Saying most people don't need more than 25 Mbps is like saying 640k is enough for anybody.
The challenge is any definition of capacity (speed) requirements is only a point-in-time gauge of sufficiency given the mix of apps popular at the time & any such point-in-time gauge will look silly in retrospect. ;-) If I were a policy-maker in this space I would "inflation-adjust" the speeds for the future. In order to adapt to recent changes in user behavior and applications, I'd do that on a trailing 2-year basis (not too short nor too long a timeframe) and update the future-need forecast annually. And CAGR could be derived from a sample across multiple networks or countries. In practice, that would mean looking at the CAGR for the last 2 years for US and DS and then projecting that growth rate into future years. So if you say 35% CAGR for both US and DS and project out the commonplace need/usage then 100 Mbps / 10 Mbps becomes as follows below. If some new apps emerge that start driving something like US at a higher CAGR then future years automatically get adjusted on an annual basis. Of course 100/10 is an arbitrary benchmark for illustrative purposes, as is the suggested 35% CAGR. I suspect that in the case of US, the Internet will see much more significant growth in US demand and that new applications will emerge to take advantage of that & further drive demand growth (similarly for low latency networking). Jason DS 2022 100 2023 135 2024 182 2025 246 2026 332 2027 448 2028 605 2029 817 2030 1,103 2031 1,489 2032 2,011 US 2022 10 2023 14 2024 18 2025 25 2026 33 2027 45 2028 61 2029 82 2030 110 2031 149 2032 201 /eom