To evaluate the in vitro fracture load of extensively damaged vital teeth after either direct or indirect restauration, severe tooth substance loss was simulated for 96 molars. Subsequently, two cavities were prepared with little (design 1) or more substantial (design 2) residual tooth support. All molars were provided with a 2-mm ferrule design and then divided into 12 test groups based on their occlusal surface size. They were restored with composite or with either of two types of single crown (cast metal or milled zirconia). After thermal ageing (10,000 cycles at 6.5 and 60 °C), 1.2 million cycles of chewing simulation were applied (64 N). Maximum fracture load was determined with a loading angle of 45°. Statistical analysis was performed by use of Kaplan-Meier modelling, Student's t-tests, one-way anova, post hoc Tukey's HSD tests and linear regression analysis. Regarding mean fracture load without ageing, the indirect restorations outperformed composite (design 1: direct: 508 ± 123 N, indirect: 741 ± 248 N; design 2: direct: 554 ± 167 N, indirect: 903 ± 221 N). After artificial ageing, however, these differences were no longer significant (design 1: direct: 328 ± 189 N, indirect: 506 ± 352 N; design 2: direct 399 ± 208 N, indirect 577 ± 292 N). Instead, the fracture load of the aged composite restorations was comparable with that for zirconia (design 1) and cast metal (design 2) crowns. Fracture loads of direct composite restorations after artificial ageing might fulfil clinical requirements.