Just wondering if you guys could make it possible to bring-your-own compute resources and external IPs.?
Maybe would be easier to answer those separately since, I think own IPs might be harder than compute resources.
However, would appreciate any answer regarding this idea.
P.S.: we have some ideas and concepts for this. Would like to chat about this privately first before sharing every detail public.
We’re exploring this for larger customers, especially folks selling cloud services that need to worry about margins.
Surprisingly, bring-your-own IPs is simpler than other infrastructure. It’s not cheap, but we can announce /24s for people who need that today.
Compute is harder. We might never run our stack on other peoples’ hardware. We will probably run dedicated hardware for folks, though, and figure out a way to sell it at cost with a SaaS like fee alongside it.
Thanks @kurt.
Our motivation grows from the thought to get to more smaller regions with such a solution. Which are currently not yet available.
With some ideas to make them available to the whole ecosystem as a region.
Ah! The big hurdle to getting into new regions is anycast. It costs us about $15k/mo to add a new anycast region. There are cheaper ways to do it without anycast, though, they’d just require folks do DNS setups to route traffic there.
What regions are you considering? I can probably guess which are easy.
Oh, I thought it would be as easy as just putting more external IPs behind an anycast IP.
We already had the idea of using another DNS solution and route to them via some kind of Loadbalancer. However, I don’t want to host the apps there without the scheduler on other regions manually, on bare metal hosts.
We are currently looking into more European countries. The current list of regions to add in no order is:
Oh well the good news is, we are already live in Spain. 
You can add mad to an app today. We’re pretty interested in adding more EU regions as well, so stay tuned.
I missed that announcement. Added!
I also mark this topic as solved for now, since you gave pretty good input what to expect.
P.S.: check you inbox
not sure if resurrecting this old thread is the right way to go but I feel that the discussion will be better in a single place, let me know if I should rather create a new thread
I’d also be interested in bringing my own compute, so I’m wondering if any of what was said here has changed in the last three years
the use case is mostly for hobby stuff where I can get cheap hardware (either my own, or cheap vps) but I don’t really feel up for managing the software, and I’d rather pay fly.io $5 or $10 per month to abstract that for me, but still let me run everything on my own hardware, where I don’t have to share resources with anyone else and it’s under my control
This is still not something we’re actively working on. Our stack is pretty complex and it’s not trivial to run on arbitrary hardware in locations we don’t manage. We have the ability to provide dedicated hosts for customers that want to isolate their workloads from others - this is just hardware that’s dedicated to your orgs apps that run in our datacenters. It’s probably not what you’re after if you just want something for hobby stuff though.
yep, that wouldn’t be worth for my use case, thanks for the answer though! 