The short answer
A useful first estimate hides in plain sight: look up the longest warranty you can buy for each piece of hardware in the exhibit. That number is the industry telling you how long it expects the thing to live — and across commercial displays and business computers alike, it tops out around five years. So an exhibit specified for a ten-year horizon — the figure most museums plan around — will spend its second half entirely outside anyone’s coverage, with at least one mid-life hardware refresh guaranteed and continuous smaller costs the whole way through.
For the recurring line itself, the honest benchmark depends on what the exhibit is made of. Where AV hardware dominates, the commercial AV industry’s standard service contract runs 8–12% of installed system cost per year, and public awards bear that out — from $5,750 a year maintaining a single interactive table to six figures museum-wide (browse the maintenance-contract rows yourself in The Exhibit Cost Record). Software-heavy exhibits work differently: support there is more standardized and mostly detached from build cost, with service contracts generally running $5K–$20K a year depending on scope. Our own model is a one-year warranty followed by prepaid support-hour banks rather than a standing retainer — but however yours is structured, the budget line belongs in the plan from day one, next to the staff hours below.
Read the warranty as a forecast
Manufacturers won’t warranty hardware for longer than they expect it to survive — that’s the whole business model of a warranty. Firms that install the hardware follow the same logic, which is why an integrator’s coverage rarely outruns the manufacturer’s. The major commercial display lines cap total coverage — standard plus every extension you can buy — at five years; business computers cap at five as well. That isn’t a sales tier: it’s a lifetime estimate from the people with the best failure data in the world.
The panels themselves are rated in hours — commercial LCDs typically 50,000 to 60,000 hours to half-brightness — which at exhibit duty cycles works out to roughly eight years or more of usable life. Note what that rating actually promises: not that the display dies at the end, but that it has dimmed to half of what it was. A gallery doesn’t get a dramatic failure; it gets a slow fade nobody notices until a new panel goes in next to an old one. Budget the replacement at purchase time, not when the fade becomes visible.
“Fixed static” software still moves
Even software you never touch is sitting on an operating system that changes underneath it. A “finished” exhibit still has to survive Windows updates — which periodically require modest but real changes to keep things running — along with driver updates, security patches, and the slow drift of every service it talks to. None of this is a rebuild; all of it is somebody’s time, several times a year, for the life of the exhibit.
And most exhibits aren’t static anyway. Content-management systems, web services, remote monitoring — these carry real recurring costs even when you host them yourself. On-premises isn’t free; it just moves the bill from a subscription line to server hardware, backups, and staff hours.
The hours nobody budgets
The largest ownership cost is often invisible because it’s salaried: how much time does your technology, IT, or operations staff spend starting systems up, shutting them down, and responding when something fails? How much time does content staff spend in the CMS keeping interpretation current? A gallery of interactives can quietly consume a meaningful fraction of a staff position — and because no invoice arrives, it never shows up in the exhibit’s cost story. Put it there deliberately, in hours per week, when you plan.
The content-management system you won’t use
Institutions almost always ask for a content-management system, usually out of a healthy fear of being locked into their original firm for every text change. The fear is legitimate. But here is what actually happens in the majority of institutions: nobody uses it — because updating exhibit content is nobody’s job. The CMS was specified as insurance, and insurance you never claim still gets paid for: in build cost, in hosting or server upkeep, in the update-and-patch obligations of one more running system.
The honest question isn’t “do we want the ability to change content?” — everyone does — it’s “whose job will it be to change content, and how many hours of it do they have?” If there’s a real answer, a CMS is one of the best investments in the building. If there isn’t, fixed content honestly specified is cheaper to build, cheaper to own, and no less capable in practice than a CMS nobody logs into. We build and sell a content-management system, so read this advice accordingly: we would rather you not buy one you won’t staff.
The year-5 cliff
Year five is when the ownership costs stop being gradual. The coverage on nearly everything expires around the same time — the five-year warranty ceilings on displays and computers are synchronized clocks — just as the operating system wants updates the software wasn’t built for and a display model is discontinued. The original firm may have restructured or moved on by then; and even a firm still standing may no longer have the person, because much of this industry builds with freelance technical talent, and the developer who wrote your exhibit’s code may have rolled off the project six months after opening. None of this is a surprise individually — warranty horizons, product cycles, and staff turnover all run on roughly five-year clocks — but institutions that budgeted the exhibit as a one-time capital project meet all of it at once, unfunded.
The durable mitigation for the people problem is artifacts, not loyalty: documentation, source code, and a real handoff are the only continuity that survives both your staff turnover and your firm’s. And be clear about who holds them — don’t rely on the original creator to maintain the documentation for you. Assume the documentation is yours to keep current, the way you’d keep the building’s drawings. Held that way, it works in every direction: when something needs fixing in year seven, you can hand it to the original firm and whoever is on their roster that year — or to anyone else.
What “maintenance” should actually mean
Maintenance contracts span four very different things, usually without saying so: break/fix (someone comes when it dies), preventive care (someone comes before it dies), content support (someone helps your staff change what it says), and platform stewardship (someone owns the software’s survival across OS updates and hardware refreshes). Most contracts sell the first and imply the rest.
When you evaluate one, the clauses that matter are not the ones that get negotiated hardest. Guaranteed response time is overrated — a dead kiosk for three days is an annoyance. An escalation path and a parts strategy are underrated — a discontinued display with no spares plan is a dead exhibit.
Design decisions that set the ownership bill
The choices made in month two of design dominate what year eight costs. The technology stack decides who has to maintain the software and how often (we wrote a separate guide on that, because it’s a governance decision disguised as a technical one). Commodity hardware decides whether year-6 replacements are a purchase order or an engineering project. Staff-editable content systems decide whether interpretation updates cost a phone call or a contract. Documentation and source-code handoff decide whether anyone but the original firm can ever touch it. This is lifecycle strategy as a design requirement — the argument of our discipline, in budget form.
Budgeting it honestly
Put ownership in the capital ask, so operations isn’t surprised into it: an annual maintenance line, a mid-life hardware refresh planned for year five or six rather than hoped past, and the staff hours named. Institutions with the means should consider endowing exhibit maintenance the way they endow positions — the exhibit is infrastructure, and infrastructure without an operating budget is deferred maintenance with a ribbon on it.