No comparison of speed.
WPEngine refugees represent my highest client density. Their tech is great for low/no traffic sites, then circles the drain under high traffic load. Also you have zero access to system logs, so no way to debug how to fix the problem.
Kinsta I haven’t used.
If you have a low traffic site, you can use any hosting.
If you expect to have significant traffic, the consider dedicated iron/servers (OVH or KimSufi).
Dedicated servers require you do your own admin or hire someone, so you will have additional costs.
You other option is to look for LXD hosting, which provides the best of all worlds, so long as the hosting company personnel know what they’re doing.
Look for someone who can deliver between 100K-1M+ requests/minute throughput for your WordPress site, when running h2load on the command line.
I use LXD for 100s of sites + the 1M+ requests/minute is very doable on garden variety OVH machines.
@Pothi brings up a good point about mod_brotli which also applies to every layer of tech.
The only way for you to control this + use latest tech, is by using dedicated servers.
For example, on 2018-06-29 PayPal began rollout of banned use of TLSv1.0 + TLSv1.1 SSL protocols.
For all my LAMP stacks, I fixed this months before the apocalypse date hit.
The fix, a single line change in my global Apache config file.
Many big sites were hit by this. For example, Ontraport forms all broke, when PayPal rollout was applied to the Ontraport site.
My point. The only way for you to control your destiny is admin your own machine or have someone smart admin your machine or LXD container for you.