Windows 1.0 most certainly did not have Win32; it had limited DOS support, and its multi-tasking was cooperative only (so one wayward program could lock up the whole system).
Like MS-DOS, early versions of Windows were closely tied to the x86 architecture. Towards the end of the 80s, a number of threats for the PC ecosystem were emerging. The first was RISC CPUs, which provided greater performance than x86-style CPUs, and led many to consider that CISC designs like x86 were a dead end (see How much better was DEC Alpha than contemporaneous x86? for example). The second was Unix, which, despite its fragmentation, was considered by many as the future of high-performance computing. In 1988, Bill Gates and Nathan Mhyrvold wanted to build a “Unix killer” operating system, capable of running on a variety of platforms.
As has been explained in answers to your previous questions, DOS and Windows couldn’t easily be adapted to other architectures, so this new operating system had to be developed from scratch. As luck would have it, Dave Cutler was unhappy at Digital at this point, and Microsoft hired him to build its new operating system, which would ultimately become Windows NT.
NT was not intended as a continuation of DOS and Windows; it was intended as the operating system of the future, starting from a clean slate (NT meant “New Technology”). The success of DOS and Windows ultimately meant that NT had to maintain backwards compatibility; but that wasn’t a design goal, and Cutler was against it. The goals for NT were vastly greater than those for DOS and Windows: it was supposed to support all kinds of workloads, from single-user desktops to multi-user servers, on a number of platforms, hosting a variety of application ecosystems (OS/2, POSIX-compliant Unix, Windows etc.), making the best use of the latest hardware of the time.
Why did Microsoft create a whole new OS in 1993: Windows NT 3.1? Which was roughly the same as Windows 3.1? (Win32 API/DOS support/multi-tasking)
Microsoft didn’t create a whole new OS in 1993, it was in development since 1989. Nor did it know in 1989 what Windows would end up like in 1993; Windows 3.0 and its success weren’t predictable in 1989.
Windows NT 3.1 was similar to Windows 3.1 only in that it shared a similar look and feel, general UI principles, and support for the Windows 16-bit API. All of that is veneer; the underpinnings of NT were utterly different to those of Windows 3.1. NT introduced the new Win32 API, which was subsequently ported over to Windows 3.1 (partially, through Win32s) and Windows 95 (with slight differences, notably around Unicode support). It also introduced fully preemptive multi-tasking between Windows applications; Windows 3.1 still relied on cooperative multi-tasking between Windows applications.
The Win32 API was important for NT when it was released, but its importance for the industry as a whole wasn’t a given back then. In fact when development started on NT, its primary target for compatibility was OS/2, and its main API was supposed to be an extended OS/2 API. The switch to an extended Windows API (which became Win32) happened after the sudden success of Windows 3.0. On release, NT supported a number of different APIs, and it’s only because the Win32 ecosystem “won” that others were dropped (to come back later in a different guise, in one instance). The future of PCs as a platform also wasn’t a given when NT was released, and it’s only because x86 (and then x86-64) won that other platforms were dropped. I’ve mentioned this before, but in 1993 it seemed that computing had many possible futures; see this issue of PC Magazine to get a feel for the situation at the time.
Both operating systems (DOS with Windows, and NT) evolved in parallel to support the dominant application ecosystem of the time. Windows continued to receive improvements, and technology from NT, to tide users over on lower-end computers until they became powerful enough to run NT; and NT also was improved to better support workloads that end-users cared about (notably, games). Once NT was capable of running everything most users needed, and became marketable as a desktop operating system, the old DOS-based Windows lineage was abandoned.
I think the existing Windows kernel already had multi-tasking. So it would not be so hard to port it to a new architecture.
The two are entirely unrelated. A kernel can support multi-tasking and still be extremely difficult to port; one example is Windows’s VMM. It’s also possible to have a portable kernel which doesn’t support multi-tasking.
So if it is possible to add x86 to an RISC kernel it should also be possible the other way around.
The NT kernel was designed to be portable from the beginning. The whole idea behind writing it for the i860 rather than x86 was to ensure that the developers’ familiarity with x86 wouldn’t leak into the design.