This paper uses longitudinal data of more than 13,000 firms to analyze the effects of on-the-job training on firm level productivity and wages. Workers receiving training are on average more productive than workers not receiving training. This makes firms more productive. On-the-job training increases firm level measured productivity between 1 and 2%, compared to firms that do not provide training. The effect of training on wages is also positive, but much lower than the effect on productivity. Average wages increase only by 0.5%. Sectoral spillovers between firms that train workers are found, but only in firms active in the manufacturing sector. In non-manufacturing no spillovers seem to take place. The results are consistent with recent theories that explain on-the-job training, related to imperfect competition in the labor market, such as monopsony and union bargaining.