They buy them;
Just as Amazon (and other big ones) do.
Have you ever compared the prices of these VPS to IPv6-only VPS and wondered why IPv6-only ones are so much cheaper?
By definition there are only ~4 billion IPv4 addresses available. No matter how fair you distribute them; that is simply not sufficient for todays -let alone future- demand.
This topic is temporarily closed for at least 4 hours due to a large number of community flags.
This topic was automatically opened after 22 days.
The client can always select to use “server -4 *.pool.ntp.org” to get IPv4 only, or “server -6 *.pool.ntp.org” to get IPv6 only. There is no need to break the geographical structure by adding a special IPv6 subdomain.
True, but still i’m looking forward to have ipv6 enabled across the board, instead of relying on client config (which might not always be possible on some very simple IoT clients).
I suspect i might have to reduce the server bandwidth setting when a flood of ipv4 clients will be routed to my ipv6 servers 
On client-side I use a self-hosted instance of AdGuard Home (Works similarly to Pi-hole but is better in every way) to redirect IPv6 DNS lookup requests to NTP’s 2.(…) IPv6 addresses. For the main pool•ntp•org domain, I use:
|pool.ntp.org|$dnstype=AAAA,dnsrewrite=2.pool.ntp.org
|0.pool.ntp.org|$dnstype=AAAA,dnsrewrite=2.pool.ntp.org
|1.pool.ntp.org|$dnstype=AAAA,dnsrewrite=2.pool.ntp.org
|3.pool.ntp.org|$dnstype=AAAA,dnsrewrite=2.pool.ntp.org
This workaround is something that most client users out there would be unlikely to set up successfully, however.
Interesting observation:
Ubuntu is planning to replace systemd-timesyncd with Chrony starting with release 25.10.
It will come pre-configured with their own NTS-enabled servers, which also happen to support IPv6. In the default configuration, the traditional NTP Pool servers are still present, but commented out, so they remain available for those who prefer them: