However I have to admit that we don't necessarily run the "target" applications that Vercel was really designed for. We have a full suite of Lambda backend services that communicate with each other in VPCs and over SQS. We have scheduled executions using CloudFront and listen to events happening in several other AWS subservices.
Now, we also have two "simpler" React applications that we have actually test-driven on Vercel before, but the cold start time of functions was honestly quite horrible and detrimental in comparison, and the problem that really drove us away very quickly was the lack of integrated database and authentication solutions (e.g. AWS Amplify/DynamoDB/Cognito and Firebase/Firestore/Auth). I know about Railway and 0auth, but we have no interest in using several different providers with uncertain confidence in them just for basic application logic, especially because they don't integrate back into Vercel at all (like e.g. functions being called when a DB record changes) and seem to be more expensive than even AWS itself.
If Vercel was more like a "Heroku for serverless", and at the very least offered some of these basic features directly, and then had some good ways of connecting them together, we would potentially reconsider.
However I have to admit that we don't necessarily run the "target" applications that Vercel was really designed for. We have a full suite of Lambda backend services that communicate with each other in VPCs and over SQS. We have scheduled executions using CloudFront and listen to events happening in several other AWS subservices.
Now, we also have two "simpler" React applications that we have actually test-driven on Vercel before, but the cold start time of functions was honestly quite horrible and detrimental in comparison, and the problem that really drove us away very quickly was the lack of integrated database and authentication solutions (e.g. AWS Amplify/DynamoDB/Cognito and Firebase/Firestore/Auth). I know about Railway and 0auth, but we have no interest in using several different providers with uncertain confidence in them just for basic application logic, especially because they don't integrate back into Vercel at all (like e.g. functions being called when a DB record changes) and seem to be more expensive than even AWS itself.
If Vercel was more like a "Heroku for serverless", and at the very least offered some of these basic features directly, and then had some good ways of connecting them together, we would potentially reconsider.