Right, now I'm actually watching the video, and I like when Sega Lord X does these type of 'alternative history' video game speculations. They're great thought exercises.
His HES idea is interesting, but the Neo Geo already had an HES model of sorts: the Silver SKU. A little looking and it shows that released at $399 at launch; expensive as hell compared to SEGA and Nintendo, but not $649 expensive. That model already had no pack-in and a standard joystick controller, so it's basically the HES he proposed while still retaining the memory card slot (something I'd rather SNK keep), minus the cartridge slot still being too big.
Actually, had the cartridge slot been redesigned for smaller carts, that'd of probably shaved another $50 off the SKU, and if SNK wanted market share, they could've taken a $50 or even $100 subsidy to hit $299 or $249. And that is definitely within reach of MegaDrive/Genesis and SNES pricing, for much more performance in 2D games.
But, then that introduces a pretty big problem. If you make the cartridges smaller, you're basically letting most of that processing power go to waste. Plus, the Neo Geo wasn't capable of scaling up sprites; it could only scale down. So smaller cartridges hurt it much more than systems like Genesis, PC-Engine or SNES. So you've got three choices, both of which add cost to the system:
1: Redesign the hardware to support hardware sprite scaling fully, and Mode 7 effects (to compete with SNES and soon-to-release SEGA CD, which both support it). Probably the costliest option on the backend, but mainly in R&D and some hardware modification changes. Can technically also be re-spun into the MVS platform to recoup the R&D costs, and would be the easiest option to do that with.
2: Release an add-on cartridge implementing full hardware sprite scaling, Mode 7 and maybe even rudimentary 3D processing capabilities. The issue here is outside of the simplest games, the add-on would be mandatory considering no support for AES-style cartridge sizes, and including the cartridge in HES SKU either means higher MSRP or more costs to eat by SNK. Also, not as easily re-usable in MVS hardware; it'd probably need to be implemented into an add-on board design.
3: Release a CD-ROM add-on. In this case, they can skip adding full hardware sprite scaling (technically speaking), since CDs can store more than enough data of any size in the era. But the CD-ROM add-on would need a decently sized buffer, I guess at least 512 KB, to make that approach feasible. Being a CD-ROM add-on, people are going to expect FMV support, so if the base system doesn't have that, the add-on would need it instead.
All of this would probably make it the most expensive option because CD-ROM drives & tech was not cheap for mass-market devices in 1991, and in case of an HES with support for only smaller cartridges and lack of full hardware sprite scaling, would be mandatory. Could also run into similar issues as SEGA CD with overreliance on shallow FMV titles (at least in the West) to try justifying it (and the HES as a whole).
Personally, I think a combination of 1 & 2 would be the best approach. Upgrade the hardware to fully support sprite scaling, rotation and Mode 7-style effects at the very least (processing power for rudimentary 3D could likely be optional), but do it through a board revision integrated into the hardware itself. Still lets SNK spin off that into MVS system revisions, and while it may add a bit more cost to the system w/ backend production costs, they can still afford to eat the added cost while targeting $299 or maybe even $269.
So a potentially aggressive $269, only $69 more than SNES's launch price (same year), same feature set but significantly more powerful 2D capabilities, and with good enough market adoption that price can come down in due time. IF they could throw in some processing power for rudimentary 3D graphics and keep MSRP around $299 at most, considering the ways they could advertise that as an additional advantage over SNES & Genesis....I'd of done so if I were SNK.
EDIT: Also SLX does mention compression helping with cartridge sizes, but the tricky thing is Neo Geo did not use tiles in the way other 2D systems did. It used sprites for everything, including background elements. Compression with sprites would generally not be as efficient as with tiles, though I'm sure many actual Neo-Geo games used cool tricks to squeeze compression where possible.
Really, it's the lack of upscaling sprites in hardware that would exacerbate the smaller cartridge situation for a more mass-market style Neo Geo system. You'd need to store sprites at the largest intended sizes on the carts, and figure out compression schemes from there. Indexed color palettes is one example (I know Neo Geo supports 16-bit color with 1-bit reserved for transparency on-off of a select color via color keying), but maybe supporting 12-bit color with no transparency bit would've been another option.
Or, focusing on graphics with 8-bit grayscale indexed palettes and smaller 8-bit or even 4-bit color indexed palettes (assuming sprites could support more than one palette...I'm not actually sure if that's something Neo Geo supported or not), etc.
EDIT EDIT: Also his suggestion of a SNK/NEC/Hudson Soft partnership is both interesting and ironic because, well, I've been thinking of & working on a concept of those same three companies in a partnership. But that one is more focused on a 5th-gen solution, not a 4th-gen console.
And it's more than just a console. But, I'm still working out a lot of the details. If I ever got enough vested funds to gather a team, would really like to make it a real product offering & business model. Hopefully I can talk more about it at a future date; I feel retro gamers and even people just feeling modern gaming's getting soulless & too "best practices" corporate, would really enjoy it.