Get Quotes

Your guide to the right provider

How to choose a data center

We're data center people, not a scraped database. Our team brings 20 to 30 years at some of the largest data center platforms, in the channel and in commercial real estate brokerage focused on data centers. We've seen more than 10,000 requirements and worked with nearly every provider, the good, the bad and the ugly. From a few cabinets running everyday enterprise workloads to a private suite of $700,000 NVIDIA servers, these are the straight answers we give buyers every week.

Start with your situation

Nobody wakes up wanting colocation. You have a problem to solve. Find yours below and we'll tell you what matters, what to ask and where to start.

Our server room is out of power, space or patience.

This is the most common reason buyers call us. Start by measuring what your gear draws today and what it will need in three years, then look for a facility within easy driving distance so your team can still get hands on. A few cabinets in a well-connected retail facility usually beats another round of UPS and cooling upgrades in an office building. Ask about remote hands, so a reboot doesn't mean a drive.

Our data center lease or hosting contract ends soon.

Work backward from the end date. Carrier circuits often take weeks to a few months, and larger spaces need build-out time, so start six to twelve months out. Use the move to right-size power, renegotiate network costs and lock in renewal caps you didn't get last time. Your current provider's renewal offer is a data point, not a deadline.

We need a disaster recovery site.

Pick a site far enough away that one storm, flood or grid failure can't take out both, and close enough that replication fits your recovery point and recovery time objectives. Check that both sites can reach the same carriers and clouds. Ask how the provider handles a regional disaster, including fuel delivery and staff access.

We're deploying AI or GPU servers.

Density and cooling decide everything. GPU racks often draw 30 kW to well over 100 kW, so you need a facility that supports that density today, usually with liquid cooling such as direct-to-chip or rear-door heat exchangers. You also need contiguous cabinets for short copper runs, heavy fiber pathways and power that's energized when the hardware lands. Many buildings simply can't take it.

Our cloud bill keeps climbing.

Move the steady, predictable workloads to owned hardware in colocation and keep the spiky ones in the cloud. Private connections from your cabinet to AWS, Azure, Google Cloud or Oracle can cut data egress costs. Pick a facility with on-ramps to the clouds you use, and model the savings with your real traffic before you commit.

The auditors are coming.

Ask for the provider's current SOC 2 Type II report and any others your framework needs, such as PCI DSS, HIPAA, ISO 27001 or FedRAMP, plus a responsibility matrix showing which controls you inherit. A private cage with its own access control and logs makes audits easier. Confirm the provider will support your auditors when they visit.

We need to be closer to our users or our networks.

Latency is about where your users, partners and clouds are, not where your headquarters is. Carrier hotels and buildings with internet exchanges give you the most networks in one place. For regional users, a secondary market close to them can beat a distant hub on both latency and price.

We're consolidating after an acquisition or growth spurt.

Inventory every site, contract and end date first. Consolidating into fewer, better facilities usually cuts cost and complexity, and a provider with locations in several markets can put everything under one contract and one support model. Line up contract end dates so you aren't paying for two sites longer than necessary.

Why colocation

Colocation means you own the servers and a provider supplies the building: space, power, cooling, security and connectivity.

Why do companies choose colocation?

The most common reasons, according to 451 Research's 2025 survey of colocation customers, are better reliability, disaster recovery, not wanting to run their own data center, direct connections to clouds and carriers, compliance and moving spend from capex to opex. Many also come because their own server room ran out of power or space.

Colocation, cloud or my own data center?

Most companies run a mix. Colocation fits steady workloads, hardware you already own, compliance requirements and anything where the cloud bill keeps climbing. Cloud fits spiky or short-lived workloads. Running your own data center only makes sense at scale, with a team to staff it 24x7.

Can colocation lower my cloud bill?

Often. Moving steady workloads off public cloud cuts compute costs, and a private connection from your cabinet to AWS, Azure, Google Cloud or Oracle can reduce data egress fees compared with sending the same traffic over the internet. Providers publish their own savings figures, so model it with your real traffic before you count on it.

Should we build our own data center instead?

Only if you have the scale and the team. Building means capital for land, shell, switchgear, UPS, generators, cooling and security, then maintenance capital to keep it all running, 24x7 staff, commercial utility rates and annual compliance audits. For most companies, colocation buys the same reliability without tying up capital or hiring a facilities team.

Are companies moving workloads back from the cloud?

Many are rebalancing. Steady, predictable workloads often cost less on owned hardware in colocation, while the cloud keeps the spiky and experimental work. Colocation becomes the anchor of a hybrid setup, with private connections to each cloud. It's rarely all or nothing.

What do I still manage in colocation?

Your servers, storage, network gear and everything running on them. The provider runs the facility: power, cooling, fire suppression, physical security and the building network. Remote hands can handle basic physical tasks for you so you don't have to drive in.

Choosing a location

Location drives latency, cost, risk and how easily your team can get to your gear.

How do I choose a market?

Start with who needs to reach the servers and how fast: users, offices, partners and clouds. Then weigh drive time for your team, power cost and availability, tax treatment, network density and natural hazard risk. Primary markets offer the most carriers and providers. Secondary markets often cost less and have more room.

How far should a disaster recovery site be from my primary site?

Far enough that one event can't take out both: a different flood zone, power grid, fault line and storm track. Close enough that replication latency fits your recovery point objective. Many companies pick a second metro a few hundred miles away. Tell us your RPO and RTO and we'll suggest pairs that work.

Does the state matter?

Yes. Power prices, sales tax exemptions on data center equipment, property tax and data residency rules vary by state. Some states also add rules for very large power users. For a few cabinets the difference is small. For hundreds of kW it can decide the deal.

What about secondary and edge markets?

Secondary markets often offer lower prices, more available power and lower disaster risk than the big coastal hubs, and they put your gear closer to growing user bases. The tradeoff is fewer carriers and providers. A provider with sites in several markets can keep contracts and operations consistent as you spread out.

Flood zones, weather and climate

Every market has hazards. What matters is whether the building was designed for the ones it faces.

Should I avoid flood zones?

Avoid FEMA's 100-year flood zone, which carries a 1% chance of flooding in any year, and look hard at the 500-year zone (0.2%). Ask where the generators, fuel tanks, switchgear and utility feeds sit relative to flood elevation. Equipment on the roof or raised pads matters as much as the building's address.

What weather and natural risks should I check?

Hurricanes and storm surge on the Gulf and Atlantic coasts, earthquakes on the West Coast, tornadoes and hail in the Plains and Midwest, wildfire and smoke in the West, ice storms and extreme heat anywhere. Ask what the building is rated for, such as wind load and seismic design, and how long it can run on generators without a fuel delivery.

How does climate affect a data center?

Cool, dry climates allow free cooling with outside air or economizers for more of the year, which lowers power use and PUE. Hot, humid climates rely more on mechanical cooling and water. Extreme heat tests cooling plants and the grid at the same time, so ask how the facility performed in its last heat wave.

What other site risks should I look for?

Flight paths, rail lines and highways that carry hazardous cargo, nearby chemical plants and single points of failure in the utility or fiber feeds. A strong site has two utility feeds, diverse fiber entrances and fuel suppliers on contract for priority delivery.

Choosing a facility

Two buildings in the same market can differ more than two markets do.

What makes one facility better than another?

Redundancy, density, connectivity and operations. Look for concurrently maintainable power and cooling (N+1 or better), enough density and cooling for your gear, the carriers and clouds you need on site, strong security and a team with a clean operating record. Then check that the space and power you need are available when you need them.

What do Tier III and Tier IV mean?

They're Uptime Institute design standards. Tier III is concurrently maintainable: any component can be taken offline for maintenance without dropping your load. Tier IV is fault tolerant: a single failure anywhere won't take you down. Many good facilities are built to Tier III standards without paying for formal certification, so ask how the power and cooling paths are designed.

What do N, N+1 and 2N mean?

N is exactly enough equipment to carry the load. N+1 adds one spare unit. 2N is two complete, independent systems, each able to carry everything. More redundancy means fewer ways to go down and usually a higher price.

How do I check a facility's track record?

Ask about power or cooling outages in the last five years and how they were handled. Ask how often generators and UPS are load tested, how changes to electrical and mechanical systems are approved and whether you can see recent incident reports. No operator is perfect. What matters is how they document, communicate and fix things.

Is the facility ready for AI and high-density gear?

Ask for the maximum kW per cabinet it supports today, not someday, and how it's cooled. Most retail colocation was built for 5 to 10 kW per cabinet. GPU racks run 30 kW to well over 100 kW and usually need liquid cooling such as direct-to-chip or rear-door heat exchangers, plus room for contiguous cabinets and heavy fiber.

Choosing a provider

A data center is more than four walls. There are a lot of good operators, and they differ in design, track record, support model and price.

Aren't all data centers basically the same?

No. Two facilities with similar specs can perform very differently. Design choices, maintenance discipline, staffing levels, how fast problems get fixed and how the provider treats customers at renewal all vary. The right fit depends on what you value most: lowest price, white-glove support, density, a national footprint or a local team that knows your name.

Does company size matter?

Large national providers bring multiple sites, standard processes, deep interconnection and room to grow, but can be less flexible on terms. Regional and independent operators often offer more flexibility, faster answers and local staff, with fewer locations. Neither is better across the board. Match the provider to how you buy and how you grow.

Public, REIT or private equity: what does ownership mean for me?

Public companies and REITs publish their financials and tend to have scale and standard pricing. REITs pay out most of their income as dividends, so they lean on capital markets to fund growth. Most private equity-backed providers are built to be sold. The playbook: buy or build sites, fill them with long-term customers with good credit, then sell to another fund or a larger operator, which often repeats the cycle. That doesn't make them bad partners, but your provider may change hands during your term. Founder and family-owned operators tend to hold for the long run.

Why don't a provider's website and spec sheet match?

Because keeping specs current is hard, even for the provider. The engineers who run the building and the marketing team who writes the website don't always talk. Inventory and specs usually live in spreadsheets, with facility managers and on desktops, in more than one version. Some providers refresh their sheets every year, some every six months and some you can't tell. We've read every operator website and spec sheet we could get, and contradictions in megawatts, square footage, carrier counts and even open dates are common. Treat any published number as a starting point. Get current figures in writing, from the people who run the building, before you sign.

What happens if my provider is sold or goes out of business?

Acquisitions are common, and contracts usually transfer to the new owner. Pricing, support and renewal terms can change afterward. Data centers rarely go dark in a bankruptcy because the buildings are valuable. When Cyxtera filed for bankruptcy in 2023, its data centers kept running and were sold to Brookfield, which later built Csquare around them. Still, check the assignment and change-of-control language in your contract.

What really changes when a data center is sold?

More than the logo. New owners usually bring new leadership, and the sales and support people you've worked with often leave. Pricing tends to rise, especially at renewal, when contracts can reset to market rates (mark to market). In the worst case, a new owner who wants one big tenant for the building can try to move or push out retail colocation customers when contracts allow it. That happens. Protect yourself up front: cap renewal increases, limit the provider's right to relocate you, require long notice and paid moving costs if they do, and make sure your terms survive a sale.

Why do new data centers need an anchor tenant?

Because of how they're financed. Private equity funds buy the business or the land, but new builds and retrofits usually run on bank construction loans, and banks want a creditworthy anchor tenant, typically investment grade or backed by one, before they lend. That's why big new campuses are often pre-leased to one hyperscaler before they open, and why space for smaller buyers can be scarce in a new building. It also tells you who the landlord will prioritize.

Why do some data center lists name companies that no longer exist?

Many lists are copied from older lists, and this industry changes hands constantly. Cyxtera and Evoque merged into Centersquare, now Csquare, in 2024. Sungard AS went bankrupt in 2022 and sold its U.S. colocation sites to 365 Data Centers and its recovery business to 11:11 Systems. Data Foundry is now Switch in Austin and Houston. Verizon sold its data centers to Equinix in 2017, Peak 10 and ViaWest became Flexential and Zayo's zColo went to DataBank. If a list still recommends a dead brand, it hasn't been touched in years. We track who owns each building today, so you negotiate with the company that holds the keys.

How do I check a provider's financial health?

Look at ownership history, debt load, credit ratings and public filings where they exist, and how many times the company has changed hands. Ask how expansion is funded. We track current ownership on every provider page, because it changes more often than most buyers expect.

What does ColocationQuotes know that other data center sites don't?

Verified information from the source. Most data center sites copy each other. We researched every operator website and every spec sheet we could get and found plenty of issues. Where we couldn't verify a number, we left it blank and noted to call the provider. We also flag resellers and name the building they're really in, track who owns each provider and when it last sold and show which operators share a carrier hotel. The one outside source we use is PeeringDB for network counts, which is self-reported and labeled that way. Every verified listing shows where we checked it and when.

Is my provider the building owner or a reseller?

Ask, because it changes who you are really depending on. Plenty of providers sell space inside someone else's data center: they lease a suite or a cage from the building operator and resell it in smaller packages, often bundled with servers, cloud or support. That can be a good deal for a few racks. The catch is that the reseller doesn't control the power, cooling, security or carrier access, and its SLA sits on top of the landlord's. If the reseller's lease ends or it gets bought, your gear is in a building you never signed with. Tell-tale signs are a suite number in the address, vague building specs and an address you recognize from a bigger operator's list. Before you sign, ask who owns and operates the building, whose SLA covers power and cooling, how cross connects are ordered and billed, and what happens to your contract if the reseller leaves the building. On our listings, when a provider sells space inside another operator's building, we name the host.

Retail or wholesale provider?

Retail providers sell cabinets, cages and small suites with more services included. Wholesale providers sell large blocks of power and space, usually hundreds of kW and up, with fewer services and lower unit prices. Some do both.

Space: from 1U to a campus

Size the space to your power, not the other way around.

Why is it so hard to find out who has space?

Published availability goes stale fast, and even provider teams don't always have a current view of their own floors. In tight markets, what's left is often scattered pockets: a rack here, a cage there, power in one hall and space in another. That works for some requirements and not others, especially if you need contiguous cabinets or room to grow. Buyers can lose weeks going back and forth only to learn they don't fit. We check fit first, so you only spend time on the options that work.

What is per U colocation?

You rent rack units in a shared cabinet. One U is 1.75 inches of vertical space. It suits a single server or small appliance, often with power and bandwidth bundled. You share the cabinet and its access with other customers.

Half rack or full rack?

A full rack, or cabinet, is usually 42U, with taller 45U and 48U options. A half rack is about 20U with its own lock. Choose by power as much as by space: a half rack with light power can cost less than a full rack you only fill halfway.

What is a cage?

A locked enclosure around a group of cabinets inside a shared data hall. Cages give you a private perimeter for compliance, room to add cabinets and control over who walks up to your gear.

When does a private suite make sense?

When you need your own room, with dedicated access control and often dedicated power and cooling. Suites usually start in the range of a few hundred kW and suit large enterprises, regulated workloads and service providers.

What is build-to-suit?

A data center built or fitted out for one tenant to its specifications, on a long-term lease, often 10 years or more. It's used for very large or specialized requirements such as AI training clusters.

What is a sale-leaseback?

You sell a data center you own to a provider or investor and lease it back. It frees up capital and moves the building off your books while your operations stay put. It works best when the facility is in good shape and you plan to stay for years.

What is a data center campus?

Several buildings on one site that share land, substations and fiber. Campuses give you room to grow next door, diverse buildings for resilience and often lower unit costs at scale.

How much power you need

Power drives most of your monthly bill, so sizing it right is the biggest savings lever.

How do I figure out how much power I need?

Use measured draw from your current gear over several weeks, not nameplate ratings, which run high. Add headroom for growth. Most retail cabinets run 5 to 10 kW. AI and GPU racks can run 30 kW to well over 100 kW.

How much power does a circuit give me?

Watts equal amps times volts, times 1.73 for three phase. Under the National Electrical Code, circuits are loaded to 80% for continuous use. A 30A 208V single-phase circuit supports about 5 kW and a 30A 208V three-phase circuit about 8.6 kW. Common options run from 20A at 120V for light gear to 50A and 60A three phase at 208V, and 415V for high-density and AI racks. Match the NEMA plug type, such as L6-30 or L21-30, to your PDUs.

What are A and B circuits?

Two circuits from separate power paths so your gear stays up if one fails. Each must be able to carry the full load alone, so an A and B pair still gives you one circuit's worth of usable power, not two.

Internet and cross connects

Connectivity is often the reason to pick one building over another.

What internet options do I have?

Blended IP from the provider, which mixes several carriers behind one connection. Dedicated internet access from a single carrier. Or your own multi-carrier setup with BGP, your own IP space and an autonomous system number for full control and failover. Carrier-neutral facilities let you choose from many carriers in the building.

What is a cross connect?

A physical cable between your cabinet and another party in the same building: a carrier, cloud on-ramp, internet exchange, partner or your own gear in another cage. It's usually billed monthly plus a one-time install fee. Virtual cross connects run over a software-defined fabric to clouds and other sites.

Single-mode fiber, multimode fiber or copper?

Single-mode fiber (SMF) carries signals for kilometers and is the standard for carrier and cloud handoffs. Multimode fiber (MMF) is cheaper for short runs of a few hundred meters, such as between nearby cabinets. Copper (Cat6 and Cat6A) reaches up to 100 meters at 1G or 10G and suits management and short in-cage links. Match the type to the far end's port.

What are meet-me rooms and internet exchanges?

A meet-me room is where carriers and customers interconnect inside a building. An internet exchange lets networks trade traffic directly, which cuts latency and transit costs. Carrier hotels are buildings with dense meet-me rooms, where most networks in a market meet.

Carrier hotels: where the internet meets

A carrier hotel is the building in a city where the networks meet. Everyone wants to be in one, own one or connect to one, and for good reason.

What is a carrier hotel?

A building, usually downtown, where dozens or hundreds of carriers, internet exchanges, cloud on-ramps and content networks all have equipment and connect to each other. Many started life as office or telephone buildings and were retrofitted after telecom deregulation in the 1990s. Examples include 60 Hudson and 111 8th Avenue in New York, 350 East Cermak in Chicago, One Wilshire in Los Angeles, the Infomart in Dallas, the Westin Building in Seattle, 1 Summer Street in Boston and 910 15th Street in Denver.

Why does everyone want to be in one?

Network density. When every carrier you might buy from is in the same building, you can switch providers, add a second carrier or reach a cloud with a cross connect instead of a new fiber build. Each network that moves in makes the building more valuable to the next one, and you can't replicate that ecosystem by putting up a new building. That's why carriers, cloud providers and investors fight to be in, own or connect to these sites.

Do I need my servers inside one?

Usually not all of them. A common setup is to keep your network gear and a small footprint in the carrier hotel and run your main compute in a newer, cheaper, denser facility nearby, connected by dark fiber or a wavelength. You get the carrier choice without paying carrier hotel rates for every rack.

What are the tradeoffs?

These are often old buildings. Expect low ceilings, floors not built for heavy GPU racks, limited power per cabinet, older cooling and higher efficiency losses. Space and power are scarce and go fast, so pricing is firm. Some operators have left legacy carrier hotels when the buildings could not meet modern density needs. Buy here for the network, not for density.

One building, many operators: who do I buy from?

In many carrier hotels the building owner runs the meet-me room, and several data center providers lease floors inside. You can often buy similar space from more than one operator in the same building, which is a strong negotiating position. Ask who controls the meet-me room, how cross connects are ordered and billed, and whether you can run fiber directly between floors instead of paying meet-me room fees.

Isn't it risky to put everything in one building?

Concentration is the point and the risk. A fire, flood or cut near one building affects many networks at once. Don't assume two carriers means two separate paths: ask each one for route details, including where the fiber enters the building and whether routes share conduit near the site. Pair a carrier hotel with a second location on different fiber routes for anything critical.

Pricing models and what it costs

Two quotes with the same monthly total can cost very different amounts over the term.

How is colocation priced?

Retail is often a standard space and power rate, for example a full rack with A and B redundant 30A 208V circuits, about 5 kW usable, might run around $1,400 a month. That's an example, not a quote. Real rates vary by market, provider and term. All-in per kW bundles space and power into one rate per committed kW. Metered power charges rent per committed kW plus the electricity you use at the utility rate times the facility's PUE. MG&E, also called kW+E, charges base rent per kW and passes electricity through separately. Per U pricing covers single servers in a shared rack.

What about annual increases?

Most contracts include an annual escalator, either a fixed rate, often around 3%, or a CPI-linked increase. Push for a fixed rate or a cap on CPI, and cap the increase at renewal too.

What if we need out early?

Early termination usually costs the remaining contract value, in full or a negotiated share such as 50%. Negotiate an exit for repeated outages, a payoff that steps down over time and the right to downsize, move within the provider's portfolio or assign the contract if your company is sold.

How much does colocation cost?

It depends on power, density, market, term and services. Across the primary U.S. markets, CBRE put the average asking rate for 250 to 500 kW deployments at $196 per kW per month in 2025, up about 13% a year since 2020. Retail cabinets and smaller markets price differently, so we get you current quotes for your load.

What else shows up on the bill?

One-time install fees (NRC), cross connects, remote hands beyond any included block, additional circuits, bandwidth and annual escalators. Ask for all of them up front and compare total cost over the full term, not just the monthly rate.

The gotchas: how buyers get nickel and dimed

The monthly rate is the number everyone negotiates. The money is made in everything around it. Here's what we watch for on every deal.

Am I buying more space and power than I need?

Overselling is the most common one. Buyers get sized on nameplate ratings or a rep's guess, then pay for committed kW and cabinets they never fill. Size on measured draw, buy what you need for the next 12 to 18 months and get expansion rights for the rest.

Can I upgrade my power later?

Ask what the whips can handle before you sign. The whip is the cable that runs from the power distribution unit to your cabinet, and it's sized for the circuit you buy. If you start on 208V 30A single phase and need more later, the provider may have to pull new whips and charge installation fees all over again. Have the whips sized for where you'll be in three years, or get the upgrade price and the install fees for it written into the contract now.

What one-time fees should I expect?

Installation fees for cabinets, cages, circuits and cross connects, plus cage builds and sometimes access control setup. All of them are negotiable, and on larger deals many get waived. Get every one listed in the proposal so nothing shows up on the first invoice.

Where do the monthly add-ons hide?

Cross connects, remote hands, extra access badges, storage, shipping and receiving, escorts and after-hours work. Remote hands are often billed in minimum blocks of 15 to 30 minutes with higher after-hours rates. Lock the rates for the full term and ask for a block of remote hands hours included.

Why is my quote so long?

Some providers quote a low base rate and itemize everything else: power, circuits, cross connects, remote hands blocks, IP addresses, cage panels, badges and storage. A quote with 20 line items can look cheaper on the cabinet line and cost more in total. Ask every provider for the all-in monthly total and all one-time fees on one page, for the same space, power and term, then compare those numbers side by side.

What happens if I go over my power?

Draw past the 80% limit on a circuit and you risk tripping the breaker and taking down your gear. Some providers also bill overage on metered power or force an upgrade. Know your draw, watch it monthly and leave headroom for startup surges.

What about renewal and exit?

Contracts often auto-renew unless you give notice 90 to 180 days out, and renewals can reset to list or market price (mark to market), especially after the provider has been acquired. On the way out, expect decommissioning and restoration fees for cages and cabling. Calendar the notice date, cap the renewal increase and get exit costs defined up front.

Terms, MSA and SLA

The contract decides what happens when something goes wrong and what you pay at renewal.

What contract length is typical?

Retail colocation is commonly one to three years. Wholesale deals often run five to ten years, and build-to-suit leases longer. Longer terms usually earn better pricing. Make sure you can grow within the term without starting over.

What is an MSA?

The master services agreement sets the legal terms for the whole relationship: liability, indemnification, insurance, termination and renewals. Each service you buy is added with an order form under it. Negotiate the MSA once, carefully.

What should I negotiate?

Cap renewal increases in the contract itself. Lock pricing for cross connects, remote hands and extra circuits for the full term. Strike automatic renewal or require notice. Get the right to walk away after repeated outages. Make new services end on the same date as your main agreement.

What does an SLA cover?

Power, cooling and sometimes network availability, plus the credits you get when the provider misses them. Check whether uptime is measured monthly or yearly, what triggers a credit, how large it is and whether scheduled maintenance is excluded. Credits are usually a slice of one month's bill for the affected service. They don't cover lost productivity, missed sales or lost revenue, and most contracts exclude those consequential damages outright. The credit isn't your protection. The facility's design and track record are. Ask about incident history, negotiate the right to exit after repeated misses and check your own business interruption insurance.

Moving in

Most delays in a data center move come from the network, not the building.

How long does it take to move into colocation?

A few cabinets in available space can be ready within weeks of signing. The usual long pole is carrier circuits, which often take weeks to a few months to deliver, so order them as soon as you pick the building. Cages, suites and high-density builds add time for power, cooling and cross connect work.

What should I plan for in a data center move?

An inventory of every device and its dependencies, the order workloads will move, network cutovers, a rollback plan, power and cabinet layouts, shipping and receiving, insurance for equipment in transit and a freeze on changes during the move. Keep your old site live until the new one is proven.

Can the provider help with the move?

Many can. Ask about receiving and staging space, secure storage, smart hands for racking and cabling, structured cabling services and partners who handle migrations. Confirm loading dock and freight elevator sizes before your gear arrives.

What amenities should I expect on site?

A place to work: customer work areas or offices, Wi-Fi, meeting rooms, break areas, parking, a loading dock, staging space and secure storage. It matters most during the move and anytime your team is on site for a long night.

Support, outages and maintenance

Most of your experience with a provider happens after you move in.

What are remote hands and smart hands?

Remote hands are on-site staff who handle simple physical tasks for you, such as reboots, cable swaps, visual checks and receiving shipments. Smart hands, or advanced hands, handle skilled work such as installs and configuration. Some contracts include a monthly block of time, and the rest is billed by the half hour or hour.

What should the customer portal show?

Power use per circuit, temperature and humidity, tickets and remote hands requests, cross connect orders, compliance reports and access logs showing who entered your space and when. A good portal saves you from emailing for basic information.

What happens during an outage?

A good provider notifies you right away, keeps you updated and follows up with a written root-cause report. Ask how they've handled past incidents, how they communicate and what changed afterward. Remember that recovery takes longer than the outage itself while your systems come back up.

How much downtime does an SLA allow?

99.9% allows about 43 minutes a month. 99.99% allows about four minutes. 99.999% allows about 26 seconds. Measured yearly, the same percentages allow far more in a single event, which is why the measurement period matters.

How is scheduled maintenance handled?

In a concurrently maintainable facility, maintenance shouldn't drop your load. Ask how much notice you get, when work is done and whether your A and B feeds are ever on the same path during maintenance. Planned work should not be excluded from the SLA at a Tier III facility.

Compliance and certifications

The provider covers the building. You still own your systems and data.

Which certifications matter?

SOC 2 Type II is the baseline for most buyers. SOC 1 matters for financial reporting. ISO 27001 matters for international and security-focused buyers. PCI DSS for card data, HIPAA for health data, and FedRAMP, NIST 800-53 and FISMA for government work. Ask for the current reports, not just a logo on the website.

What is the difference between SOC 2 Type I and Type II?

Type I checks that controls are designed correctly at a point in time. Type II tests that they worked over a period, usually six to twelve months. Ask for Type II.

Is a data center HIPAA certified?

There's no official HIPAA certification. Look for a third-party HIPAA assessment of the facility's physical safeguards, and ask whether the provider will sign a business associate agreement if your compliance team requires one.

What does the provider cover versus me?

Typically physical security, environmental controls, power and access logging. You cover your systems, logical access, encryption and data handling. Ask for a responsibility matrix so your auditors know exactly which controls you inherit.

Rather have someone do this for you?

Send us your requirements. We'll shortlist the providers that fit, get you pricing and help you negotiate.

Figures cited: 451 Research Voice of the Enterprise: Data Centers, Colocation 2025; CBRE North America Data Center Trends H2 2025 (via Pinpoint Capital Advisors, Q2 2026); FEMA flood zone definitions.

Zero risk

Give us one shot

You're not committing to anything. Send us your requirements, see what we come back with and tell us to pound sand if it isn't better than what you'd get on your own.