This paper presents epidemiological data from ongoing HIV-prevalence surveillance in a large South African manufacturing company located in various provinces, as well as one location in Swaziland. A survey of the workforce in 12 of the manufacturer's business units (BUs) throughout these locations was carried out in 2001–2003, and each was repeated three years later, in 2004–2006. The company is typical of manufacturing in South Africa, in that it is labour-intensive, male-dominated, and has high proportions of semi-skilled and unskilled employees. The surveys were anonymous, cross-sectional and unlinked. The workforce was stratified at the level of analysis by workplace, geographical location, job status (permanent or contract), job skill level, and age group. In 2001/3 a total of 5 177 permanent employees, representing 76% of those on duty at the time, plus 1 966 contract employees were surveyed. In 2004/6, 5 445 permanent employees, representing 81% of those on duty, and 2 688 contract employees were surveyed. HIV prevalence in the 2004/6 survey ranged from 0.9–30.7% throughout the 12 BUs, with a mean of 15.4%, showing an increase from the previous mean of 13% prevalence in the 2001/3 survey of the same locations (p = 0.0001). Differences in HIV prevalence across the BUs were determined by geographical location and the profile of the workforce. Similar proportions of contract employees were found to be HIV-positive in the initial survey and repeat survey (23% and 24.5%, respectively). HIV prevalence was inversely proportionate to permanent employees' job skill level. The 2004/6 data show lowest HIV prevalence in the upper-management category (3.6%) and highest prevalence in the unskilled category (22.6%). HIV prevalence did not increase in the upper- and middle-management categories but increased significantly in the skilled (p = 0.02), semi-skilled (p < 0.001) and unskilled (p < 0.001) job categories. HIV prevalence was similar among contract employees and unskilled permanent employees. Race (not considered in the study) may in part confound these findings. In both surveys, HIV prevalence was highest in the age group 30–39. Prevalence did not increase between surveys in the age group 18–29; however, significant increases occurred in the age groups 30–39 and 40–49 (p < 0.001), and possibly also in the age 50+ category (p = 0.09).