[plett]

Because dual stack is a hassle. It doubles your operational workload - which is a big deal in a large network. IMHO, we'll only see large scale IPv6 adoption when it becomes possible to deploy an v6 only network and translate to/from v4 at the edge for legacy edge cases.

Facebook already does this - everything is IPv6 only inside their datacentres, and the only dual stack devices are their load balancers at the edge. They had to do this because they were running out of RFC1918 address space to use internally for IPv4, and the traffic was already going through the dual stack load balancers anyway.

On the access side, mobile networks are also going v6 only. EE here in the UK (sometimes) only gives you an IPv6 network and uses 464XLAT to NAT you when you want to reach an IPv4 only address. They can do that because they know what devices all their users are using, and can whitelist recent versions of both Android and iPhone which have CLATs in them which makes 464XLAT possible.

Fixed line residential and business connectivity is much harder - laptops and smartphones will work just fine with IPv6 only, but people expect their games console or old networked print server to work, and frequently they only support IPv4. What's needed there is islands of IPv4 on their LAN and a 464XLAT compatible CLAT inside the customer's router - then the entire ISP network can be v6 only apart from a few NAT64 devices at the edge. I'm sure that will come, but it hasn't happened yet.

It will happen - mass adoption is inevitable IMHO. There are new internet users and services being added every day, and there aren't enough IPv4 addresses for all of them. It's either IPv6 and 464XLAT or IPv4 only with CGNAT. And CGNAT is expensive - keeping state for all those connections makes for expensive boxes with lots of memory.