What's more expensive and has long term track record success?

Pay today or pay tomorrow. I don't think it is an outlandish thought IF $100M/mile tunnels are achieved that the TCO for an elevated roadway will be higher. Elevated roadways have to be replaced eventually. Tunnels are more durable, look at the Holland Tunnel for example, 89 years old and no need for replacement in sight. And it's $48M price tag would be $3.3B today for 1.62 miles.

Elevated roads require surface space. A tunnel requires a small entry-point.

Well, given that both tunnels and elevated roads are still being built, the most expensive likely varies depending on the situation. If Musk really reduces the cost of tunneling by 10x, that calculus will shift in the direction of tunneling, obviously.

Dirt roads are extremely cheap and have been around forever. That doesn't make them the best choice in a modern society.

It's a lot easier to tunnel through sky than rock.

But rock comes with supporting structure, whereas you need to build those into "sky tunnels". This limits just how much "sky tunnels" you can make in two ways - there's only so much space on the ground you can use, and there's only so much weight it can support. That, and people will get anxious if their sky is totally obstructed by highways in the sky.

Tensile-strength-to-volume ratios for many materials is ridiculously high. Tunnel wall, ceiling, and floor quantities represent massive overengineered for that purpose

A cable tramway is one option for an aerial system. The load-bearing capacity of a 15mm steel cable is impressively high, and it can be strung along pylons.

Not that you'd want to bolt street cars to such a system, but it's an example of a possible people-mover that doesn't require persuading a multiple-kilometers-long bolus of solid granite it'd prefer being elsewhere.

Whilst I'm aware that Musk's premise here is that he can be the boringest guy ever, I'll note that the Gotthard Base Tunnel, in Switzerland, required 17 years of excavation to extend 35.5 km.

https://en.m.wikipedia.org/wiki/Gotthard_Base_Tunnel

Tunnel systems can also give rise to some interesting cascade failure modes. Cab-forward deisel locomotives aren't generally considered a risk factor, until they are. Summary of an accident report, here, five souls lost to bad design of stock and tunnel, inadequate maintenance, poor judgement, and panic response:

https://ello.co/dredmorbius/post/qfkx-7qsggpi0bgnvplwmq

Also, noting: the incident inquiry found for operator error of the crew. I find that conclusion difficult to justify based on the presented evidence.