stress * 2 * thickness * dx = p * 2 * r * dx.
[ stress * walls cross-sectional area] = [ internal pressure * projected internal area ]
Solve for stress, you get:
stress = p * r / t
Regarding your question, yes, as the radius goes to infinity, the stress goes to infinity. The area where the pressure is applied grows with r, but the cross-sectional area where the stress is applied is still (w * thickness * dx.) This equations work well for thin-walled cylindrical pressure vessels (r > 5t is general rule of thumb). For a cube, you would have to develop the equations, but keep in mind that you will have a singularity/discontinuity on the walls because of the right angle.
edit: good to have a reference just in case: http://ocw.mit.edu/courses/materials-science-and-engineering... [PDF ALERT]
So a planar face of a cube cannot satisfy the equilibrium equations? Interesting ... so then a cube will necessarily bulge so the radius is enough to satisfy the equations, right?