At 4.5% loss per year, you'd still have 16% of cars running at 40 years. That's pretty normal.

By that logic, shouldn’t about 25% of US persons be 150, given the annual death rate of 9.28 per thousand?

No. Humans age in a way that cars don't, so "that logic" would not attempt to apply the same curve to humans.

If you're done nitpicking, you're welcome to explain your number better. You forgot to say how to apply "4.5%". I'm sure an exponential fit has issues, but a linear fit would be much worse, and anything fancy needs more data points.

That's why I gave a range. That average stat actually seems to line up with the low end of that range, and since every car isn't scrapped at the same age it's going to be a distribution. There are not many cars from 1985 on the road today, but there sure are some. And since we're talking software which doesn't actually degrade, it shouldn't be the thing limiting the overall lifetime.