Many people are confused by classifying where aarch64-future will go .. while some think Tier1 is dependent on thunderX, others think it depends on supported (consumers-) - boards. Thinking, discussing & arguing is a complicated thing while the truth looks much better than we all think : IT´s all there we just have to use it
So Cloud-based aarch64- services plug the gap between expensive, rare metal servers and cheap consumer boards:
Example: You are working on a project with others, member 1 owns a Rock64 with fbsd12 installed, member 2 has access to a thunderX-Server with current-head installed and member 3 even doesn't have any access to relevant aarch64-based hardware… oh, that's really unfortunate, we cannot continue working on the project, so let's argue in mailing lists which is the best way to goto Tier1 …
O.K.- —- STOP HERE : all nonsense, we don't need to stop working, it's the opposite: we will even work 100 times faster from now on —— whatever argument may be heard as the best in mailing lists or wherever : IT IS PROBABLY WRONG
because all that is no problem in the real world of today :
Everything we need exists on aarch64 , from slow to fast, from expensive to cheap, from perfect to crap….
so reorder to go back to our project:
all three project-members can hook up VERY EASILY and immediately FreeBSD_aarch64-instances on e.g. AWS .
And some will be surprised.
FreeBSD supports AWS EC2-instances very well , just choose regions and checkout, what FreeBSD-versions are available as ready- to boot- AMI (Amazon Machine Image)…
and you know what ? it's incredibly fast compared to any available consumer-board….
you will then choose your needed storage which applies to your AMI, hosted on EC2 on your AWS-account .. and of course you choose RAM/CPU-power …..
We wanted 3 things in the project:
1. the exact same FreeBSD-Version(s) because we were working on a version-related issue 2.: We wanted to compile not only 1 instance of our port, instead we wanted 2 or 3 .. 3.: We wanted SPEED and save time…
to be honest: I started using it in a project to keep it cheap with 1 CPU and 2 Gigs of RAM, 10 Gig storage… but that was absolutely not what I wanted in the project…
So I very happily ended up in using the 16-core , 32 GIG RAM-instance and attached 50 GB storage… if you have started with the default 10 Gig and fail in storage - usage( what I did:-) : extend your S3-storage and do the :
service growfs onestart
And afaik there are even bareMetal-instances available…
While aarch64-EC2-instances on AWS are not cost-free, you can get it from very-low-cost and slow instances to real fast .
If you only want to use AWS temporarily(that's what it's made for) , don't forget to remove your running instances/storage/ from your account.
So this article isn't thought as an AWS-tutorial(not needed because wisywig there eis self-explained),
nor is it thought as advertising for a company which doesn't need advertising 
It is thought to remind you :
You can speed up development by using Cloud-Services occasionally.
Then you can
download your (AWS-) compiled project to your local slow snail- consumer - board
and continue there …
let's say you spent 3$ for a hammer-fast compilation but have saved 15 hours (working-) time :
Let's agree: not the worst deal you have made 