Knocking off the knock-offs
First, I got a good guffaw out of the line about entities that use governments to advance their own commercial interests. There isn't enough time in the day to list all the times Big Blue has benefited from the intervention of local, state, and federal governments around the globe.
The 'cheap knockoff' comment harkens back to the early days when Amdahl (backed by Fujitsu and then owned by it) and then Hitachi created clone mainframes that compete against IBM's System/370s some 35 years ago. As anyone who has watched the engineering done by Amdahl/Fujitsu and Hitachi in the long and strange mainframe market knows full well, it was these companies that often innovated ahead of Big Blue - as is often the case with a convicted monopolist even after it is compelled to open up its platform (as IBM was compelled to do with the 1956 antitrust decree and its amendments in 1969).
Amdahl had logical partitions on mainframes way ahead of IBM's, and Hitachi's Skyline bi-polar mainframes from April 1995 through October 1999 kicked the living guts out of Big Blue's System/390 bi-polar and CMOS alternatives. The final Skyline Trinium machines put putting 2,000 MIPS and 320,000 I/O operations per second in a single system image. That's good enough that they can still be sold in Japan more than a decade later and do useful work - provided you want to use Hitachi's software.
The fact of the matter is that the competition engendered by that consent decree helped spawn the modern computer industry because it illustrated that the United States government had limits on how much monopoly power it would tolerate not just with IBM, but with any computer platform provider. Monopoly is always measured against a relevant market, and the clever lawyers at Cravath, Swaine & Moore, IBM's hotshot, white-shoe law firm in New York, were able to argue to get that consent decree overturned in 1997, with a four-year window for sunsetting it.
Amdahl and Hitachi didn't exit the market so much as being pushed out of it. Once the consent decree was inactive in 2001, IBM would not have to sell its mainframe software on Amdahl or Hitachi boxes, and so it is not at all surprising that these companies mothballed their respective mainframe hardware lines in 2000. They did so not because there wasn't plenty of money to be made with mainframes - which there was, and certainly still is - but because it was abundantly clear that thanks to Moore's Law the money would increasingly come from the systems software, not the hardware.
Amdahl and Hitachi were not getting a cut of the IBM mainframe software action, so what was the point? At some point, you'd have to almost give the hardware away just to get the software and services contracts.
I think it is safe to say that IBM's mainframe market is moving at an increasingly glacial pace in terms of technology change. But then again, the same thing will start to happen with all server platforms once virtualization is sorted and fully baked into x64 chips. That is not against the law; and as a practical matter, is it against the law to tie an operating system to a device? Cell phone operators do it, Apple did it with its own variants of Power platforms and still does it with Intel-based systems.
Microsoft was gung-ho about having Windows run across MIPS, Power, Itanium, and x86/x64 platforms, but for all intents and purposes it is really just an x64 operating system, tied as much to the x64 architecture as IBM's mainframe software is tied to 64-bit System z hardware these days.
Just as IBM did when it argued to get out of the 1956 consent decree, Big Blue's lawyers will say that the mainframe is a small and battered platform in a vast and competitive market, and that this is the relevant market that should be used to determine if the company has monopoly control. If IBM has a mainframe monopoly, then Apple has a PC one, too.