And I agree with you. The current OS market is stale.
The POSIX era is closing, it fails to properly model non-serial or disk based data operations. Flowing data like tcp/ip connections are horribly handled in the Unix like model. Linux is a bloated monster with thousands of system calls, and the BSD's are stable but user/developer hostile (MacOSX included at a systems level Darwin has a lot of memory management brain damage).
The problem largely with drivers is a war you can't win.
Windows has backwards compatibility written into it at a very fundamental level. This was a good lesson Windows learned from IBM early on. If you write something for windows 95, it'll likely work with windows8. System32 is still System32. Most your API's, IPC, Draw calls, etc. All still exist, most are actually faster.
The biggest selling point of an IBM mainframe isn't that they are fast. It isn't they have 400+ PPC chips, each with 24 virtual processors. Its that they are binary compatible with IBM System360. If you invested 10 man years writing an application in the 1970's, it still runs today, without the need to write or fix things. It just works. You can have faith in IBM and those guys with gold medallions and bellbottoms who are now balding managers.
Does it fully make economic sense? No. The cost of IBM equipment, support, etc. Is likely way more then 10 man years of software time over 40 year history that application has been deployed.
The fundamental problem is purchaser ignorance. The purchaser may see there are more competitive solutions. But the chain of command doesn't want to invest, it wants to be comfortable. You aren't fighting silicon valley investors so much as you are fighting wall street.
Compare this to say Uber. Which solved a fundamental problem. Hailing taxi's is stupid. Lets make it easier. Ask for a taxi on your app and wait. The quality of your life has improved. The balance sheet looks the same (largely) a taxi costs X dollars. But my throat isn't sore, my arm doesn't hurt, maybe this system is better.
In The Valley the issue would largely be the same.
Linux is king.
If you actually do invent a better solution the Linux. You'll need to convince google, facebook, twitter, amazon, etc. To redeploy all of their applications to your platform. And at best you could offer maybe a 5% performance increase. 5%, to redo billions in software? That won't delay a single upgrade in the data center. Data center costs stay the same, software development costs go up. How does that balance sheet look?
I'll give you a hint. Bad. There is no market, or demand for it. At best it'll end up being another marginalized fringe OS.
P.S.: If you want to do start with Plan9 work up from there, make the kernel interface work like 3.x linux kernal so you can use linux code. Might work.
P.P.S.: Disruption is based on a better product. OS's can never be a better product. OS's fundamentally trade off speed for usability or the opposite.