There can still be use cases for onsite replication, it's all about the RPO. For example, if I have systems which I consider critical, I may want to replicate them from one storage to another, or one cluster to another, that way, if a system gets damaged, perhaps a Windows update gone awry, or ransomeware, or just a complete storage system failure, you can failover that VM to the other system right away. I've even seen customers replicate VMs on the same machine in some cases, just to have a very quick way to recover if something were to damage the running VM beyond repair.
Backup + Instant Recovery can get a similar RPO/RTO, but then you still have to move the system back to production storage so and, for larger VMs, or VMs with high I/O requirements, the performance loss might not be acceptable. Replicas don't have this issue, they are just another VM and can be powered on and running from the same or different production storage immediately.
So it's really all about what your requirements are. Can you live with the time it takes to restore your VMs, or is the performance of running your VMs from instant restore acceptable? If so, then backups alone are probably OK, but if not, then having a local replica can really be useful, generally being able to return a system, regardless of size, back to full production status in a matter of a few minutes.