Retail vs Wholesale Colocation: How to Choose

Colocation sales conversations now run in two very different vocabularies, and deciding which one applies to your workload has become one of the most consequential infrastructure calls a team can make. One conversation prices a monthly bundle per rack or per cabinet, with some kilowatts and a port included, and promises move in within weeks. The other prices a private suite by committed kilowatt, hands you a space you largely operate yourself, and ties you to a term measured in years. Teams that choose between the two by whichever quote shows the smaller number tend to either stretch a three rack workload into a five year committed suite or pay retail rates for capacity that should have been bought at wholesale scale, and both mistakes are expensive to reverse because neither is a quick contract to exit.

This guide explains what each model actually sells, where the total cost truly diverges, and how to match the lease model to the behavior of your workload rather than the momentum of a sales cycle. It covers the pricing mechanics that hide the difference, the space and control each model puts in your hands, the contract terms that decide how locked in you are, and the hybrid strategies that are becoming the norm as enterprises run a retail edge and a wholesale core at the same time. By the end you will be able to normalize any quote to dollars per committed kilowatt and know which model your workload will survive inside.

At a Glance: Retail vs Wholesale Colocation

What Retail and Wholesale Colocation Actually Are

Retail colocation is the model most people picture when they hear the word colocation. You rent a rack, a cabinet, or a cage inside a shared data hall that the provider operates and that other tenants use alongside you. The provider runs the cooling, the power plant, the security, and the access control, and it answers for the environment under an SLA rather than leaving the building to you. Your monthly bill is built around the space unit, with a block of power, a cross connect or two, and sometimes bandwidth bundled into the rate. Move in dates are measured in weeks because the space already exists and the provider’s job is to hand over keys rather than build anything.

Wholesale colocation flips the unit of sale. Instead of renting space by the rack, you lease a dedicated capacity block, a private suite, a data hall, or a floor of a building, priced per committed kilowatt. You and the provider agree on a power commitment that usually starts somewhere between 100 and 500 kilowatts and can scale to multiple megawatts, and the suite is separated from every other tenant, often with its own cooling plant. The provider delivers the shell, the power, and the cooling infrastructure, and you, or a contractor you hire, fit out the white space, choose the containment, the racks, and the PDUs, and run the suite from then on. Deployment is measured in months rather than weeks, and the lease typically stretches three to ten years, because the buildout is a capital commitment the provider needs time to earn back.

Pricing: Per Rack vs Per Kilowatt

The pricing mechanics are where the two models stop sounding similar. Retail quotes are bundles. A rate card lists a monthly figure per rack or per cabinet with a defined kilowatt allowance, an included cross connect or two, and a menu of adders for extra power, extra ports, remote hands, and after hours access. The bundle means you pay for capacity you may not use, which is fine when the included amount matches your draw and quietly expensive when it does not. Wholesale quotes separate the components. You pay a monthly rate for each committed kilowatt, a metered pass through for the energy the suite actually consumes, and a one time buildout for the fit out, and every additional dollar is explicit and negotiable.

That split is why the same market carries two very different price tags. In 2026 US primary markets, retail pricing lands roughly between $200 and $400 per kilowatt per month once you divide the bundle by its included power, while wholesale asking rates for 250 to 500 kW blocks sit closer to $160 to $220 per kilowatt per month and fall further still at megawatt scale. The gap is real, but it is paid in the currency of commitment: power minimums, multi year terms, and a buildout you fund. The line between the two models has also been moving, because power scarcity in major hubs has prompted some operators to raise their retail minimums to 50 or even 100 kW and others to push the wholesale threshold from 250 kW to 500 kW or 1 MW, so the boundary that applied last year is worth confirming this year.

Before comparing any two quotes, normalize them to dollars per committed kilowatt, because a per rack bundle and a per megawatt lease otherwise look like different currencies. Divide the retail bundle by its included kilowatt allowance, add the cross connects and remote hours you know you will use, and place the wholesale rate next to it with the buildout amortized across the term. Derive the committed kilowatt figure from the number of racks you will fill and the density per rack you intend to run, because a quote is only comparable when the assumption behind it is a number you can defend rather than a number your sales contact supplied.

The utilization trap applies in both directions. A wholesale commitment that runs at half load is more expensive per delivered kilowatt than a retail bundle that matches your draw, while a retail deployment that grows past its included power pays overage rates that quickly exceed what the same capacity would have cost at wholesale. Project utilization across the full term, not the first quarter, and price the honest scenarios: the flat baseline, the growth ramp, and the early exit. Teams routinely discover that a model is only cheaper under the scenario they neglected to model.

Space, Control, and Who Runs the Environment

Retail sells you an environment as much as it sells space. The provider owns the physical design, the temperature, the maintenance schedules, and the security, and your operational obligation is limited to your own hardware. Remote hands handle tasks your team is too far away to do, managed colocation layers monitoring and patching on top, and when a cooling unit fails at 3 AM the facility’s staff notice before you do. The trade off is that the environment is shared and standardised: density is bounded by what the hall was built for, changes like raising your per rack power require a conversation, and you live with the constraints that suit the building’s average tenant rather than your specific workload.

Wholesale puts that control in your hands and the cost of exercising it on your books. Density is a specification you choose, which matters more every quarter as AI and GPU workloads push racks far past the 5 to 10 kW envelope that legacy retail halls were built around and into 20 to 100 kW racks that need liquid cooling and heavier electrical distribution. Physical isolation is a specification too, for regulated, compliance sensitive, or simply security conscious tenants who do not want their gear sharing a hall with strangers. But the suite is only an advantage if someone runs it. Chilled water plant, CRAC units, fire suppression, and switchgear need preventive maintenance, monitoring, and incident response, and in a wholesale model that work lands on you or on a managed services contract you pay for separately, which is the line item most first time wholesale tenants underestimate.

Contract Terms and Exit Costs

Retail and wholesale contracts are written for different product lifecycles, and the difference shows up in how free you are to change your mind. Retail terms commonly run 12 to 36 months, with a modest early exit or a short tail to run out your equipment. That short commitment is an operational feature: you can add a market, shrink a region, or replace a facility at the cost of a quarter, and the flexibility shows up in the price you pay per kilowatt. Wholesale terms run 3 to 10 years because the provider builds capacity around your commitment, and the buildout is amortized into the rate, which is precisely why the rate looks good. The honest way to read a long term rate is as the sum of your commitment plus the provider’s security, not as a market price for capacity.

Two clauses decide how a long lease feels in its third year. The first is the renewal escalator, often 3 to 5 percent a year, which quietly compounds past the headline you negotiated. The second is expansion rights, contractual options on adjacent capacity at a defined price, which in today’s power scarce market have become the most valuable line in any wholesale agreement because new capacity in major hubs is pre leased years out. Exit terms matter as much as entry terms: understand the decommissioning obligation, whether the suite can be subleased, and what happens to stranded capacity if the workload falls short of the commitment. A retail exit is a cabling job; a wholesale exit is a contract negotiation that started the day you signed.

Connectivity: The Retail Ecosystem Advantage

For workloads that live on interconnection, retail colocation is a market and wholesale is a building. A retail hall or carrier hotel hosts dozens of carriers in the building, provisions a cross connect in days, and offers direct cloud on ramps, internet exchanges, and fabrics you can join with a quick ticket. That density is why multi cloud architectures, inference workloads, and latency sensitive services cluster in retail facilities: the value is not the rack, it is the ecosystem around the rack, and it arrives already plumbed and priced.

A wholesale suite can reach the same connectivity, but you assemble it. Cross connects into the meet me room are still available, and large tenants negotiate dark fiber or dedicated wavelengths, but the suite has no default network, and the paths, the fiber entrances, and the carrier contracts are yours to build and to negotiate. For a stable core that talks mostly to your own sites and clouds over long term circuits, wholesale is perfectly adequate. For a workload that needs to reach many networks, peer around, and change links often, the retail ecosystem is a structural advantage that no power discount fully compensates for.

When Each Model Wins

The decision tree is short and honest. Retail wins when your footprint is small or variable, when you need to be live in weeks, when your value comes from interconnection, or when your team has no appetite for operating a data hall. Wholesale wins when your capacity is large and stable for years, when density or isolation forces a custom environment, or when the core of your architecture justifies trading operational simplicity for a much lower cost per kilowatt. The middle, roughly 50 to 250 kW, is genuinely contested, and the right answer depends on the three variables your forecast controls: how stable the load is, how long you can commit, and who runs the environment.

Size the workload first, because the model follows the number. A deployment that needs 20 racks at 10 kW each for the next five years is a wholesale candidate on paper, but only if the load is actually that predictable and your team can absorb the operations. The same racks behind a workload that could be half that size next year belong in retail, where you pay the flexibility premium but keep the option. Forecasting the full term honestly, including the scenarios where the business changes direction, is worth more to the decision than any pricing benchmark, because the cost of being wrong in wholesale is counted in years.

Combining Both: The Hybrid Model

Most enterprises that scale past retail, and many that have always been at wholesale scale, end up running both models at the same time, and the pattern has a logic worth copying. The stable core, the capacity that will still exist in five years and needs the lowest possible rate per kilowatt, goes into a wholesale suite. The variable layer, the edge presence, the interconnection dense services, the markets where you want optionality rather than commitment, lives in retail facilities across one or several metros. Wholesale minimums stay full because the core stability underwrites them, and retail absorbs the flexibility without forcing you into an expensive construction project every time demand moves 30 percent.

The hybrid is also where the discipline of running multiple facilities separates teams that manage it from teams that drown in it, because capacity stops being per site and becomes per workload. Wholesale minimums have to stay full, retail has to stay lean enough to flex, and the circuits that bind the whole footprint together are bought with the total workload in mind rather than one facility at a time. Teams that treat the fleet as a single system stay ahead of the capacity conversation instead of discovering it at a renewal.

The Decision in Practice

Start the decision early enough to act on it. Build the rack plan and the power forecast, decide how much stability you can honestly commit to, decide who operates the environment your workload requires, and then price the whole term in dollars per committed kilowatt including buildout, energy, cross connects, and escalation. Map the two models against that forecast and let the honest number choose for you rather than the pitch. For most teams the answer is retail for everything until the workload proves it deserves wholesale, which is a position you can always revisit because retail is cheap to leave.