Likely they had a license for the Intel chips, and then used that to make their own chips based on the Intel chips, but are technically not, and therefore able to sell without paying Intel royalties or whatever.
Unless you are very explicit, yes, derivative works aren't covered by normal IP licensing and case law was established very long ago testing interpretations. I don't have Lexis unfortunately to pull the dockets, but towards the end of this year I will, and I am going to try and provide as much summary research as possible via a blog - because silicon IP is entering our lives in a big way soon as architectural complexity demands standard latency mitigating by implementing software in silicon (Oracle is claiming that expression as a common law trademark since a couple of SPARC generations, so even our normal terms of ref are going to be fraught.)
Another reason silicon IP could become commonplace is chip shortage. It's not so far fetched to imagine a simplified computer without specialized coprocessors but with a single reconfigurable gate array allocated to the problem at hand.
Exposure to trade secrets is the most important standard problem with derivatives. Access to even extremely detailed and informative documentation doesn't necessarily convey infringement certain IP, but it is not the most popular pastime of a litigation lawyer to create strategies for layman's comprehensions of differences between novel creation non obvious to the skilled in the art and generically intelligible deductions reasonable to expect no protection from laws even if both parties went to enormous expense to negotiate and conclude the terms of access and application of the descriptions essential to contested ip.