Suppose your company makes the mid-range device. You have competitors above and below you. The one above you uses a fast expensive processor, the one below you uses a slow cheap processor. You need a medium processor for a medium price. If you design your own, it gets cheaper than the existing middle one but still not as cheap as the cheap one.
So if you design one, it's too big, costs too much to fab, for the low end. It's too slow for the high end. You're the only one who can use it. You and companies outside your market who don't compete with you at all, because you make network switches and they make cars.
You also make a good point about the GPL above, but there isn't any reason why processors couldn't use a GPL-style license, is there?
I suspect that the most likely route to open source higher performance cores is some sort of cross industry consortium collaboration - rather than a single firm - with GPL cores and a central body doing a lot of the work.
I do wonder whether you eventually end up with something that economically looks a lot like a cross between Red Hat and Arm though - say with membership fees for support rather than licensing fees.
This is the real reason to upstream, regardless of license. When that incentive doesn't apply and companies only release for compliance, we get Android-vendor-style source dumps and "BSPs" that are not very useful — not upstreamable and often barely even work as documentation.
It's a bit harder with silicon since a chip is "done" once you've sent it to manufacturing, but still somewhat applies when you maintain an evolving line of chips.