Right, but what if its not feasible on land due to thermal expansion, but is feasible underwater with slightly higher power requirements to overcome air resistance? Just a thought.

The design is intended to be robust with regard to leaks (pumps can overcome some leakage to keep interior at a near-vacuum). This would not work underwater.

Pumps can overcome a small amount of leakage - from tiny cracks between bolt threads, etc - places that an un-reactive gas can make it through a huge pressure differential. In shallow water, one has the viscosity of water to work with, which makes this much easier in liquid phase.

What I don't know is how pumps, and cracks, would react to gaseous H2O versus gaseous N2; Or if perhaps a liquid with low offgassing potential could be used as a boundary layer to seal cracks instead of water (which, while much higher viscosity than a gas, remains lower than most liquids).

One big advantage of any submersed liquid approach: Earthquake resistance.

Additional notes: My main theory before the announcement was that the concept used a fast-speed-of-sound, low-density STP hydrogen tube and maglev rather than slow-speed-of-sound air tube, or an expensive-to-seal vacuum tube. An STP hydrogen tube hovercraft levitation model still seems pretty damn attractive to me, especially with a sealing liquid buffer of some sort - it cancels out the problem of water-vacuum interface, and in water, flammable seals, as well.