Let's assume that the successor to the Switch takes on the same form as its predecessor; that is, it's a handheld/console hybrid, with more or less the same power consumption, and that it will come out in 2023 or 2024. We'll call this the Switch 2. Now what if I told you that I think it is perfectly reasonable that this device can be capable of running even fairly demanding games at 4K 120fps? First, let's have a technical primer that I am totally ill-equipped to give:
The Switch uses the Tegra X1 chipset, made by Nvidia. Nvidia is probably best known for their desktop GPUs, such as the RTX 20-series cards. The Tegra X1, based on a 20nm process, came out in 2015. A revision based on a 16nm process came out in 2019 and was used for Switches made since ~August 2019 and every Switch Lite. While more powerful than the original, Nintendo chose to forego that benefit for compatibility and instead only enjoy the lower power consumption. Let's assume that Nintendo sticks with Nvidia and has them provide the chipset for the Switch 2.
Recently, Nvidia unveiled version 2.0 of their DLSS (Deep Learning Super Sampling). This uses machine learning to train AI to upscale low resolution images from games into high resolution ones. In other words, a game can render at 1080p and DLSS can output a 4K image. Now you may say that that's not truly 4K, and won't look as good as native 4K. Indeed, I said the same. But it seems Nvidia really delivered on their promise with DLSS 2.0! Take a look at this Digital Foundry video to see it in depth: https://www.youtube.com/watch?v=YWIKzRhYZm4 The takeaway here is that DLSS 2.0 is basically free performance, better quality, or a mix of both with basically no catch. Well, the catch is that you need a 20-series card since it requires Nvidia's Tensor (AI focused) cores.
Now fast forward to 2023. Nvidia has by now released a new mobile chip with Tensor cores as well as DLSS 3.0, which can near flawlessly upscale 4X (1080p to 4K) with an even smaller performance cost. I'll skip the AI argument (though will make it upon request), but I think this is more than reasonable to expect by then. In other words, this means that for the Switch 2 to output 4K@120fps, the system actually only needs to render at 1080p 120fps. So can we expect that? With said mobile chipset likely being based on a 7nm process and keeping a 15W power limit, I think so. Might have to skip ray-tracing, though.
P.S. For those of you wondering about handheld mode, let's say the Switch 2 screen is 1080p. This means that 4X upscaling will only require a 540p render, which as the DF video shows, actually works great with DLSS 2.0. So the introduction of DLSS may actually be a greater benefit for handheld mode!