I rely on Restic at the moment which seems to need fast read access to data, but their incremental snapshotting is great. It'd be ideal if I could find something like that supporting these "cold storage" solutions.
There are 3rd parties that will do it for you (Iron Mountain is at least one) but that's an extra cost and Google takes no responsibility for it. I assume this is an example of a place where Amazon is able to leverage its wholistic business, with a Cloud service that can also take advantage of their physical logistics system. Google's service here is quite significantly cheaper and has some nice features though, but even if it's not worth a $4/$1.23 premium for Amazon I could definitely see continuing to pay Amazon some premium ($2 vs $1.23 say) for that alone anywhere with limited high speed WAN availability.
We also have a Transfer Appliance , that comes in two sizes (100T and just under 500T). We don’t currently support shipping one filled up with your data for recovery/export though.
You can also request a "B2 Fireball"  from them. It's basically a small array that they mail to you for $550 with 70tb of storage. You fill it up and send it back to them within the month, and they'll load the data into your account.
 https://www.backblaze.com/b2/cloud-storage-pricing.html (Bottom of the page)
Amazon's equivalent to B2 Fireball is "AWS Snowball" (amusingly enough, not sure if there is a bit of fun name riffing between the two here), which is a service fee of $200/50TB and $250/80TB device, any onsite days after the first 10 at $15/day.
It's interesting how the pricing mix is on this feature though. Amazon offers lower potential ingress pricing depending on your use, though notably if you kept the Snowball a whole month the pricing would get very close to the Fireball (+20 days @$15/day brings the price to $500/$550 respectively, though the former with 20TB less and the latter with 10TB more).
Backblaze and Google are both much cheaper to get data out of though, Amazon's Glacier and descendent services remain very much deep freeze focused.
There may also be a minimum storage period, like Amazon has.
Let's wait and see.
The actual product is pretty painful when you need to do a recovery, especially if you don't know where the file lived on disk. I haven't tried newer Arq Cloud Backup destination to see if it improves the search experience.
That said my experience is from more than a year ago and I would try it again if they were able to bring their search on par with current consumer backup offerings.
If you only backup from a single machine it has a local cache of already backed up data, this has the large advantage that it basically only needs to push the delta data to the remote, not do any kind of synchronization to check what is already there or not.
... with one very big difference - you can only point borg at an SSH host. You can't point borg at S3 or B2 or Glacier, etc.
rsync.net supports both borg and restic, but even the heavily discounted plans are much more expensive than "Cold Storage" or Glacier, because they are live, random access UNIX filesystems ...
Also worth pointing out that my storage is calculated after compression and deduplication. So depending on the data a Borg backup can be much smaller than the actual data.
Restic seems more made from the ground up to utilize the existing power of a filesystem as a database, so it needs remote storage that offers quick interactivity (esp. checking existing files), i.e. it's impossible to use something like Glacier as a backend.
It's not a problem for me since I just backup to a local drive and (am planning to setup) synchronization to a remote dumb storage.
Actually, since it’s google I likely wouldn’t consider them regardless.