2W, maybe. Pico, no. Driving a payment system and displaying an interactive video is still a job for a SBC not a microprocessor, even a modern one.
Raspberry Pi have found a price level and form factor that works for customers. The raw speed of the 5 is probably overkill for most user applications (the 4 is too IMO) - though edge AI and robotics need it, so there is demand. But ultimately the speed is there because a faster ARM chip fits within the cost-power-size requirements of the board.
Pico was meant to shine for much of the education half (e.g. micro-python controlled mini robots without the hassle of complex and poorly supported developer environments, OS management and configuration, or hardware complexity at a couple $ cost point), not the interactive media half (which would target the Zero 2W usage case over a full blown PC style Pi).
The compute module variants do tend to shine in the edge robotics niche where you need to do some processing in a custom form factor and don't want to design or source a SoC directly... but you don't really need the "rest of the PC" the standard Pi models provide. Pis (any model) are a bit of a shit fit for edge AI in general though. The memory bandwidth is god awful (even compared to cheaper SBCs) and the hardware to meaningfully offload matrix operations just isn't there like it is in a lot of other embedded boards fit for purpose. This leaves you spending more $ to use more Watts for less AI compute vs typical embedded AI options.
The Pi foundation's big break was targeting ultra low cost markets (i.e. sub $50 TCO) with a well supported device that has/had a lot of hacker community mindshare. Any product they've made beyond that focus (i.e. not the embedded or microcontroller variants) has been a pretty bad technology fit which ends up in more drawers than deployments despite the hype.
