I am a big fan of go big or go home. With my current setup, none of those devices would work at all, for many reasons.
The Celeron NUCs do not support more than 8 gigs of ram, and they are slower than you think. Of course you can run HA on them, but if you want to get real and use it as a VM/container platform… go BIG.
My 6 year old machine is a quad core Haswell Xeon workstation, it was upgraded from 8GB to 32GB last year because I was starting to run out of mem, this was even before I started using containers (not using VMs). Even though it is 6 years old, it is several times faster than that Celeron NUC. I am sitting around 20% CPU, which would max that little thing out, and the HA ecosystem will require me to add more containers and VMs as I add more features. After some more ram and a few SSDs, this thing still has capacity to play around.
Currently running
Ubuntu 18.04 with 20.04 kernel
Docker: HA, Rhasspy, BlueCherry with motion detection
Databases: Prometheus, MongoDB, MySQL
DNS, Filtered DNS, NTP, NUT, Duplicati, Mosquitto
Grafana, Netdata
Unifi controller, Adaptec RAID controller
Samba and NFS file servers
You do not have to go as big, but maybe there is a good middle ground. The 8th gen NUC has the Iris Plus graphics ( NUC8i5BEH), which you wouldnt think would be a big deal if you are not using a monitor, but it has 128MB EDRAM, which acts as an L4 cache, makes them very quick for multitasking and running lots of stuff. It is 3 times faster single threaded than that Celeron, and 4 times faster multithreaded, and even more for graphics and applications that support AVX, which the Celeron does not.
If you end up wanting to do more, you will very quickly regret getting a Celeron, especially if you want to run VMs, which need allocated RAM