Maybe. It's certainly possible, though I don't personally think so. My gut says it's another mixup based on most logical assumption, with another system coming up soon.
Lite has the same exact Mariko SOC as v2, capable of 25-30% superior performance over the original switch, it's why it still gets a couple more hours battery life despite having a smaller battery. There would have been zero difference to them looking at components like the soc, which they logically prioritize, from the parts suppliers, which is where they tend to get their upcoming information from. Game and watch has a cortex STM32 while less powerful than the switch is an incredible understatement, its overpowered for a little box that plays an NES game, and has a clock, and whatever else it does. A goofy very Nintendo esque utility I don't think anyone could have seen coming by looking at boxes of parts being shipped and kibitzing about it.
Definitely seems to fit the bill for a casual handheld reminiscent of Nintendo's legacy portables, maybe something like a gba classic, better than calling the soc that is 25-30% more powerful than the handheld that was currently playing doom 2016 and Witcher iii the casual model.
But that's just my interpretation.
The reporting was about a new, cheaper model of Switch releasing in 2019 to act as a successor to the 3DS's market. I don't see how Mochizuki could interpret a supplier telling him 'They're buying STM32H7 chips that are no more powerful than the OG 3DS CPU and have no hardware accelerated graphics*, and obviously are incompatible with Switch software,' as 'Budget Switch coming later this year.' It's far more likely Mochizuki heard not just about the die shrink, but also about the smaller battery, or smaller screen, or the non-detachable controller and smaller shell, et cetera.
Nintendo probably chose that particular chip for the G&W SMB because it was the cheapest**, most easily sourced microcontroller that could handle emulating the NES without even a hint of slowdown. I'm sure they could have managed with an M0+ based chip, based on posts online about people emulating NES on RP Pico, but going with an M7 probably saved more time (and thus money) not having to optimize the software as much as a weaker processor might require, than they would have saved on a slightly cheaper microcontroller. Remember, it was a limited edition release, so saving pennies each unit isn't as important as saving work-weeks of expensive engineering efforts.
*No hardware accelerated graphics in the sense we tend to think of when discussing video game platforms. It does have a display controller of course, and a "graphic oriented memory management unit", as well as a JPEG codec, the latter of which could technically count as it is hardware accelerated and does relate to graphics, but unless they designed the emulator to render the framebuffer as a JPEG file for some god-forsaken reason I doubt the codec was used for anything. I'd guess it's mostly the same NES emulator as they used for the 3DS virtual console (since they're both 32-bit ARM systems, granted one is ARMv6 and the other ARMv7), but that's baseless speculation on my part.
**I couldn't find exact prices, but the STM32H7B0 used in the Game & Watch Super Mario Bros is listed as the most power efficient option in the 'value' section of the line-up, as well as the only 'value' option with the graphics optimized MMU.
Sorry for having such a long response to your rather innocuous theory, I just love geeking over Cortex M and this thread has never given me the chance to do so until now. So. . . Thanks!
