The discovery of 'high-risk' de novo copy number variants (CNVs) associated with neuropsychiatric disorders such as schizophrenia offers the opportunity to translate these findings into useful tools for clinical geneticists. However, this will require estimation of penetrance for these variants, which has not yet been properly considered. To facilitate this process, we estimated the penetrance of CNVs associated with schizophrenia, at 15q13.3, 1q21.1, 15q11.2, 17p12, 2p16.3, 16p13.1 and 16p11.2 with a novel Bayesian method applied to pooled data from published case-control studies. For these CNVs, penetrance for schizophrenia was between 2 and 7.4%, which contrasts with the much higher penetrance for schizophrenia of the 22q11.2 deletions found in velo-cardio-facial syndrome. The highest penetrance was for 15q13.3 deletion (6-9% in individual studies) and the lowest was for 15q11.2 (2%). CNVs confer much higher risk for schizophrenia than common variants, but their penetrance is substantially lower than Mendelian disorders or other syndromic conditions. Since these CNVs predispose to multiple disorders, including epilepsy, autism and intellectual impairment, penetrance estimates will also need to take into account diagnostic specificity, and their overall penetrance for any neuropsychiatric disorder is likely to be much higher. Thus, although CNVs are still far from being clinically useful or relevant to genetic counselling for specific disorders, their detection may hold an important clinical value in predicting negative developmental outcomes.