How to Price Dedicated Servers: Cost Floor, Margin and the Market

Dedicated server pricing gets set one of two ways: copy a competitor and shave a little off, or add a percentage to the hardware cost and hope. Both produce prices that are wrong in ways you will not notice for a year.

Here is how to build the number from the bottom, and what to do with it once you have it.

Build the cost floor

You cannot price without knowing what a server costs you per month, and almost nobody's first attempt at that number is complete. The components:

Hardware, amortised. Purchase price divided by the service life you actually get, not the accounting life. A machine bought for a three-year depreciation schedule that you keep running for five has a lower monthly cost in years four and five — which is real margin and also a maintenance risk.

Power. Draw at realistic load, not nameplate, multiplied by your all-in cost per kWh, multiplied by your PUE — because every watt to the server costs you more than a watt at the meter once cooling is included. This is frequently the largest recurring component and the one most often estimated rather than measured.

Space. Rack units consumed, at your cost per U — whether you own the facility or lease the rack.

Connectivity. Port cost plus the bandwidth actually consumed at your blended transit cost, plus your IPv4 cost, which is no longer negligible.

Failure and spares. Drives fail, PSUs fail, RAM fails. A percentage per server per month for parts and the labour to fit them, plus the cost of holding spares.

Support. Your average ticket load per server per month at your loaded staff cost. This one varies enormously by customer segment and is the reason a cheap self-managed server and an expensive managed one can have similar margins.

Unsold capacity. The uncomfortable one. If you run at 80% utilisation, the 20% sitting idle is a cost your sold servers have to carry. Pricing off a 100%-utilised model guarantees you lose money.

Then price the offer, not the hardware

Cost gives you a floor. It does not give you a price, because customers are not buying components — they are buying an outcome with a set of guarantees attached, and those guarantees are most of the value.

The same hardware supports very different prices depending on delivery time, network quality and redundancy, DDoS protection, the SLA and what it actually pays, support responsiveness, and the out-of-band access and self-service that determine whether the customer ever needs you. A provider competing purely on specification is in a commodity market by choice.

Where the market puts a ceiling

Your price has to survive comparison. Not with everyone — with the providers your customers actually consider, which is usually a short list in your region and segment.

Two disciplines here. Compare like for like: an unmetered 10 Gbps port and a 1 Gbps port with a transfer cap are not the same product even on identical hardware, and comparison tables routinely pretend otherwise. And check the total, not the headline: setup fees, IP charges, bandwidth overage and remote-hands rates move the real number substantially.

If your price sits above the market, you need a stated reason a customer can evaluate. If it sits below, know why — and if the reason is that you have not costed it properly, that is the finding.

The structural decisions

Setup fees. They recover the real cost of racking and provisioning and they suppress conversion. The common resolution is a setup fee waived on longer terms, which recovers the cost from customers who leave quickly and rewards the ones who stay.

Bandwidth model. Included transfer, unmetered port, or committed rate with 95th-percentile overage. This decides who your customers are more than any other choice on the page, because each model attracts a different usage profile — and a generous transfer allowance is an invitation to the exact customers who will consume it.

Term discounts. Meaningful on dedicated hardware because your cost is genuinely committed once the machine is racked — see annual versus monthly billing.

Older hardware. Fully amortised machines can be sold cheaply at real margin. Do it as a clearly labelled lower tier rather than by quietly shipping older hardware at the current price.

Review it on a schedule

Every input above moves. Power costs move sharply, hardware costs move, transit costs fall, IPv4 costs rise. A price set two years ago against inputs that have all changed is not a price, it is a historical artefact.

Recompute the floor annually, at minimum, and sort your existing customers by margin while you are there. The accounts that have fallen below the floor are the ones to act on, and acting on forty of them is a very different exercise from a blanket increase — see raising hosting prices without losing customers.

How FluxBilling fits

The inputs to this calculation are exactly the records an integrated platform already holds: which hardware exists, what it cost, where it is racked, what power is allocated to it, which IP addresses it holds, and which customer and invoice it is attached to. FluxBilling keeps hardware inventory, rack and power allocation, IPAM and the billing record in one schema on every tier, so revenue per machine and revenue per rack unit are queries rather than a reconciliation between a spreadsheet and an invoice list.

Products carry per-location pricing, which matters once power and transit costs differ by site — see billing across multiple datacenters.

An honest limit: FluxBilling does not model your cost base or compute margin for you. It holds the operational facts the calculation needs; the cost model is yours to build.

Closing thoughts

Compute the floor for one server this week — a real one, with its actual power draw and its actual share of your unsold capacity. Compare it to what you charge. For most providers who have never done it, the answer on at least one product line is uncomfortable, and it is far better to find it in a spreadsheet than in a year-end review.