoa at 4 h after feeding were higher when buffalo were fed with CC than in the CR3:1 treatment (p<0.05). Buffalo fed with HCM had a lower roughage intake, nutrient intake, population of total viable and cellulolytic bacteria and microbial nitrogen supply than the LCM fed group (p<0.05). However, nutrient digestibility, ruminal pH, ammonia concentration, population of protozoa and fungi, and efficiency of microbial protein synthesis were not affected by cottonseed meal levels (p>0.05). Based on this experiment, concentrate with a low level of cottonseed meal could be fed with cassava chips as an energy source in swamp buffalo receiving rice straw.