AWS is more often than not a better solution than colo when factoring in the on-site engineers, techs, and operational complexity costs a company will pay to monitor and respond to hardware-related events. One could build out a datacenter management team with on-call engineers, or, they could pay AWS to handle all that, and focus on innovation and products that make their company unique and (hopefully) profitable. AWS makes a lot of sense for companies that wish to inoculate themselves from the hardware layer, and it would probably take a company many magnitudes larger than Heap to realize any real benefits from self-hosting at a colo. This isn't even considering the fact that uptime matters, and you'll need more than 1 colo to really do it right.
I say this as someone who built, manages and operates datacenters and colo spaces.
I say this as someone who built, manages and operates datacenters and colo spaces.