Stress measurements in deep boreholes have universally shown that stresses in the Earth's crust are in equilibrium with favorably oriented faults with friction coefficients in the range 0.6–0.7 and with nearly hydrostatic pore-pressure gradients. Because of the lack of any fault-adjacent heat-flow anomaly as predicted by a conductive model of frictional heating, the San Andreas fault has long been thought to be an exception, i.e., far weaker than this standard case. Borehole stress measurements near the San Andreas fault have failed to confirm this weak-fault hypothesis, being either inconclusive or in conflict with it. Directions of maximum horizontal stresses reported to be nearly fault normal in central California are now known not to be regional stresses but a result of active folding within folds that have been rotated 20°–30° clockwise from their original orientations. Everywhere in southern California it is observed that the maximum stress directions rotate to smaller angles (30°–60°) with the San Andreas, within 20 km of it. The sense of this rotation is opposite to that expected from the weak-fault hypothesis and indicates that the shear stress on the San Andreas is comparable in magnitude to all other horizontal stresses in the system. In the “big bend” section of the fault, this rotation is predicted from a transpressional plate-boundary model in which the San Andreas is loaded by a deep shear zone with a locking depth of 10 km. If the adjacent minor thrust faults are assumed to obey Byerlee friction, the crustal-average shear stress on the San Andreas in that region must be in the range 100–160 MPa, regardless of the pore pressure in the fault. These stresses are many times greater than permitted by the weak-fault hypothesis. In the more transcurrent regions farther south, the San Andreas shear stress will be smaller than this estimate, but similar stress rotations observed there indicate that the San Andreas cannot be weak relative to minor faults in that region. These stress rotations can only be consistent with the weak-fault hypothesis if it were assumed that all faults in California were equally weak, which is known to be untrue. The conclusion is that the heat-flow model is flawed, probably in its assumption that all heat transfer is governed by conduction.