Colocation
Your own hardware in our European datacenters, on the same network as our cloud and bare metal. A third of a rack, two thirds, or all of it. Power and cooling are inside the availability we commit to. Cross-connect to any carrier in the building, or take our transit.
Reply within 24 h
No commitment
An engineer reviews it
The most demanding brands in Europe run on us.
Two ways out of one rack
A company with offices, hardware it owns and workloads in public clouds. The rack sits inside the network we operate, beside its instances and its bare metal. Traffic leaves it two ways and only two: through our edge, whichever path it takes after that, or straight out on a cross-connect that never touches us.
Topology: your hardware sits inside the DIVERSITY network, beside your cloud instances and your bare metal servers, all three on one internal network with no public hop between them. Your offices, branches and own datacenter reach it over Private Link, public internet traffic reaches it over our own IP transit, and we peer directly with the public clouds at European exchange points. One run leaves your hardware without passing through our network edge at all: a cross-connect to any carrier in the building, on your own contract with them.
Public internet
Your end users
Head office
Branches and sites
Your own datacenter
Public clouds
AWS, Azure, GCP, Oracle and Cloudflare
Any carrier in the building
Your own cross-connect, on your own contract
DIVERSITY network
Network edge
Madrid, Amsterdam, Munich and Helsinki
Your hardware
A third of a rack, two thirds, or a full rack
Cloud instances
VPS, VDS, MEM and MNG
Bare Metal
Single tenant, same network
Our IP transit
Aggregated, not resold from one carrier
Private Link
Direct peering
At European exchange points
Cross-connect
One network inside
No public hop between them
Private Link, three ways
- Our router over the connectivity you already have
- Our fibre and our router, both
- Your own tunnel, over WireGuard, IPsec, or GRE inside IPsec
Our transit
Aggregated rather than resold from one carrier, direct on the major European internet exchanges. This is what most racks take.
Your cross-connect
A circuit from your rack to any carrier in the building, on your contract with them. It does not pass through our network, and we do not sit in it.
Private Link
Head office, branches and your own datacenter. Our router over your connectivity, our fibre and router, or your own tunnel.
Direct peering
AWS, Azure, GCP, Oracle and Cloudflare, at European exchange points.
Inside the network
Your hardware, your cloud instances and your bare metal servers on one network, with no public hop between them.
What you are buying
Three sizes, and they are thirds of a rack, not the quarters and halves you will have been quoted elsewhere. What separates them is not only space: a full rack is badged around the clock to your own people, and a shared one is escorted, because a rack with other tenants in it cannot hand out credentials without exposing them.
Rack space
1/3 rack
A third of a rack, shared with other customers
Escorted access to the rack
2/3 rack
Two thirds of a rack, still shared
Escorted access to the rack
Full rack
A whole rack, with nobody else in it
Badged access, 24x7, held by your people
The rack itself
Up to 20 kW, where the site supports it
Independent feeds, dual-corded kit
Contained aisles, front to back
27°C ±5°C in the cold aisle
Power and cooling
Power
- Independent no-break feeds to every rack, backed by UPS and generators
- At some sites that is three separate paths instead of the usual two
- Your equipment has to be dual-corded and draw on both the A and the B feed, or it is outside what we can commit to
- Real draw has to stay under 80% of the rack's power connection
- Up to 20 kW per rack, where the site supports it
Cooling
- Cold aisle containment, with doors and aisle ceilings
- Airflow is front to back, so hardware has to be built for it
- We hold 27°C ±5°C, measured 50cm from the rack at a height of one metre, inside the cold aisle
- We hold relative humidity between 35% and 65%, to keep static under control
- A blanking plate in every empty U, which we require rather than recommend
Fire and containment
- We sample the air continuously for smoke
- Gaseous suppression, so a fire goes out without soaking the racks around it
- Rooms are compartmented, and the walls carry a rated fire resistance
Connectivity
Our transit
- Aggregated across the transit providers available to us, not resold from one carrier
- Direct on the major European internet exchanges
- Peering with AWS, Azure, GCP, Oracle and Cloudflare
- What most racks take, and never compulsory
Cross-connects
- To any carrier present in the building
- On your own contract with them, not through us
- It does not pass through our network, and we do not sit in it
Private Link
- Our router over the connectivity you already have
- Our fibre and our router, both
- A tunnel over your own connection, with WireGuard, IPsec, or GRE inside IPsec
Access and remote hands
Getting in
- Full rack: 24x7, badged, and the credentials are your own people's
- A third or two thirds of a rack: escorted, with no badge issued
- The difference is the other tenants in a shared rack, not a service level
Remote hands
- Our own engineers, doing the thing you would have driven there to do
- A prepaid bag of hours on the advanced support plan
- Billed by the hour on the free support baseline
Before you ship
- Tell us the kit and what it draws, and we will tell you which site fits
- It has to be dual-corded and move air front to back
- Blanking plates in every empty U, without exception
Cover and terms
Service level
- 99.9% availability, by contract, with automatic credits if we miss it
- Here it covers the service, the power and the cooling
- It stops at what you deploy inside the rack, which is yours
Support
- Free on every product we sell, with a 4-hour target response on support requests
- Weekends and out of hours need the advanced support plan, which is paid
- An engineer, not a tiered ticket queue
Contract
- Twelve months minimum
- Renewed tacitly for twelve more unless you give notice
- Three months before the end of the period, not three months from whenever you ask
Where you can buy it
MAD1 Madrid
- 0.5 MW installed power capacity
- 2 carrier uplinks
- Rack space available at short notice
AMS1 Amsterdam
- 2 MW installed power capacity
- 6 carrier uplinks
- Rack space available at short notice
MUN1 Munich
- 2 MW installed power capacity
- 8 carrier uplinks
- Sold here, with a lead time
HEL1 Helsinki
- 0.5 MW installed power capacity
- 2 carrier uplinks
- Sold here, with a lead time
Latency and uptime, in figures
Two figures instead of adjectives: the round trip your own browser measures right now, and the availability we commit to by contract. On this product that second figure covers more than the service. It covers the power and the cooling too, and stops at what you deploy inside the rack.
--ms
Your latency to our network
Measured between your browser and our servers as you read this, not taken from a brochure.
How we measure it (opens in a new tab)
99.9%
Availability, by contract
Service, power and cooling. Automatic credits if we miss it.
Anyone can rent you a rack in Europe. Not everyone is European.
Your hardware sits in a European datacenter under European law, and DIVERSITY is a European company with no parent company, US or otherwise, so there is no entity above us for an order from outside the EU to be served on. The other half of not being captive is the network. None of our datacenters is owned by a carrier, so you never buy transit from whoever owns the building, and are never tied to them. Cross-connect to any carrier in the building on your own contract, or take our transit, which is aggregated rather than resold from a single carrier.
Data in Europe
Workloads and critical data on European infrastructure. No transit through foreign jurisdictions.
GDPR: data protection by design
Built for traceability, compliance, and control. Not a layer added afterward.
European datacenters
Madrid, Munich, Helsinki, Amsterdam. Operated by our own team.
No US parent
No exposure to extraterritorial reach like the CLOUD Act. Jurisdiction is the shield.
When to use Colocation
When the hardware has to be yours and what you want from us is the rest: the power, the cooling, the network it sits on and the contract around it.
Moving your own servers out of a comms room and into a contained aisle
A third of a rack now, two thirds or a whole one when you outgrow it
A second site for disaster recovery, in a different European country
An appliance, a licence dongle or an HSM you cannot virtualise
Dense kit that wants up to 20 kW in one rack, where the site supports it
Racks beside your instances and your bare metal, with no public hop between them
A cross-connect to the carrier you already have a contract with
Offices and branches that need a private path in, not a VPN concentrator
Use cases by industry
Where this product earns its place. Sectors where it replaces something more expensive, more painful, or simply riskier.
Regulated and public sector
Where an auditor wants to know which building the equipment is in and who can walk up to it. Full racks are badged to your own people, shared racks are escorted, and the company you signed with is European, with no US parent.
Banking and finance
A second site in a different European country, with our own Layer 2 network between the two rather than a path across the open internet. DORA asks who your provider answers to: one company, one jurisdiction, no US parent.
Healthcare and pharma
Systems that cannot be virtualised and data that cannot leave your control, on hardware you still own. Power and cooling are inside the availability we commit to by contract, which is the part a hospital basement cannot promise.
Manufacturing and industry
Licence dongles and appliances that have to be a physical box somewhere. They can sit in the same building as the rest of your estate instead of in a cabinet next to the production line.
Software and SaaS
The half of the estate that is still yours, on the same network as your instances and your bare metal. One contract, one operations team, and no hop across the public internet between the two halves.
Media and rendering
Dense hardware that wants up to 20 kW in one rack, where the site supports it, with front-to-back airflow inside contained aisles. Take our aggregated transit, or cross-connect to whoever already moves your traffic.
Want to know what fits first?
Tell us what the hardware is, how much it draws and where it has to be.
You get a rack size, a site, a lead time and a price from an engineer, normally in under 24 hours.
Reply within 24 h
No commitment
An engineer on the call
Four datacenters. One standard.
Independently audited and carrier-neutral, with our infrastructure in them managed by our own engineers. Where one site does more than another, the difference is written into your technical dossier rather than averaged into a claim about all four.
Our datacenters are independently audited, not self-assessed. Across the estate that means ISO 27001 for information security, ISO 9001 for process, ISO 14001 for environmental management, and SOC 2 Type 2 where we publish it; our Amsterdam capacity additionally sits in a facility audited against the internet exchange's own colocation standard. Scope and audit dates differ by site and both come with the technical dossier, because a certificate without its scope statement is decoration.
Mitigation at the network edge, before traffic reaches your systems. It engages on its own and does not wait for you to open a ticket, which matters because the first ninety seconds are the ones that decide whether an attack is an incident or a graph. Per-customer thresholds are agreed when you sign up rather than discovered during the first event.
None of our datacenters is owned by a carrier, so you never buy transit from whoever owns the building, and are never tied to them. We sit directly on the major European internet exchanges, and we peer with AWS, Azure, GCP, Oracle and Cloudflare, so hybrid traffic takes a short, known path instead of whatever a transit provider chooses that day.
High-sensitivity smoke detection that samples the air continuously rather than waiting for a ceiling sensor to trip, and gaseous suppression that puts a fire out without soaking the racks around it. Rooms are compartmented so an incident stays in the room it started in, and the walls carry a rated fire resistance rather than an assurance.
If part of your infrastructure stays on your own hardware, it can live in the same building as your cloud, up to 20 kW per rack at sites that offer it. Same network, same operations team, no hop across the public internet between the two halves of your estate, and one contract rather than two vendors blaming each other.
Every rack takes independent no-break feeds, and where the facility provides it that is three separate paths rather than the usual two, each behind its own generator. UPS and generators are start-tested on a schedule, not assumed, because the failure mode everyone actually meets is a generator that had not been run since commissioning.
Redundant cooling per room, sized to hold temperature if a unit drops out, with temperature and humidity measured continuously rather than on walk-rounds. Our Amsterdam site runs at a PUE of 1.3 on entirely renewable electricity and pipes half its waste heat into neighbouring homes and university buildings instead of venting it.
The same standard at every site.
Pick a site for latency or for jurisdiction, not for what you get. Every item listed is the floor at all four. Where one facility does more than another, the difference is written into your technical dossier rather than averaged into a number.
Key data
The numbers that matter when uptime is the contract.
sustained peak across the European network
live, per datacenter, on our status page
European infrastructure since 2015
Madrid, Munich, Helsinki, Amsterdam
Things people ask before signing
Straight answers, no upsell.
What are we buying from you that a datacenter operator would not give us?
Floor space is the easy part to compare, so start with the rest. The rack is plugged into a network we run, on transit we aggregate rather than resell from one carrier, and it sits beside your cloud instances and your bare metal servers with no public hop between them. Power and cooling are inside the availability we commit to by contract, so they are something we owe you. Remote hands is our own engineers. And there is one contract and one operations team for the whole estate, instead of a landlord, a carrier and a cloud provider each explaining that it is one of the others.
What sizes can I take?
A third of a rack, two thirds of a rack, or a full rack. They are thirds, not the quarters and halves most of the market quotes, so compare on what you get rather than on the fraction.
How much power can one rack draw?
Up to 20 kW, where the site supports it. The figure is a property of the facility your rack is in, not of the product, so ask us for the site you are buying in before you order dense hardware. Two conditions hold everywhere. Your equipment has to be dual-corded and draw on both feeds, because a single-corded server sits outside what we can commit to on power. And its real draw has to stay under 80% of the rack's power connection.
What is the temperature and airflow envelope?
We hold 27°C with a permissible deviation of 5 degrees either way, measured 50cm from the rack, at a height of one metre, inside the cold aisle. Where you measure is part of the number: put a probe in the hot aisle and you will read something else, correctly. Racks sit in contained rows with doors and aisle ceilings, airflow runs front to back, and every empty U carries a blanking plate, which we require rather than recommend. We hold relative humidity between 35% and 65%, mostly to keep static under control.
Can I cross-connect to my own carrier?
Yes, to any carrier present in the building, on your own contract with them. That circuit does not pass through our network and we do not sit in it. Most racks take our transit instead. We aggregate the transit providers available to us rather than reselling one carrier, we sit directly on the major European internet exchanges, and we peer with AWS, Azure, GCP, Oracle and Cloudflare, so the route to a given destination is usually the aggregate rather than one provider's view of it. Either way it is your choice, and you can run both.
Which sites can I buy in?
All four: Madrid, Amsterdam, Munich and Helsinki, and a rack in any of them sits on the same network as the other three. Space at short notice is a narrower answer, and today that is Madrid and Amsterdam.
Who can get into the rack, and when?
If you have a full rack, your own people hold the credentials and the access is 24x7. If you have a third or two thirds of a rack, access is escorted and no badge is issued, because the rest of that rack belongs to somebody else and handing out unescorted credentials would expose them.
Can your engineers work on my hardware for me?
Yes. Remote hands is our own engineers doing the thing you would otherwise drive there to do: a reseat, a swap, a cable, eyes on a front panel. On the advanced support plan it is a prepaid bag of hours that is already set up, so there is nothing to negotiate at two in the morning. On the free support baseline it is billed by the hour. We do not publish rates on the site, so ask and you will get the number.
What does the 99.9% cover here?
99.9% availability, by contract, with automatic credits if we miss it. On this product it covers the service, the power and the cooling, so keeping your rack fed and in its thermal envelope is a commitment we owe you. Where it stops is at what you deploy inside the rack. Your hardware, your operating systems and your applications are yours, and the power side of the commitment assumes your equipment is drawing on both feeds.
What happens if a power feed fails?
Nothing you should notice, provided your equipment is drawing on both. Every rack takes independent no-break feeds backed by UPS and generators, and at some sites that is three separate paths instead of the usual two. A single-corded server defeats all of it on its own, which is why we ask about cording before you ship.
Are the datacenters certified, and what does that cover?
Our datacenters are independently audited, not self-assessed, and the certificates belong to the buildings rather than to us. On this page that distinction matters, because the building is what you are buying. Scope and audit dates differ by site, and so does the physical envelope: the feed topology, the containment and the thermal numbers are all written into the technical dossier for the site you are in. Ask for that dossier before you assume a scheme or a figure covers the room your rack will be in. A certificate without its scope statement is decoration, and the scope is the part an auditor reads.
How soon can I have space?
In Madrid and Amsterdam, at short notice. In Munich and Helsinki the answer is a lead time we give on request rather than one we would quote on a web page, because it depends on how much space you need. Ask and you get a real one, before you plan a migration around a site rather than after.
How long is the contract, and how does it renew?
Twelve months minimum, and it renews tacitly for twelve more unless you give notice. Notice is three months before the end of the period, which is not the same as three months from whenever you ask: miss the window and you are in for another year. That is standard in this industry, and you should have it now rather than in month thirteen. Put the date in a calendar when you sign.
What does it cost?
We do not publish rates, for this or for anything else we sell. What moves the number is the size, the power you actually draw, whether you take our transit or bring your own carrier, and how much remote hands you expect to use. Tell us those four things and an engineer comes back with a price, normally in under 24 hours, along with what it does and does not include. No sales sequence attached to it.
Who can be ordered to let someone near my hardware?
We can, under European law, on a European order. Your own access to the rack is arranged through us, and on a shared rack there is no unescorted route to it at all.
Why does it matter where my provider is headquartered?
The US CLOUD Act lets American authorities compel any provider under US jurisdiction to hand over data in its possession, custody or control, wherever it is held, without notifying you. Colocation looks like it escapes that, because the disks are yours and they are in Europe. What does not escape it is any company in the chain under US jurisdiction, or a subsidiary of one, that can reach the room, the network, or the records of who comes in and what is installed. Neither DIVERSITY nor the operator of any of our buildings is owned or controlled by a US company, so there is no entity above us for such an order to be served on.
Can you see what is running on our servers?
No. We do not hold credentials to your hardware and there is no console we log into on your behalf, which is the structural difference between colocation and a managed service. What we control is the space, the power, the cooling, the network your rack is plugged into and who walks into the room. If you want us able to do more than that, you say so and we agree it in writing first.
What they say about our work
“Half our estate is still on our own hardware. Putting those racks in the same building as the cloud removed a whole class of problem we had stopped trying to solve.”
Infrastructure manager
Industrial group, Portugal
1 of 3
What else we operate
Private Link for a private path in, and Cloud instances and Bare Metal for the half of the estate you would rather not own. Same network, same operations team.
Shall we talk?
An AI deployment, critical infrastructure, or a quote.
Write or call and you'll get an engineer, not a ticket bot.


