It's interesting how his chart deflects somewhere around 80%, since at the 2017 Rocky Mountain IPv6 Summit that's about where T-Mobile US (who probably have the highest penetration of IPv6-capable endpoints of any ISP) noted their transition started slowing drastically.[0] The main reasons are that the remaining devices are old (e.g. 3G phones), have limited capabilities/needs (e.g. embedded IoT devices which just need to phone home once in a while), or are using odd configurations (e.g. for some reason T-Mobile doesn't currently support IPv6 for tethering).
Another good talk from that conference is about how Carrier Grade NATs are (amazingly) even worse in practice than they sound in theory[1]. During the Q&A portion one of the audience members points out that as technologies like CGN become required to provide IPv4 service it will cause the cost of supporting IPv4 to rise over time and naturally push ISPs to kill IPv4 as the use of v4 drops while the costs to support it do not.
I can also recommend a short talk by Cloudflare about their experience serving IPv6 traffic[2] and another on the history and future of IPv6[3].
Another good talk from that conference is about how Carrier Grade NATs are (amazingly) even worse in practice than they sound in theory[1]. During the Q&A portion one of the audience members points out that as technologies like CGN become required to provide IPv4 service it will cause the cost of supporting IPv4 to rise over time and naturally push ISPs to kill IPv4 as the use of v4 drops while the costs to support it do not.
I can also recommend a short talk by Cloudflare about their experience serving IPv6 traffic[2] and another on the history and future of IPv6[3].
[0] https://www.youtube.com/watch?v=nNMNglk_CvE
[1] https://www.youtube.com/watch?v=fbk4H6EmZzI
[2] https://www.youtube.com/watch?v=2RGzR5eMfcg
[3] https://www.youtube.com/watch?v=p_HbX-batjs