Two colocation quotes can both show $180 per kW per month and still be thousands of dollars apart by the time the invoice arrives. One may be net of electricity. Another may bundle power into the recurring rate. Cross-connects, remote hands, connectivity and one-time charges may sit outside both.
That is the central problem with colocation pricing in 2026: $/kW/month is the right unit, but it is not automatically a comparable price. Before deciding whether $160, $200 or $380 per kW is expensive, the quote has to be normalized to the same scope.
These figures come from McKinsey, Colliers and Lightyear and are intentionally shown separately. They use different definitions, so combining them into one “average U.S. colo price” would create false precision.
The most defensible 2026 benchmark depends on the product
For large wholesale and multi-tenant deployments, current research converges around the high-$100s per kW per month. Colliers' 2026 Data Center Marketplace report says primary U.S. markets generally price at $160–$200/kW/month, net of electrical charges, with secondary markets typically 10%–15% lower.
McKinsey's 2026 analysis reaches a similar range from the business-model side: wholesale leases typically earn around $150–$200/kW/month, while retail colocation is higher at roughly $200–$380/kW/month. The spread reflects more than scale. Retail customers tend to buy flexibility, smaller increments, network density and a heavier service layer.
A separate quarterly index from Colocation Price Index currently places the North American wholesale average at about $196/kW/month, excluding metered power, and Northern Virginia at about $215/kW/month. That is useful as a live market check because it reinforces the same broad conclusion: scarce, deliverable power in major markets is keeping wholesale pricing around or above the upper-$100s.
Why Lightyear's $380 median is not in conflict with a $180 wholesale quote
Lightyear's current pricing guide is especially useful because it labels its metric clearly. Its H2 2025 retail median is $380 per usable kW, and “all-in” means the space cost plus power cost, while cross-connects and connectivity remain outside the figure. The 10th percentile was $236 and the 90th percentile $663.
Six months earlier, the median was $324. Lightyear therefore recorded a 17.3% increase between H1 and H2 2025, with increases also visible at the low and high ends of its distribution. Its 2026 colocation pricing guide attributes the current environment to tight capacity, rising renewal rates and weaker volume discounts in constrained markets.
A $180/kW wholesale rate that excludes electricity and a $380/kW retail rate that includes space and power are not rival estimates of the same product. They are closer to two different layers of the market.
Net-of-power pricing
$ / kW / month
The recurring colocation charge is quoted separately from metered electricity. This is common in larger deployments and is the basis used by Colliers for its primary-market range.
Power-inclusive pricing
space + power ÷ usable kW
Space and power are combined into an all-in kW rate. Lightyear's retail benchmark uses this approach, but still excludes connectivity and cross-connects.
What is actually inside a colocation bill?
The base kW rate is only one line. A clean comparison starts by separating the recurring infrastructure charge from energy, network and support costs.
Contracted kW multiplied by the agreed monthly $/kW rate.
Either embedded in the rate or passed through separately based on metering and contract terms.
Physical interconnections to carriers, clouds, exchanges or other tenants, often charged monthly per connection.
Internet transit, wavelength, cloud connectivity or other network services if purchased through the facility or partner.
On-site technical labor for reboots, cabling, swaps and physical intervention.
Installation, cabling, cage build-out, setup or other one-time work required to turn the environment up.
This is why a procurement spreadsheet should never have a single column called “colo price.” At minimum, it needs to distinguish the base recurring charge, electricity treatment, network/interconnection fees and one-time charges.
A 100 kW quote can look cheap until the scope is normalized
Consider a fictional 100 kW deployment in a primary U.S. market. The numbers below are illustrative, not market quotes; the purpose is to show how the billing structure changes the comparison.
The electricity line above is simply 75 kW × 730 hours × $0.10/kWh. It is not a prediction of what a specific facility will charge. Utility tariffs, PUE allocation, demand charges, taxes, energy adjustments and provider markup can all change the real pass-through structure.
The example does show why a nominal $180/kW quote cannot be compared directly with a $300 or $380 all-in quote until the power treatment is known. If electricity is already embedded in the higher quote, part of the apparent gap disappears immediately.
Primary versus secondary markets: the discount is real, but not unlimited
Colliers currently estimates that secondary U.S. markets transact at roughly a 10%–15% discount to primary markets on colocation pricing. Applying that mechanically to a $180/kW primary-market reference would imply roughly $153–$162/kW before electricity.
The important qualification is that “secondary market” does not mean permanently cheap. Colliers notes that pricing dispersion is increasingly driven by deliverable power, lease term, facility condition, tenant credit and execution certainty rather than the old retail-versus-wholesale label alone. Power scarcity is pushing demand into alternative markets and narrowing some of the historical differences.
That trend is visible in market-index data too. Colocation Price Index currently places Northern Virginia at $215/kW/month for wholesale capacity, versus a North American average of $196, and reports a much faster year-over-year increase in Ashburn than in the region overall.
Why more committed kW does not guarantee a big discount anymore
The traditional expectation was straightforward: commit to more power, sign a longer term and receive a better unit rate. Scale still matters, but constrained capacity has weakened that relationship in the hottest markets.
Lightyear reports that large requirements above 1 MW are no longer consistently receiving the volume discounts buyers once expected because contiguous power blocks are difficult to secure. Colliers reaches a similar conclusion from the investment side: vacancy in primary U.S. markets is around 1.3%, and functional availability in the tightest hubs can be effectively zero.
In other words, a buyer asking for 2 MW is not merely a larger version of a buyer asking for 100 kW. The provider has to decide whether a contiguous block of scarce deliverable power is worth committing to that tenant, in that building, for that term.
How much contiguous capacity is required, and when?
Is usable inventory actually available in the target metro?
Longer commitments can help, but scarcity can outweigh traditional discounts.
High-density requirements may restrict the number of technically suitable halls.
Tenant quality can influence pricing and commercial flexibility.
Inclusive and pass-through models must be normalized before comparison.
Retail pricing is a different buying experience
Smaller deployments often enter colocation through a cabinet, partial cabinet or small cage rather than a wholesale power block. The physical rack matters more at this end of the market, but power still becomes the economic constraint as density rises.
McKinsey's $200–$380/kW/month retail range captures the premium associated with smaller, more flexible deployments. Lightyear's $380 all-in median sits at the top of that range, which is plausible given that its benchmark explicitly incorporates space and power and reflects a market that tightened sharply through 2025.
For a small buyer, the best unit price is not necessarily the cheapest contract. A facility with richer interconnection options may reduce network spend elsewhere; included remote-hands allowances may matter operationally; and a lower-cost building with no expansion headroom can become expensive if the deployment has to be split later.
How to compare two colocation quotes without fooling yourself
Normalize committed kW. Make sure both proposals are pricing the same capacity and redundancy assumption.
Identify the power model. Mark electricity as included, metered pass-through or separately calculated.
Add recurring extras. Cross-connects, bandwidth, support and management fees belong in monthly occupancy cost.
Annualize one-time charges. Installation and build-out costs should not disappear simply because they are labeled NRC.
Model the contract term. Escalators and renewal assumptions can matter more than a small first-year rate difference.
Price expansion risk. Check what the next 20, 100 or 500 kW would cost and whether it can be delivered in the same facility.
The cleanest output is a three-year or five-year total occupancy cost, not a single $/kW headline. That is also the logic behind the Colocation Cost Calculator planned for Data Center Scope: the useful result is the full recurring cost structure and its sensitivity to power, not just the base rate.
What a reasonable 2026 planning assumption looks like
For a large U.S. deployment, $160–$200/kW/month before electricity is a defensible primary-market screening range based on Colliers, with wholesale market references from McKinsey and Colocation Price Index clustering in a similar band. Secondary markets may begin roughly 10%–15% lower, but that discount should be verified against actual power availability.
For smaller retail deployments, $200–$380/kW/month is a more appropriate broad range, with Lightyear showing a $380 all-in median in the second half of 2025 and a very wide distribution around it. Those figures should not be treated as quote substitutes; they are useful for spotting proposals that need a closer explanation.
The practical lesson is less satisfying than a single national average, but more useful: the number after the dollar sign matters only after the billing scope is understood. In colocation, $/kW/month is the beginning of the comparison, not the end.