It's hard to generalize, except to say that all of the technologies you name may be used in appropriate situations.
For example, we know that NVMe has higher throughput and lower latency, but spinning disks are still a lot less expensive, and for an application with modest traffic, a spinning disk might be adequate, and a good way to save money.
Also the storage performance isn't so important if the whole dataset for a given application fits in RAM and is largely read-only. In other words, the I/O traffic is so infrequent that you don't need a high throughput technology.
So the question "what are most people using" is more or less irrelevant.
It's more important to consider what your application needs. That depends on many factors, like data size and query patterns for your application.
That said, cloud database platforms tend to provide storage hardware at the higher end, because they kind of have to assume any application might need that performance. Also because in the cloud, they almost always have multiple customers' containers sharing physical hosts. AWS offers "dedicated servers" but almost no one uses them besides government clients.
