If you scroll down the DeadBro homepage, you'll find three badges that raise eyebrows: EU-hosted, dedicated bare metal, and not on AWS, Azure, or GCP. In 2026, running a SaaS on your own hardware is unusual enough that people ask if it's a philosophy or a stunt.
It's neither. It's arithmetic, with a privacy bonus. Here's why.
APM is a write-heavy business
An APM ingests data constantly and reads it occasionally. Every tracked request that hits your Rails app becomes a write into our system: traces, SQL timings, memory stats, GC data, queue time. On busy days that means millions of writes — when we moved 300GB of request payloads out of Postgres, we were handling close to 3 million writes per day against only a few thousand reads.
Now put that workload on a hyperscaler. Cloud pricing is built to look cheap at the front door and charge you at every hallway: per-operation fees on object storage, IOPS provisioning on databases, egress on everything leaving the building. A "write now, maybe read later" workload is close to the worst case — you pay retail for millions of operations whose data mostly just sits there.
On our own dedicated servers, the equation flips. The hardware is a fixed monthly cost, and NVMe drives don't charge per write. When that request table exploded from 10GB to 300GB in a month, the answer wasn't a five-figure cloud bill — it was standing up MinIO on a separate box we control and pointing the firehose at it.
Why that matters for your bill, not just ours
This isn't inside baseball. Our infrastructure cost is the main input to your price.
Most APM vendors charge per host or per agent, partly because their own costs scale with how much of someone else's cloud they're renting. Our costs are flat and predictable, so pricing can follow tracked requests instead — and a free tier can stay actually free, because an idle app costs us close to nothing. The moment we'd move to metered cloud infrastructure, someone has to pay for those millions of daily writes. That someone would be you.
Bare metal is also just… fast
A dedicated server gives you the entire machine: every core, all the RAM, NVMe with no noisy neighbors and no virtualization tax. We tune Postgres for the actual hardware it runs on instead of a t-shirt-sized instance type. For a product whose agent overhead is measured in microseconds and whose dashboards should feel live, predictable performance is a feature — and consistency is exactly what shared cloud tenancy struggles with.
The data side: your traces are sensitive
Here's the part that made EU hosting non-negotiable rather than just nice. Think about what APM data actually contains: URLs, request parameters, error messages with stack traces, user context. Monitoring data is production data wearing a lab coat.
DeadBro runs on dedicated servers physically located in Europe, operated by us. Your APM data stays in the EU, full stop. There's no US hyperscaler in the chain, which means no CLOUD Act ambiguity to explain and a much shorter GDPR story for your DPO: one sub-processor list, one jurisdiction, one clear answer to "where does this data live?" If you're an EU company that has been through a vendor privacy review recently, you know how rare that answer is.
The trade-offs, honestly
Bare metal isn't free lunch, it's a different lunch. There's no autoscaling slider — we plan capacity ahead and keep headroom. Backups, failover, and monitoring are our responsibility, not something a managed service abstracts away. And when hardware misbehaves, that's our pager going off, not a status page we get to point at.
We think that's the right trade for this product, for two reasons. First, the workload shape is predictable: ingest grows with customer traffic, not in random spikes, so capacity planning is boring in the good way. Second, we monitor systems for a living — running our own is not exactly outside the job description. We drink our own champagne: DeadBro watches DeadBro, uptime checks and all.
What this means for you
You don't pick an APM for its hosting philosophy. You pick it because it finds your N+1 queries and pages you when production breaks. But the infrastructure underneath is why the rest of the offer holds together: request-based pricing with no per-host fees, a free tier without a card, EU data residency by default, and dashboards that stay fast because nobody is sharing our disks.
Boring hardware, owned outright, in the right jurisdiction. That's the whole trick.