The stuff I'm looking at are subscription-based that we use at my company... Wood Mackenzie, IHS, S&P Global, etc., so I can't really link them here unless you have a subscription.
In their latest long-term forecasts, Wood Mackenzie has nat gas power demand falling from 28.7 Bcf/d in 2026 to 28.4 in 2031; S&P Global has it falling from 29.7 in 2026 to 23.1 in 2031. IHS has it peaking a few years later (2029) but falling after that. Of note, IHS's long-term forecast is the most stale (from February) and the other two both revised down 2026-35 power demand in the most recent updates so it will be interesting to see if IHS follows suit in their next update.
And to be clear, I'm talking about L48 demand, not world demand.
After working with this data for internal forecasts for the past few years, I just feel like there is a LOT of recency bias when prices make dramatic moves. Like in the summer of 2020 I remember getting into arguments about whether or not gas would ever get above $4 again because it was below $2 and there was nothing above $2.40 or so on the forward curve, even though forecasters kept talking about production was going to fall and prices would rise. Now I'm having the opposite argument where nobody can imagine it going below $5, lol.