Anyone with a load on nodes, GPUs, or provers eventually hits this decision: run the infrastructure yourself or hand it to a team like us. There is a line between these options, and it is not where people usually draw it. We answer this on every first call, and we answer honestly, because there are cases where you should not hire us. Below is how we work through it.
The cheap part is misleading
First you have to see what the cost is even made of. Renting hardware from a hoster is the visible, cheap part; the real money is in the operational layer: on-call, observability, incidents, the discipline around keys. We broke this down in detail in the post on node TCO without the price list, and the key takeaway is that the cheap price-list line covers a small fraction of the real cost, and the "yourself or outsource" decision is precisely an argument about the operational half.
So comparing "rent hardware for X" against a managed team's price is meaningless. You are comparing the visible cheap part against the full cost, and under that comparison self-hosting always looks cheaper than it is.
The line is not fleet size, it is 24/7
The main myth: "once we grow to N nodes, then we'll think about a team." The line is not in the number of nodes but in the coverage requirement. The moment the infrastructure has to run 24/7 and its downtime or slashing costs real money, you need round-the-clock on-call, and sane on-call is a minimum of three engineers: one on shift, one on backup, one resting. That floor does not depend on whether you have 6 nodes or 60.
And there is the whole arithmetic. Building that team yourself runs, on the 2025-26 market, from 26 thousand dollars a month in the cheapest jurisdiction to 78 thousand in the US, flat, at any fleet size. On a small fleet that floor makes your own team the most expensive option per node, not the cheapest. One admin is cheaper, but one admin is not coverage, it is a single point of failure that burns out and one day sleeps through an incident.
When running it yourself makes sense
Honestly: there are cases where you should not hire us.
- Infrastructure is your core competency. If you are an infra company yourself and SRE is what you sell, keep it in house.
- You already have a team of 3+ strong engineers for round-the-clock coverage, with room to spare. Then the floor is already paid, and outsourcing duplicates what you have.
- The load is non-critical. If a few hours of downtime costs nothing and there is no slashing risk, you do not need 24/7 on-call, and therefore not a team for it.
If any of these is you, we will say so on the call. Selling coverage to someone who does not need it is a way to lose a client in six months.
When to outsource, and what you buy
Outsourcing makes sense when infrastructure is critical to you but is not what you sell. You are not buying "admins," you are buying an already-staffed round-the-clock rotation that several clients share, so it costs you a slice of a team rather than three full salaries. The same 3 to 4 engineers on your fleet we keep for around 15 to 20 thousand dollars a month, flat and at any fleet size: cheaper than building a rotation in even the lowest-cost jurisdiction, and it already works rather than being hired over six months.
Beyond the rotation you buy the discipline that does not ship with a rented server: failover that cannot double-sign, monitoring that wakes you for a reason, and the understanding that slashing is not about uptime.
No vendor lock-in
The main fear in handing infra outside is getting stuck with a contractor. So we come in and leave reversibly: configs, runbooks, and infrastructure as code stay with you, and off-boarding with knowledge transfer is part of the contract, not "we'll write the docs later." If you decide to take the infra back or switch teams, you carry away a working state, not empty keys. Vendor lock-in is zero by design, because otherwise trust in outsourcing infra does not get built.
How to tell where you are
A short test: does your infra need to run around the clock, and does its downtime cost real money. If not, run it yourself. If yes, cost the operational floor as a fixed team cost and compare honestly: your own rotation from 26 thousand a month against a slice of one already running.
If you want to work through your specific case with no sales pitch, the first call is with an engineer, not a manager: that is what we cover through operate. Get in touch, and we will tell you honestly where your line sits, even if it is not in our favor.