Your description is correct for what it promises in the future.
But today, if you are using NVMe, and you are using Win11 (it has new NVMe kernel IO method support), and you are using DirectStorage, then you will get...Everything still loaded through main memory via the CPU.
Now you will get a performance benefit of it not having to be copied into user-mode memory space on the way to the GPU, it can stay in a Kernel IO buffer to be copied direct to GPU, not via user-mode process memory space.
You will also get a multiplatform API to use, that maybe smaller indie game titles can consume easily to get the benefits on platforms that support (such as XBox).
If you are on Win10, or not on NVMe, then DirectStorage still uses ReadFile in chunks through the process address space, just like before. But you still get a nice new API to use.
Now in the future Microsoft DX12+ team hopes that GPUs will provide optimisation (new silicon) and support to allow for NVMe to GPU transfers over PCIe directly. But today the closest thing is the proprietry Nvidia RTX IO. Or as others have said the AMD card that has storage built into it.
There is another news reports within hours of DirectStorage GA release citing Linux and HPC use of GPU for computation and I think ultimately that will be the driver to getting in GPU silicon a general purpose mechanism for volume data transfer directly from storage that I'm sure DirectStorage will consume. But it is all a few years away still.