Noisy intermediate-scale quantum devices have the potential to be useful for quantum simulation of materials. A prominent approach for near-term quantum simulation is based on variational quantum algorithms (VQAs). In this talk, we propose a VQA to prepare the groundstate of the Fermi-Hubbard model. In particular, we investigate various state-preparation circuits and benchmark their performance in presence of realistic noise. We find that Hamiltonian-inspired variational forms perform better than a hardware-efficient approach. This work constitutes a first step towards the simulation of strongly correlated fermionic systems. *This work was undertaken thanks in part to funding from NSERC, FRQNT and the Canada First Research Excellence Fund.