There was centralization of GPU mining at one time as well, until prices came down and small farms began to grow.
The same process will happen with ASIC miners. As prices get cheaper more rigs will be purchased by individuals and that will solve your much worried about but mostly inconsequential centralization argument.
You may be right in that one day ASICs will become cheaper and more affordable for the average miner, however as it stands today it's infeasible, and the main issue is that the largest manufacturer of these machines also happens to be the largest mining pool which is where the centralization stems from. By the time ASICs become cheaper, the top manufacturers/pools will most likely have come up with even more powerful machines that makes the cheaper ASICs obsolete, which will not solve the issue of centralization one bit. This will be an endless cycle.
You're also missing the overall goal of this project, as we believe there is a flaw in PoW as a whole, which is why we're switching to PoS. PoW has served it's purpose up until now, however there's heaping amounts of evidence to show that PoW is unsustainable for the long-term. Not just in terms of environmentally, but also in terms of efficiency and centralization. If you'd like, we go into detail in our whitepaper (pages 7-10) with plenty of sources to back up our claims:
https://bithereum.network/storage/whitepaper.pdf. We also go into detail from pages 10-12 about how PoS helps solve this issue. Specifically Ethereum's Casper model which solves the issues of the nothing at stake problem as well as long range attacks and others.
The goal of starting with an ASIC resistant Equihash algorithm is so that for the year or so leading up to our PoS implementation, we hedge against the risk of a conglomerate hijacking the network and ensure it's long-term sustainability. While also allowing the average person anywhere in the world to participate in mining which gets crypto into the hands of more people and helps the goal of global adoption.