Assuming reliability is the priority I would suggest going with Tailscale Funnels or a cheap VPS acting as intermediary.

I don’t have a lot of experience with dealing with GCNAT, but perhaps you could look into some solution with UPnP or RFC 6887.

object_store does indeed also support WebDAV among a variety of other protocols, Apache Druid or Apache Pinot probably would be better examples. My only experience with WebDAV is with Nextcloud and hasn’t been that great because it has been very slow, probably should look into it sometime.

EDIT: Apparently it supports CAS, and even has a locking mechanism

Scaleway, Exoscale, Cyso, Contabo, UpCloud, and others too

Many cloud providers offer S3-compatible storage, so it’s a common protocol to use in applications. There are even some databases like SlateDB that fully rely on object storage for everything. Supporting more API’s is extra work (unless you’re using OpenDAL) so most people pick S3 compatible API’s because they’re the most widely supported across all cloud platforms.

S3 isn’t just an AWS thing anymore. It has kind of become the standard object storage protocol, and almost every cloud provider uses it aside from a few the made their own API’s (e.g. Azure Blob storage)

Many cloud providers offer S3-compatible storage, so it’s a common protocol to use in applications. There are even some databases like SlateDB that fully rely on object storage for everything. Being able to have local S3 compatible storage is useful if you want the storage of your local machine while still doing so over a widely compatible protocol.

I tried that and my account got randomly deleted

I don’t think they only run on these chips. There are some companies in the US that provide Deepseek V4 presumably running on standard Nvidia chips.

Hetzner or Netcup

There is Cockpit which allows you to manage the server and has simple management for containers. However, I recommend using something like Dockge with compose because it makes it easier to change the configuration of containers without recreating them manually.

Companies like Claude with their AI subscriptions might be losing money, but I doubt paid-per-token AI API usage is not making them money. There are several companies like e.g. DeepInfra and Fireworks that have sprung up to sell specifically that. I don’t think simply multiplying API cost with expected usage is sufficient to estimate how much will be charged however, because I suspect that OpenAI and Claude currently have a significant profit margin since they seem to be the defacto duopoly in the US.

The chart above shows that quite clearly, the vertical axis is the combined score on various benchmarks. The horizontal axis shows the price. OpenAI and Claude do score higher, but the price difference is enormous, even if it wasn’t a log scale (70$ vs 1.3k$ for similar results!). The competition of these companies could drastically reduce the margins of US companies,

I therefore think the pricing will depend on whether the large US AI companies manage to lobby the government to enact laws to cripple the competition of Chinese companies under the guise of security.

Not everyone on the internet lives in the US