I agree insofar as that there is no dedicated runtime for crystal in AWS Lambda, but it can well be used there or in fargate containers. I wrote a (proof of concept) minimal runtime for Lambda some time ago [1] and had pretty decent results.
Given that both fly.io and Heroku have working buildpacks for Crystal/Lucky, I’d say you can serve pretty heavy load before you’d need to reach for more specialized machines. Of course, that depends on the nature of your application as well. Maybe my understanding of 'large scale' is different from yours, though.