Later this year, 5000-9000 boards will enter the market. It will be late 2019 before 25,000 boards are even manufactured. By then, the crypto market will probably have doubled or tripled, which means the number of supported FPGA's also doubles or triples. So it is not clear that the FPGA's can ever catch up to the market.
I would add that the secondary market for these boards, once rendered unprofitable for altcoin mining, is nil.
PS -- The '300A' Bittware card is a single FPGA? VU13P maybe? Seems like overkill.
The secondary market for these boards is huge. Xilinx sells 250 boards per month to non-crypto developers. The boards can also be used for AI-mining and other distributed cloud-supercomputing applications. In my opinion once these boards are everywhere, very few people will be mining with them, as there will be more profit to be made in other fashions. Just look at Amazon. They rent out these FPGA's for $30 per day to non-miners. The cell phone giant Huawei has also launched a cloud-FPGA platform to rent out FPGA's and compete with amazon. Crypto is a tiny segment for the applications of these boards.
The Bittware CVP-9 is a 300A VU9, and the CVP-13 is a 300A VU13. Not overkill; even at these currents we are still power limited in our applications, by a significant margin. We would need 500-700A to run at full voltage, full logic, full clock speed.