Apart from the obvious web3 stuff there are some interesting usecases for classical enterprise software.

Caveat, my experience is limited to clusters below 100 nodes.

You can run your private IPFS DHT if you want and the protocol has some cool load shedding features. There is also IPFS cluster which adds data availability features to your IPFS deployments. When we tried private IPFS networks the cluster project was relatively early and we found that managing DA in the application that owns the data ended up being more straight forward so we didn't use that but I've heard good things lately.

If you don't need to enforce access management across your internal data calls it can be an interesting alternative to stuff like minio or ceph. YMMV but my experience was that, running a private IPFS cluster can be easier and much more flexible than running a similarly scaled minio or ceph deployment. It can be seen as a shift left like DevOps is. Instead of having each internal app on your project depend on a Data team that manages Ceph or another data storage and access technology you run (most of) it yourself.

I’m actually using it for sharing enterprise components so that a document saved in one company can be sent to another company and the document loads content from IPFS. All this without needing a central server