If you're shortsighted enough to think the only "cost" is the device, yeah - it's a "rounding error"
But the TCO of employee connectivity is not merely "the device"
It's everything else that goes into it
Want a standard developer environment? Create an appropriate VDI instance
Want a standard CAD environment? Create an appropriate VDI instance
Want a standard xyz environment? Create an appropriate VDI instance
Then, based on the user role (at the moment of login), spin-up and connect to the proper image
This also allows someone who does two jobs to connect to multiple VDIs based off work needs, and do the 'proper' work on each one
Sure some of this is about data governance ... but most is about enabling end users to be as productive as possible
Updates? Done offline in offhours just to the master image(s)
New version of $OS$ needs to be deployed/trialed? Build/update one image at a time, verify, then push out as the new Golden Master