You make the software button once and it's there for the many millions of cars. You have to actually manufacture and stick in the many millions of buttons otherwise. Besides the actual action was going to be software on the bus anyways. Your window switch hasn't been directly connected to a motor in decades. It's sending a "window down" message to the bus that goes to the window actuator unit that then drives the motor. You're still paying someone to make it computerized anyways, you were going to pay a team of designers to draw it up and make the plans for the physical switch as well.
The screen was going to be there anyways due to backup camera requirements and because consumers want AA/Carplay.
Evidence suggests that their engineering teams are either not that big or not that good given how garbage most vehicle UI/software is, and it's a price you pay (mostly) once per touchscreen software design, which will span several models, where as the component + install cost needs to be paid for every vehicle in perpetuity.
If you haven't been there, you cannot imagine how bad most car manufacturer's software departments are. They are big, expensive, and crawling with bad practices. Management usually doesn't have a clue about software, so there's a lot of maneuvering with goals being anything but producing good software quickly and cheaply.
It's a little deeper than this, software for each module is typically provided by a tier 1 or tier 2 supplier according to a spec provided by the OEM. Sometimes the tier 1 or tier 2 supplier is also subbing out the software or stuck with some system on chip that sucks.
So for a made-up example, GM wants to build a smart dash in the latest SUV, maybe Bosch or Continental has one with a SoC inside and their own software hell. OEM works with supplier to integrate, bugfix, skin, and customize. But they don't write it from scratch.
AFAIK, car manufacturers want to bring more software in house as a core competency, which is probably good because the "Tier 1"s are generally even worse at software than them and have worse aligned incentives.
The fact that software is bad is not evidence that it was built by a small team or had a low budget. A depressing amount of high-budget, large-team software is awful.
This is absolutely true, but if you scratch the surface of teams like that what you'll usually find is terrible management more interested in shuffling paperwork and CYA than in quality and excellence.
When I worked at Toyota (well, NUMMI) in the '90s, the engineers from Toyota Japan that told me: "I'd kill my mother to save $1 on each car produced." Yes, at Toyota's scale, $1/car is a lot of money.
This makes the incorrect assumption that the infotainment system would be removed, reducing the cost of the engineering.
Adding a virtual button in an infotainment system is much cheaper than a physical button. Especially since the most cost effective routing of those physical buttons would be to the infotainment system that is going to be there regardless.
Remember that someone needs to manufacture those buttons, install them in the factory, stock them for replacement and keep them around several countries in the world in warehouses for when they break.
Now replace all that with a single screen and suddenly costs savings everywhere \o/
Can you find any annual report from a car manufacturer that shows parts sales contributing significantly to profit?
Yes, dealerships make money from servicing and parts: "the service and parts department, which accounts for the other 49.6% of the dealership's gross profits".
But a car manufacturer doesn't capture that, so a manufacturer has no financial incentive to increase profits for dealerships.
The screen was going to be there anyways due to backup camera requirements and because consumers want AA/Carplay.