Great hardware, billing department from 2003

I've been with Hetzner for eighteen months now. On the hardware side, the EPYC nodes and redundant networking have been solid — I measure 99.98% uptime on my end, which aligns with their SLA. :)

The billing department, however, operates on what I can only describe as an industry-standard legacy workflow. Invoices arrive as PDF attachments with no line-item breakdown. When I requested granular usage data, I was directed to an upstream portal that requires a separate login with a 48-hour provisioning delay. I escalated through three tickets before receiving a CSV export dated two weeks prior.

Has anyone successfully automated reconciliation with them? I'm evaluating whether the compute quality justifies the operational overhead.

Oh Petra, I feel for you. Before you spend another cycle on their billing portal, did you test your restore? :)

I keep a 3-2-1 checklist for every provider relationship:

• 3 copies of critical invoice data
• 2 different export formats (PDF + CSV)
• 1 offline copy because portals go down

Hetzner's compute is genuinely warm and reliable. Their billing feels like a backup you forgot to verify — looks fine until you need it.

Have you tried requesting the API access they quietly launched in March? It's buried in their partner docs.

Their "API" is a SOAP endpoint with no IRR-listed AS-SET for their filtering. I checked.

The actual routing is fine. They announce properly via their upstreams at Vultr and HostHatch, RPKI-validated, with sensible BGP communities for prepending. IRR entries match whois. No issues there.

What I find puzzling: they still require manual LOA for prefix changes, faxed or scanned. Most providers automated this via PeeringDB-linked portals years ago. The 48-hour delay Petra mentions aligns with their stated "industry-standard security verification," which appears to mean a human opens a ticket and clicks approve.

Not a technical problem. An operational one.

Our APAC team had a similar experience with Hetzner's Singapore presence. The hardware performed well for our latency-sensitive workloads, though the billing portal's timezone handling caused confusion — invoices showed UTC-5 timestamps with no localization.

We ultimately moved edge workloads to OVHcloud for this reason. Their billing automation was more mature for our region.

I do not have experience with their API. @GeorgeNmp's analysis of their routing matches what we observed.