Over the past thirty years, consensus estimates of the contemporary global carbon budget compiled by the Intergovernmental Panel on Climate Change (IPCC) and Global Carbon Project (GCP) report a robust net land carbon sink, nearly similar in magnitude to carbon uptake by the world's oceans. Evidence for a strong net land sink comes from the north-south gradient of atmospheric CO₂, global trends in atmospheric O₂, and ocean models constrained by multiple tracers. Additional direct and indirect evidence for a land sink comes from eddy covariance tower measurements, increases in the amplitude of the annual cycle of CO₂, and satellite-derived greening trends. Here, we used published data to show considerable differences between 2000-2019 estimates of the northern hemisphere net land sink derived from an atmospheric inversion (1.9 ± 0.5 Pg C/y), Coupled Model Intercomparison Project Phase 6 (CMIP6) land surface models (1.1 Pg C/y), and a remote sensing-derived estimate based on changes in vegetation carbon (0.8 ± 0.7 Pg C/y). We hypothesize that the remote sensing-derived estimate is the most robust because it is directly tied to land surface observations. At a global scale, the net land sink is about a factor of 2 lower than the estimate from the Global Carbon Project. In models, improved representation of ecosystem disturbance, limits to soil carbon sequestration capacity in mineral soils, and physiological mechanisms decoupling photosynthesis and growth have the potential to weaken carbon uptake and improve agreement with the remote sensing-derived estimate. We propose adjustments to other terms in the global carbon budget that partially reconcile the weak land sink with other available constraints. Key among these is a proposed decrease in fossil fuel emissions.