For some subset of C compiler users, gcc is an acceptable substitute, and so there's no reason to pay anyone, anywhere, $.01 for a C compiler. Their business is gone forever and it is no longer worth writing a commercial C compiler in order to get these people's money. Your market is now people with needs that gcc can't fulfill, which is smaller but quite possibly still lucrative.
Ditto for Gimp/Photoshop. I crop and scale pictures, so GIMP or Paint.NET work more than fine. Photoshop isn't worth it for me to buy. On the other hand, my wife and I have both bought laptops in the last year and I spent 3x what she did. Her cheap Dell was not an acceptable substitute, so I paid more. Even if the Dell was free, I still would have paid for my bigger faster laptop with 18.4" screen.
If in 10 years, GIMP does everything that a graphic designer or artist could want, it will no longer matter how advanced Photoshop is. Just ask film camera makers.
The book The Innovator's Dilemma (and its underrated sequel, The Innovator's Solution) talk about this extensively.
I think a free product have to be utterly superior to a paid product in order to displace commercial products.
The cost of replication or the lack of substitutes is irrelevant ... people don't use GIMP over Photoshop because a Photoshop substitute is fucking hard to develop ... it took 22 years to reach its current state.
I am not parsing this statement. Do you mean that because it take 22 years to develop Photoshop, that it is impossible for GIMP to be good enough to be a substitute good thus eventually displacing it?
In order to completely displace commercial products, yes. But for any reasonably complex or powerful product, there's a market of customers that are overserved by it and would pay less for less if they could. This is usually handled by product segmentation - think of options in cars. If you don't need the extra power, go for the 4-cylinder instead of the V-6. If cash is more important than comfort, skip the leather seats. If you need the fastest, most luxurious car and money isn't an option, check every box on the list.
I hear this a lot, but when I see places where people have actually done benchmarks GCC usually doesn't come out too bad, and when you look at the correctness of the code (for C at least) its hard to find something better than GCC.
People still pay for development tools. Lots of people paid for IntelliJ IDEA even with Eclipse around.
GCC is also a piss poor compiler by any standard, that's why many people have such high hopes for LLVM.
You also have to think about "complementary products" ... Apple is investing in the LLVM/GCC compilers because it helps them sell their hardware with their proprietary OS installed, which is by no means free.
The cost of replication or the lack of substitutes is irrelevant ... people don't use GIMP over Photoshop because a Photoshop substitute is fucking hard to develop ... it took 22 years to reach its current state.