Is the Amiga chipset so complex that it can't be replicated directly rather than via an FPGA? I have limited knowledge of hardware engineering - but I imagine with modern manufacturing methods it would easily be possible to build the AGA chipset into a single chip. However, this would likely require extensive knowledge of transistors and to have it manufactured it'd probably need to be ordered in bulk to be cost-effective. I imagine this is why FPGAs are used rather than actually doing it like this.
On the scale of the customer base/market for such things, it's far less expensive to do this in FPGA. To convert the Minimig, aoOCS or Clone_A chipsets from FPGA into a custom ASIC (the holy grail "custom chip" for some people", isn't terribly difficult. It IS terribly expensive. Jeri was able to do the 64DTV as an ASIC due to large volume. As the FPGA implementations work very well so far, I'm not sure why it would be worth such an expense for the available market to go custom ASIC. Especially as FPGA allows updates to improve things, go from OCS or ECS to AGA, etc. I'd rather be in a reconfigurable FPGA than in a fixed ASIC at this point. Show us a flawless AGA with Apollo and maybe at that point it would be more interesting to go ASIC for even faster CPU. Even then, I yself would do ASIC for Apollo and leave AGA in separate FPGA or separate ASIC, unless the full SoC thing could still adapt into an existing 680x0 socket.
