One thing is that simulation don't require hardware.

So - depending on licenses - the student could sit and play at home, without having to check out hardware from the lab.

Next thing is that some simulators also allows programmable peripherial stimuli which means that it's possible to develop and test software features that requires external hardware without having that hardware. Examples for the Keil debugger - use of a graphial application to mimic an LCD, or the simulation of a I2C-connected memory.

In a simulated environment, time has no meaning. The disadvantage is that sometimes, you fail to find timing-related bugs that requires the real hardware. But the advantage is that you can stop the program, look at things, and then start again without problems. This isn't always possible when debugging live hardware.

Another example from Keil - the built-in logic analyzer/oscilloscope that can capture the state of a large number of signals, and let the user scroll through the output. Not all developers have logic analyzer. Or their analyzer have too few channels or too short memory depth.

So simulation is not dead. It's just a different route to write and test code.

While the Keil evaluation licenses are size-limited, they still allow enough code for a developer to test specific functions of their application - even if the real application is too large, it can normally be broken down into modules that are tested individually.

Another thing with a simulator - it can often support a number of different chips (at least within the same architecture), making it reasonably easy to dry-test a different processor before deciding to buy an evaluation board for the new processor.

So while emulation is very good to have, it is not a full replacement for simulation.