Most hosted fleets sit entirely with one provider. That is usually not a decision. It is what happens when the first contract works well enough that nobody revisits it, and the second batch of machines goes where the first batch went.
It is worth revisiting, because a single-provider fleet carries a risk that does not show up anywhere in the rate. The hosting rate prices power, space and labor. It does not price what happens if that one relationship stops working.
What concentration actually costs you
The failure modes are not hypothetical, and they are not all about a host going under. A provider can stay perfectly solvent and still become the wrong place for your machines.
A site can be curtailed harder than you expected for a season. A facility can be oversubscribed, so the power you contracted for is quietly rationed. A change of ownership can bring a change of operating culture. A local regulatory change can put the whole site under a restriction that has nothing to do with you. A dispute over an invoice can escalate to suspension.
In every one of those cases, an owner with machines in one place has the same two options: accept the outcome, or move the entire fleet at once under time pressure. An owner with machines in two places has a third option, which is to move some of them, slowly, while the rest keep hashing.
That third option is the whole argument. It is not that a second host is better. It is that having one changes what the first host failing costs you.
The case against splitting
It is only honest to put the costs on the table first, because for a lot of fleets they outweigh the benefit.
Splitting costs you volume. Hosting rates step down with scale, and two half-sized deployments will often price worse than one whole one. If your fleet is small enough that it sits at the bottom of both providers’ rate cards, you are paying a real premium for the diversification.
It costs you attention. Two contracts, two invoices, two sets of reporting, two relationships to maintain, two escalation paths to learn. If you are running the fleet alongside another job, that overhead is not trivial.
It costs you leverage at renewal. A provider who hosts your entire fleet has a reason to keep you. A provider who hosts a third of it has less.
And it can cost you minimum order quantities. Some facilities will not take a deployment below a certain unit count or kilowatt figure, so a split that looks sensible on paper may not be available at the size you want it.
Where the line usually sits
There is no universal threshold, but the question that tends to settle it is simple: if this fleet went dark for a month, what happens to you?
If the answer is that it would be annoying, concentration is probably fine. If the answer is that it would breach a loan covenant, miss a payment you have committed to, or end the business, then you are carrying a risk that a rate reduction does not compensate.
The other useful framing is proportion. Splitting a fleet in half is rarely necessary. Moving fifteen or twenty percent of the machines to a second provider gets you most of the benefit, because it gives you a working relationship, a known escalation path, and somewhere that will already take your calls if you need to move in a hurry.
That second relationship is worth more than the machines in it. It is an option you have already paid for.
Choosing the second host badly defeats the point
Diversification only works if the two providers do not fail together. Two sites on the same grid, in the same regulatory jurisdiction, fed by the same utility, exposed to the same weather, are not two bets. They are one bet written twice.
Worth separating deliberately:
- Grid and region. Different interconnection, ideally a different state or market.
- Ownership. Two sites under the same parent company share a balance sheet and a management decision.
- Power structure. One site on a fixed rate and one on a market-exposed rate behave differently in the same month, which is the point.
- Climate. Two sites that both derate in the same heat wave do not give you much.
Apply the same due diligence to the second provider as to the first. A weak second host does not reduce your risk, it adds a second way to lose machines. The selection process does not get a shortcut because the deployment is smaller.
What makes a split workable in practice
Two things do most of the work, and both have to exist before you need them.
The first is a serial-level asset register that spans both sites. Split fleets are where ownership records stop being administrative and start being load bearing. You need to be able to say which machine is where, in writing, without asking either host. If the two providers hold the only record of what you own, you have not diversified your counterparty risk, you have doubled it.
The second is contract terms that let you actually move. Notice periods, early termination, and the retention language that determines whether leaving is a logistics exercise or a negotiation. A split that you cannot execute is a theoretical comfort.
Beyond those, the practical details are familiar from any multi-location setup: consistent pool and monitoring configuration so the two populations are comparable, one place where both sets of numbers land, and a clear view of what each host is contractually responsible for. Much of the operational thinking in our guide to multi-site operations applies, with the difference that the people running each site do not work for you.
What the second host should be told
There is a temptation to keep each provider unaware that there is another one, on the theory that it preserves negotiating position. In practice it tends to cost more than it earns.
A host who knows they hold part of a fleet, and that the rest sits elsewhere, has a visible reason to perform. The comparison is not hypothetical to them. Providers who are confident in their operation generally respond well to it, and a provider who reacts badly to the existence of a second site has told you something useful about how a future dispute would go.
What is worth keeping to yourself is the commercial detail. There is no reason to share the other provider’s rate, contract terms or capacity. The useful disclosure is simply that the fleet is split and that deployments are allocated on performance.
That framing also makes growth conversations easier. Expanding at whichever site has earned it is a normal commercial decision, and one that both providers can understand without anyone being surprised.
Measuring the two against each other
The reference point a second site gives you is only worth having if the comparison is fair, and most casual comparisons are not.
Machine populations differ. A site running newer units will post better efficiency numbers that say nothing about the facility. Ambient conditions differ, and a hall in a cooler climate will derate less in July. Pool configuration differs if it was never standardized. Any of those will make one host look better than it is.
The comparisons that survive scrutiny tend to be the ones the facility actually controls:
- Realized uptime against the contracted figure, measured by your own monitoring rather than the host’s reporting
- Repair turnaround from fault to machine back in production, counted in days
- Curtailment hours, and whether they matched what the agreement said to expect
- Invoice accuracy, and how disputes were handled when raised
- Response time on routine requests, which predicts response time on urgent ones
None of those depends on which machines are where. All of them are things a prospective third deployment would want to know.
Staging it
The move that goes wrong is the one made at speed, during a dispute, with no prior relationship anywhere else. The move that goes well is boring.
Start with a deployment small enough that losing it entirely would not matter much. Run it for a full billing cycle and a full seasonal swing before concluding anything. Compare what the two hosts report against what your own monitoring says, which is far easier when you have a second data set to compare against. Then decide whether to grow it.
Owners who do this usually find the second relationship earns its keep long before any crisis, simply by giving them a reference point. It is very hard to tell whether a host’s curtailment, derating or repair turnaround is reasonable when you have nothing to compare it to.
The honest summary
Running one host is the right answer for a lot of fleets, and there is no virtue in complexity for its own sake. But it should be a decision, taken knowing what it costs, rather than the default that nobody ever examined.
If the fleet matters to you financially, the question is not whether your host is good. It is what you would do on the day a good host stops being the right place for your machines, and whether you would rather start working that out then or now.
If you are weighing a second site or a first one, our team is happy to talk through what a staged deployment would look like for your fleet size and where it would sit. Get in touch and we will give you a straight answer, including when the numbers say to stay where you are.
Explore Rax Mining
- Bitcoin Miner Hosting — Tiered rates from $0.085/kWh, NatGas from $0.055/kWh
- NatGas MDU Units — 1MW modular datacenter containers
- Mining Profitability Calculator — Estimate your mining returns
- Our Facility — Tour our mining infrastructure

