A fundamental problem in plate tectonics is the shear strength of major plate boundary faults. This translates to the question whether the generally observed small earthquake stress drops of 3–10 MPa on major faults release most of the accumulated stress or only a small fraction of it. There is strong evidence that the San Andreas fault, a major transform plate boundary, is weak (<20 MPa shear resistance). It is not yet clear whether subduction thrust faults are also weak. We present two types of evidence from the northern Cascadia subduction zone that indicate very low coupling shear stress on that plate interface and hence very low strength of the subduction thrust fault, comparable to that estimated for the San Andreas fault. First, the well-defined surface heat flow and heat generation allow negligible frictional heating on the plate interface. The average shear stress on the fault must thus be very low over a time scale of a few million years. Second, focal mechanism solutions for small crustal earthquakes in the southern Vancouver Island area indicate that the horizontal stress in the direction of plate convergence has a similar magnitude to the vertical stress. This inferred stress state requires the present tectonic stress coupled across the subduction thrust fault to be very low. One explanation for the weakness of the fault is the presence of near-lithostatic pore fluid pressure in the region of the fault zone for which there is independent evidence. The conclusion of a weak subduction thrust fault does not conflict with geodetic observations of contemporary surface deformation which indicate that the fault is currently locked, accumulating strain energy toward a future great earthquake. The surface deformation responds to the small (<20 MPa) temporal changes of the stress field associated with the subduction earthquake cycle. This transient stress is superimposed on the larger background regional stress field in which the maximum compression is parallel to the margin. The weakness of the Cascadia subduction thrust fault and the unusual stress state of the forearc region have important implications for earthquake hazards. For example, a subduction earthquake may induce large strike-slip earthquakes in the forearc that affect a large area.