Your hardware, our building, our hands.

Ship us machines you already own and we rack, power, and connect them in one of two Des Moines facilities. When something needs a hand on it, that hand belongs to somebody on our team rather than a contractor who has never seen your cabinet.

carpathian.ai / colocation / facilities
OccupiedAvailablePowered and reachable
US Central 001Production floor, carrying live colocated and cloud workloads
US Central 002Disaster recovery site, same hardware and same network as 001
PowerRedundant
CoolingRedundant
Carrier uplinksTwo independent
Core switchingActive and standby
Rack monitoringAround the clock
Physical accessOur own staff
Two Des Moines sites
US Central 001 carries production, US Central 002 is the disaster recovery site
Redundant power and cooling
At both buildings, with racks monitored around the clock
Two carrier uplinks
Independent paths out of the building, either one able to carry the site alone
Remote hands from our staff
The people who touch your cabinet work in the building it sits in

What colocation gets you

The machines stay yours. The building, the power, the cooling, and the network become ours to worry about.

You buy the hardware, specify it the way your workload wants it, and keep root on every box. We give it space in a facility Carpathian owns and operates, feed it power with more than one source behind it, keep the air around it cold, and hand it two ways out of the building.

What that takes off your plate is the part of infrastructure that has nothing to do with your software. Standby power, redundant cooling, a second uplink, and the trip to the building when a drive fails are all on our side of the line. Racks are monitored around the clock, and the people who open the cabinet are Carpathian staff rather than a contractor sent by a landlord.

If you also run cloud instances with us, they are on the same network in the same buildings, so the two halves of a setup are not separated by the public internet.

  • Space, power, and bandwidth in a facility Carpathian owns and operates
  • Redundant power and cooling at both sites
  • Racks monitored around the clock
  • Two independent carrier uplinks, either one able to carry the site alone
  • Paired core switching, with an active core and a synchronised standby
  • Remote hands from staff who work in the building
  • The same network your cloud instances run on, if you use both

Colocation, cloud, or both

What separates these is who owns the hardware and how quickly you can change it. Colocation and cloud run in the same buildings on the same network today, so choosing one does not rule out the other.

Colocation

You own the machines. You picked the processors, the disks, and the chassis, and you can keep running them for as long as they earn their place in the rack. Changing the fleet means buying hardware and shipping it, so the unit of change is a purchase order and a delivery date rather than a click.

Your hardwareOur space, power, and bandwidth

Cloud

You rent a virtual machine on hardware we own and share between tenants. Nothing arrives on a pallet. You size an instance, boot it, resize it when the workload changes, and destroy it when the work is finished, all in minutes.

Our hardwareResize or destroy in minutes

Both together

A colocated database and a cloud application server sit on the same network in the same buildings, which is why teams often keep the machines they already bought and grow into cloud instances beside them instead of replacing a fleet in one go.

Same buildingsSame network

Bare metal

A whole physical machine from Carpathian with no other tenant on it. It is still in development and is not yet available to provision, so nothing on this page includes it. If it is what you want, tell us and we will let you know when it ships.

In developmentNot yet available to provision

US Central 001 and US Central 002

US Central 001 carries production. US Central 002 is the disaster recovery site. They run the same hardware and the same network, so a failover does not change how a workload behaves.

US Central 001

The production floor. Cabinets here carry live workloads, colocated and cloud alike, on redundant power and cooling with racks monitored around the clock. Physical access is performed by Carpathian staff who work in the building.

ProductionRedundant power and cooling

US Central 002

The disaster recovery site, carrying the same hardware and the same network as 001. Production traffic can fail over to it, and because both ends match, a workload that moves does not have to be reconfigured before it will run.

Disaster recoveryAutomated failover
carpathian.ai / colocation / network path
traffic on the active pathtraffic on the standby pathdevice reporting healthy
Carrier uplinksTwo independent uplinks, either one able to carry the site on its own
Paired core switchingAn active core and a synchronised standby, with failover that needs no one awake
Your cabinetYour machines on the same network our cloud instances run on

Why Des Moines

Iowa sits outside hurricane and wildfire zones, which keeps the natural risk profile of both buildings low. The failure modes we plan for here are the ordinary ones: a utility feed, a cooling unit, an uplink. Each of those has a second one behind it.

Des Moines is also near the geographic center of the United States, which keeps round trip times short for customers on either coast without paying for a presence in both. For most workloads that settles the latency question at a single location.

It is a place we can drive to, which is the part that matters when your cabinet needs opening. Physical access and remote hands are performed by Carpathian staff who work in the building, so a reboot, a cable swap, or a media load does not go through somebody who has never seen your equipment.

What two sites give you

Both buildings are in Des Moines, so what you get from the pair is site redundancy. Each site has its own redundant power and cooling, both run the same hardware and the same network, and production traffic can fail over between them. A utility feed or a cooling unit failing in one building does not follow the workload to the other.

It is not geographic redundancy. Two facilities in one metro share a grid region and a weather system, and we would rather write that here than have you discover it during a review. If your recovery plan requires distance from this part of the country, tell us early and we will say plainly whether we can meet it.

  • Independent power and cooling at each site
  • The same hardware and the same network in both buildings
  • Production traffic can fail over between the two
  • One metro, so this is site redundancy and not geographic redundancy

Questions we get asked

Straight answers to what buyers ask before shipping hardware into a building they have not been in yet.

Can I colocate my own hardware?
Yes. You ship us your servers and we rack, power, cool, and connect them. You keep ownership of the hardware and full control of what runs on it, and we handle the facility, the network, and remote hands.
Where would my equipment sit?
In one of two Carpathian facilities in Des Moines, Iowa. US Central 001 runs production workloads and US Central 002 is the disaster recovery site. Iowa sits outside hurricane and wildfire zones, which keeps the natural risk profile of both buildings low, and Des Moines is near the geographic center of the United States, so round trip times stay short for customers on either coast.
Is two sites in one metro geographic redundancy?
No, and we would rather say so here. Both buildings are in Des Moines, so this is site redundancy: each site has its own redundant power and cooling, they run the same hardware and the same network, and production traffic can fail over between them. It is not separation across grids and weather systems. If your recovery plan requires that distance, tell us early and we will say plainly whether we can meet it.
What is the difference between colocation, bare metal, and cloud?
With colocation you own the hardware and rent space, power, and bandwidth. With cloud you rent a virtual machine on shared hardware and can resize or destroy it in minutes. Bare metal, a whole physical machine from us with no other tenant on it, is still in development and is not yet available to provision. Colocation and cloud run in the same buildings on the same network today, and bare metal will join them once it ships.
How is the network built?
Two independent carrier uplinks leave each site, either one able to carry it on its own, into paired core switching with an active core and a synchronised standby that fails over automatically. Colocated cabinets sit on the same network our cloud instances run on, so a colocated database and a cloud application server are not separated by the public internet.
How do I get remote hands on my equipment?
Open a support request from your dashboard describing the work. Physical access and remote hands are performed by Carpathian staff who work in the building rather than by a contractor, and we send back photos or console output when the work is done. Reboots, cable swaps, media loads, and diagnostics are all routine.
What uptime do you commit to?
Carpathian commits to a 99.5% uptime SLA on cloud and dedicated services. Both sites carry redundant power and cooling, racks are monitored around the clock, and production traffic can fail over between the two buildings.
What does a cabinet cost?
We quote against your equipment rather than publishing a rate card, because the space, the power, and the bandwidth all depend on what you are bringing. Send us the specification and the power draw and we will come back with a number. If you would rather see the room before you ship anything into it, a visit can be arranged.

Send us the specification and the power draw.

Tell us what the machines are and what they draw, and we will quote the space, the power, and the bandwidth against that list rather than against a rate card that fits nobody. If you would rather see the room before you ship anything into it, a visit can be arranged.