This paper brings insights into the hydrogen infrastructure for a clearer vision of electrolyser planning so as to avoid the sunk cost of huge investments at an early stage. By means of a power plant dispatching model, we evaluate the potential of hydrogen in France from the residual load of renewables and nuclear power after generating power to meet the electricity demand. Results show that the intermittency is passed on electrolysers leading to low capacity factors and to unclear business models. By 2035, installing decentralized multiple electrolysers with learning rates seems a better strategy than a few large electrolysers with economies of scale; the trend is reversed afterwards. Yet, hydrogen cost increases by 2050 due to low capacity factors and to low residual load. Thus, massive deployment of clean hydrogen production requires ambitious power generation capacities, implying that connecting H2 to power in surplus only, might not be enough to decarbonize economies. Dedicated wind and solar power farms or imports of low-carbon hydrogen will most probably be necessary.