B4X products are for the development of Apps that can work on Android, Desktops (and any OS that can handle Java), and iOS.
There are no specific limits on the actual target CPU type of the Apps, including ARM.
So it doesn't matter that development takes place in the Desktop. In fact it is a great advantage.
However, you mention Graphic processors that aid in speeding up parallel processing used by AI models.
B4X does have an interface with Python that can use special libraries to do that for Windows and Linux.
You are right that there are very few existing interfaces with the upcoming RTX chip of NVidia for ARM technology.
But I believe that time will change that.
"Nvidia's ARM-based RTX chip technology is called the
NVIDIA RTX Spark, a high-performance system-on-a-chip designed for
Windows on ARM laptops and compact desktop systems. Unveiled in mid-2026, it pairs a custom 20-core ARM CPU with a Blackwell-architecture RTX GPU and unified memory"
Updates to Python Keras and Tensorflow libararies will follow. B4X's interface with Python will let you use them on
Windows systems.