If you are (and most of us are) still encountering cases where the closed models are not good enough, the open models are (mostly) even worse, so they're not competition.
I anticipate the money will run out, and when it does the open models catch up (that appears to be cheaper than maintaining a leading position); at that point, I expect us to face the "commoditise your complement" situation.
The true cost/token is something that I am still wondering about. The most expensive chinese model costs $3 per input to run. And we still have massive markup from the american closed source model companies.
I agree that money will run out. But what will policy do?