Atlassian Data Center is priced as an annual subscription in fixed user-tier bands, with Jira Software Data Center listing near $44,000 a year for 500 users in 2026 and rising step by step as the user band increases, so the price jumps at each tier boundary rather than scaling smoothly per user. Each product, Jira Software, Confluence, and Jira Service Management, has its own banded price list, and the total cost of a Data Center estate is the sum of those product subscriptions plus the Marketplace apps and the infrastructure to run them. Understanding the band structure is the key to pricing a Data Center estate, because the position of your user count relative to the next band boundary can move the bill materially.
This guide explains how Data Center pricing works in 2026, the tier bands for each product, the app and hosting costs around them, and how Data Center compares with Cloud. It sits under our Atlassian Cloud pricing guide, links across to Atlassian Cloud migration cost, and is delivered through our software licensing advisory practice.
How Data Center pricing works
Atlassian Data Center is the self-managed, enterprise-grade deployment option, run on infrastructure the customer controls, whether on-premises or in the customer's own cloud account. It is licensed as an annual subscription, not a perpetual license, and the price is set by user-tier bands. You buy the band that covers your user count, and you pay the band price regardless of where in the band your actual count sits. Crossing into the next band at renewal steps the price up to that band's rate.
This banded structure is the single most important feature of Data Center pricing. Two organizations with 480 and 520 users can pay quite different amounts because the second has crossed a band boundary. The practical consequence is that user count management matters not just for the total but for which band you land in, and a small reduction that drops you under a boundary can save a full band step.
Tier bands for the core products
Each Data Center product carries its own banded price list. The table below shows indicative 2026 annual list pricing for the core products at common user bands. Actual figures vary with the price list in effect and any negotiated discount, so treat these as the shape of the structure rather than a quote.
| Product | 500 users | 1,000 users | 5,000 users |
|---|---|---|---|
| Jira Software Data Center | About $44,000/yr | About $63,000/yr | About $190,000/yr |
| Confluence Data Center | About $30,000/yr | About $45,000/yr | About $140,000/yr |
| Jira Service Management DC | Per-agent tiers | Per-agent tiers | Per-agent tiers |
Jira Service Management is the exception to the user-band model: it is priced per agent, the staff who resolve requests, rather than per user, much as other service-desk tools price per fulfiller. That difference matters because the agent count is usually far smaller than the total user population, so JSM can be comparatively inexpensive for a large organization with a small service-desk team. Mixing the per-user and per-agent models correctly is part of pricing a combined Atlassian estate.
Marketplace apps and hosting
The product subscriptions are only part of the Data Center bill. Marketplace apps are licensed to the same user tier as the product they extend, so a heavily extended Jira estate carries an app cost that can rival the core subscription. Apps follow the same banded structure, which means crossing a user band can step up both the product and every app at once.
Hosting and administration are the other side of Data Center economics. Because Data Center runs on infrastructure you provide, the true cost includes servers or cloud compute, database, storage, the staff to patch and maintain the cluster, and the high-availability setup that Data Center is designed for. These costs do not appear on the Atlassian invoice but are real, and they are exactly the costs a Cloud migration is pitched as removing. A fair Data Center cost figure includes them.
Count hosting and staff in the Data Center cost: The Atlassian subscription is only the visible part of Data Center cost. Infrastructure, database, storage, high availability, and the staff to run the cluster are real and recurring. A true Data Center total includes them, which is the only basis on which a Cloud comparison is honest.
Managing the band boundaries
The banded pricing creates a specific optimization opportunity that smooth per-user pricing does not. Because the price steps at each boundary, knowing where your user count sits relative to the next band is worth real money. An estate at 1,050 users sitting just over the 1,000 band pays the higher band for fifty extra users; reclaiming inactive accounts to drop back under the boundary can save the entire step.
This makes user reclamation a direct cost lever on Data Center, not just good hygiene. Dormant accounts, departed employees still licensed, and service accounts that need not be full users all push the count toward the next band. A periodic review that removes them keeps the estate in the lowest band that genuinely fits, and the saving is largest precisely when the count is hovering just above a boundary. The same discipline is covered in our Jira enterprise pricing guide.
Data Center versus Cloud
The live question for most Atlassian customers is whether to stay on Data Center or move to Cloud. Data Center keeps control of infrastructure, data location, and customization, which matters for regulated and highly customized estates, but it carries the hosting and administration burden. Cloud removes that burden and adds Atlassian-managed updates, but converts the cost to per-user subscription that often runs higher for large estates and shifts customization into Atlassian's supported boundaries.
The decision turns on estate size, customization depth, regulatory constraints, and the value placed on offloading infrastructure. Large, heavily customized, or strictly regulated estates frequently have a stronger Data Center case; smaller or standard estates often find Cloud simpler and competitive once hosting savings are counted. The full cost of moving is in our Atlassian Cloud migration cost guide, and the Cloud pricing detail in the Atlassian Cloud pricing guide.
Negotiating Data Center pricing
Data Center pricing is more negotiable than the published band list suggests, especially at the larger tiers and at renewal. Atlassian discounts off list for multi-year commitments and larger bands, and a buyer who arrives with an accurate, reclaimed user count and a credible view of the Cloud alternative negotiates from a stronger position than one who simply accepts the band renewal.
| Lever | Effect |
|---|---|
| Reclaim inactive users | Drop a band, save a full step |
| Rationalize Marketplace apps | Cut app cost tied to the user band |
| Multi-year commitment for discount | Lower effective annual rate |
| Cap the renewal uplift in writing | Protect the price across the term |
| Price the Cloud alternative credibly | Strengthen the renewal position |
The renewal uplift applies to Data Center as it does to Cloud, and capping it in writing protects the price across the term. The strongest Data Center renewals combine a reclaimed user count, a rationalized app list, and a written uplift cap, secured before the renewal date rather than negotiated under its deadline. Firm-side help runs through our Atlassian negotiation practice.
High availability and the cost of scale
Data Center is built for high availability and scale through clustering, which is one of its main advantages over the older Server product and a real cost input. Running a clustered deployment means multiple application nodes, a shared database, a shared file system, and often a load balancer and a separate search index, each of which has to be provisioned, monitored, and maintained. The subscription band buys the right to run that architecture, but the architecture itself is the customer's cost.
For large estates this is precisely the trade Data Center offers: more control and resilience in exchange for carrying the infrastructure and the team to run it. The figure that matters for a budget or a Cloud comparison is the fully loaded one, the subscription band plus the clustered infrastructure plus the staff time, because comparing only the Atlassian invoice to a Cloud per-user price understates what Data Center actually costs to operate at scale.
Common Data Center pricing questions
Buyers raise a recurring set of questions when pricing Data Center. The first is whether the price scales smoothly: it does not, because the banded model steps at each user-tier boundary, so the cost of adding users depends entirely on whether the addition crosses a boundary. The second is whether apps can be dropped to save money: they can, and because apps are tied to the same user band, removing unused apps or reclaiming users to drop a band reduces both the product and the app cost together.
The third question is whether Data Center is being discontinued like Server. Atlassian ended Server sales and support and has steered investment toward Cloud, but Data Center remains the supported self-managed option, which is why the realistic decision for most large estates is Data Center versus Cloud rather than Data Center versus Server. The fourth is whether the renewal price is fixed: it is not, because the renewal uplift applies, which is why a written cap on that uplift is as important on Data Center as on any subscription.
A final question concerns mixed estates that run several Atlassian products. Because each product is priced on its own band, and Jira Service Management on a per-agent basis, the total is the sum of separately banded subscriptions rather than a single blended rate. Pricing the estate accurately means pricing each product against its own band and agent count, then adding the shared app and infrastructure cost, which is the level of detail a serious renewal preparation requires.
Buyers also ask how far in advance the band and renewal work should start. The practical answer is at least three to six months before the renewal date, because reclaiming inactive users, rationalizing apps, and confirming which band the estate truly needs all take time to do properly, and arriving at the renewal with that work finished is what converts the banded list price into a negotiated one rather than an accepted one.
The action plan
To price and control a Data Center estate, start by identifying the band each product sits in and how close the user count is to the next boundary. Reclaim inactive users to hold the lowest band that fits, rationalize the Marketplace apps tied to that band, and add the real hosting and administration cost to see the true total. Then decide the Cloud question on that complete figure, not the subscription alone.
The recurring lesson is that Data Center cost is governed by the band boundaries and the hidden hosting cost, not the headline subscription. Managing the count against the bands and capping the uplift are the levers that hold the price down. For an independent Data Center cost review and negotiation support, see our Atlassian practice and software licensing advisory.