There is a lack of accurate prevalence data on undernutrition and the risk of undernutrition among the hospitalised elderly in Europe and Norway. We aimed at estimating the prevalence of nutritional risk by using stratified sampling along with adequate power calculations. A cross-sectional study was carried out in the period 2011 to 2013 at a university hospital in Norway. Second-year nursing students in acute care clinical studies in twenty hospital wards screened non-demented elderly patients for nutritional risk, by employing the Nutritional Risk Screening 2002 (NRS2002) form. In total, 508 patients (48·8 % women and 51·2 % men) with a mean age of 79·6 (sd 6·4) years were screened by the students. Mean BMI was 24·9 (sd 4·9) kg/m2, and the patients had been hospitalised for on average 5·3 (sd 6·3) d. WHO's BMI cut-off values identified 6·5 % as underweight, 48·0 % of normal weight and 45·5 % as overweight. Patients nutritionally at risk had been in hospital longer and had lower average weight and BMI compared with those not at risk (all P < 0·001); no differences in mean age or sex were observed. The prevalence of nutritional risk was estimated to be 45·4 (95 % CI 41·7 %, 49·0) %, ranging between 20·0 and 65·0 % on different hospital wards. The present results show that the prevalence of nutritional risk among elderly patients without dementia is high, suggesting that a large proportion of the hospitalised elderly are in need of nutritional treatment.