How large can the dark matter self-annihilation rate in the late universe be? This rate depends on (rho_DM/m_chi)^2 <sigma_A v>, where rho_DM/m_chi is the number density of dark matter, and the annihilation cross section is averaged over the velocity distribution. Since the clustering of dark matter is known, this amounts to asking how large the annihilation cross section can be. Kaplinghat, Knox, and Turner proposed that a very large annihilation cross section could turn a halo cusp into a core, improving agreement between simulations and observations; Hui showed that unitarity prohibits this for large dark matter masses. We show that if the annihilation products are Standard Model particles, even just neutrinos, the consequent fluxes are ruled out by orders of magnitude, even at small masses. Equivalently, to invoke such large annihilation cross sections, one must now require that essentially no Standard Model particles are produced.