Arbitrary norm growth in the 3D Navier-Stokes equations

We construct a family of smooth initial data for the Navier-Stokes equations, bounded in $BMO^{-1}(\mathbb T^3)$, that gives rise to arbitrarily large global solutions. As a consequence, we rule out various hypothetical a priori estimates for strong solutions in terms of a critical norm of the initial data. To our knowledge, this is the first example of unbounded norm growth in the well-posed setting. The solutions exhibit an inverse cascade across an unbounded number of modes, with growth resulting from repeated squaring by the quadratic nonlinearity. This mechanism relies on largeness of the data in $B^{-1}_{\infty,\infty}$ and is fundamentally distinct from instantaneous norm inflation and ill-posedness phenomena.