There are differences. I don't know if the idea will actually work but here is what I understand of it. Use an analogy of cities as computer chips/networks.

1. He wants to move roads to 3d just like computer chips are 3d. This should increase bandwidth.

2. He wants to packet switch the system instead of batch. This should lower latency.

3. For the first two to work, you need to greatly reduce the cost/speed to construct them. Think Moore's law.

4. One way to do this is to speed up the boring machines.

5. Another way is to lower the additional necessary equipment. Tunnels carrying internal combustion cars need air circulation lest people die from carbon monoxide poisoning/ lack of oxygen. EVs don't need that.

6. Smaller tunnels are much faster to construct so constructing one for a subway would be much slower than one for EVs.

Arguments against this can either attack the premise that cities are necessary (remote work), that bigger cities are necessary (why not have smaller cities), that existing or other transportation can handle it better, or one or more of the above chains of thought are flawed.

This line of thinking is also found in his latest factories where the line is just a line of cars moved by carriers that aren't physically anchored to the floor). He makes an analogy of factories as computer chips.