See:

• Keith Lofstrom's reasoning here in the section entitled "The switching technology cost is proportional to the mass times the exit velocity cubed ..."
• Section IV, "Mass Drivers Versus Rockets", on page 6 of "Human-Rated Launch Infrastructure for the Interplanetary Era: Breaking Rocket and Legacy EML Cost Barriers"
• The Wikipedia page on Startram

Earlier electromagnetic launcher designs exhibit steep, often near-cubic cost scaling with exit velocity because power electronics must convert electrical energy into kinetic energy in decreasing time as vehicle speed rises. While launcher length scales with velocity squared ($v^2$), distributed acceleration requires each segment’s power capacity to scale linearly with velocity. Since the number of powered segments also scales with launcher length, total power-conversion and switching cost scales as $v × v^2=v^3$. Real-world power electronics frequently scale even worse due to voltage, thermal, and reliability constraints, making cubic scaling an optimistic lower bound. This effect dominates system economics at orbital velocities and explains why traditional coilgun and railgun concepts have historically failed to produce affordable human-rated launch systems.