The number 75 comes from a footnote. Service Life Design for Bridges, the summary guide produced under the second Strategic Highway Research Program and dated 29 April 2019, tabulates design life as the period of time on which the statistical derivation of transient loads is based, and gives the value as 75 years for AASHTO LRFD. Its footnote to that row is the part worth memorising. Load and resistance factors for structural design in AASHTO LRFD were calibrated to the 75-year period, the guide says, but deterioration of the structure over time was not explicitly considered, and there is no direct relationship between the AASHTO LRFD definitions for service life and design life.
So 75 years describes how far out the load statistics were pushed, a calibration boundary for transient loads rather than a durability promise about how long the steel will still be there.
The same words, two incompatible readings
FHWA’s Bridge Preservation Guide, publication FHWA-HIF-18-022, defines design life as the period for which a component, element or bridge is expected to function for its designated purpose when designed, constructed and maintained as per standards. Its commentary adds that design codes and material specifications are important parameters in determining that expected life. On that reading design life is a durability expectation.
On the SHRP2 reading it is a calibration window for load models and carries no durability content at all. Both documents are current federal-sector guidance aimed at the same engineers, and a project requirement that borrows the phrase from one while taking the number from the other produces a specification nobody can verify against anything.
The instability is not confined to design life. The SHRP2 guide’s terminology table lists four separate rows for the term service life, each citing a different reference, plus a further row for design service life. One of those four is the Bridge Preservation Guide’s own formulation: the period for which a component, element or bridge provides the desired function and remains in service with appropriate preservation activities. Another describes the time during which the structure performs its design function without unforeseen maintenance or repair. Those two are not the same test. A bridge kept in service by scheduled overlays satisfies the first and fails the second.
What a 100-year requirement actually adds
Long service lives are being specified. The SHRP2 guide names three recent projects carrying a 100-year service life requirement for non-replaceable components: the Osman Gazi crossing of Izmit Bay in Turkey, and in the United States the Gov. Mario M. Cuomo Bridge and the Abraham Lincoln Bridges. The qualifier at the end of that sentence carries the engineering. The requirement attaches only to what cannot be swapped out.
Which components those are is a project decision rather than a physical fact, and the guide’s own examples are instructive about how far the line moves. Its text gives foundations, substructure and decks as commonly non-replaceable, and bearings and joints as commonly replaceable. Its worked example from a cable-stayed signature bridge, presented as strictly exemplary rather than as a standard, sets 100 years for towers, foundations, abutments, piers, pier caps, deck, superstructure, approach slabs and mechanically stabilised earth walls, then 75 for drainage system piping, 60 for stay cables, 50 for bearings, 40 for barriers, 30 for expansion joints and 25 for the separate deck wearing surface and for painting and coatings. The deck is a hundred-year element in that table and the surface running over it is a twenty-five-year one.
The guide also states the logic that makes such a table coherent: requiring a 25-year or 50-year service life for replaceable components on a bridge with an overall 100-year requirement would be logical, and scheduling several component replacements to coincide limits the disruption. Accessibility for those replacements therefore becomes part of the durability specification rather than a detailing convenience.
Following the code produced reliability indices from negative 1.0 to over 5.0
The same guide reports what happened when the prescriptive route to durability was tested probabilistically. Summarising a benchmarking exercise published in fib Bulletin 76, it states that after 50 years the use of AASHTO LRFD deemed-to-satisfy provisions achieved reliability indices as low as approximately negative 1.0 for unfavorable combinations and up to greater than 5.0 for favorable ones. For the unfavorable case, a reliability index of negative 1.0 equates to a greater than 80 percent probability of failure in achieving the limit state, and the limit state in question is reinforcement depassivation, meaning the onset of corrosion rather than collapse. Favorable combinations can exceed the typical reliability index of 1.3.
The spread is the finding. Following the code produces durability outcomes ranging from comfortable to worse than a coin flip depending on the exposure the designer assumed and the concrete requirements permitted, and the guide notes that AASHTO LRFD does not quantify chloride exposure from de-icing chemicals or sea water at all. The cover requirements it does give are 4.0 inches for direct exposure to salt water, 3.0 inches for coastal and 2.5 inches for exposure to de-icing salts, before any water-cement modification factors.
The deck is where design life and service life part company
Actual service life diverges from any design figure mostly through the riding surface, and the intervals are short enough to be surprising. The Bridge Preservation Guide offers a sample cyclical agency rule for decks rated 7 or better on NBI Item 58: deck sweeping and washing every one to two years, crack sealing every three to five, deck sealing every three to five, a polymer overlay every eight to twelve, a polymer-modified asphalt overlay every twelve to fifteen. Its sample condition-based rule credits an epoxy overlay with a service life of 10 to 15 years and a deck patch with 3 to 10. All of those are presented as examples of what an owner might adopt, not as national values.
Set them against 75 years anyway and the arithmetic is unforgiving. A bridge that reaches that age has absorbed something like five to nine overlay cycles and fifteen or more crack-sealing passes, and the guide states the dependency without hedging: a steadfast bridge preservation program and quality workmanship practiced during the service life of an asset is necessary for the asset to reach its design life. Design life is contingent on work that nobody has yet funded on the day the bridge opens.
Loading pushes the same way. The deterioration modes that federal inspection policy asks risk panels to score include section loss, fatigue and fracture in steel, and reinforcing and prestressing steel corrosion in concrete, and every one of those accumulates with truck passes and with chloride rather than with calendar years. Two structures of identical design, one carrying heavy freight through a salted winter and one carrying light rural traffic in a dry climate, do not share a service life in any useful sense. Which material shifts those curves, and at what cost, is taken up in steel, concrete and the future of bridge materials; the condition ratings that record the divergence are explained in how bridges are inspected and rated; and the wider set of durability, funding and structural questions sits in bridge engineering coverage.
What the inventory is actually aged at
The stock answer to how long bridges last is an average age, and it deserves less respect than it gets. ASCE’s 2025 Report Card for America’s Infrastructure states that there are 623,218 bridges in the country, with an average age of about 47 years, and attributes its condition percentages to FHWA’s Bridge Condition by Highway System table for 2024. FHWA’s own 2025 edition of that table counts 624,193 bridges, of which 272,779 were classified Good, 309,729 Fair and 41,685 Poor, with Guam not submitting data that year. The two counts differ because they are different reporting years.
An average of 47 years describes a stock built in bursts. It contains Interstate-era structures entering their seventh decade and structures opened last season, and the mean sits between populations with different design codes, different detailing and different maintenance histories. The distribution answers the question and the mean does not, which is why FHWA publishes Bridge Condition by Year Built as a separate table cross-tabulating condition against construction era.
Which elements has the owner agreed are non-replaceable? What target was set for them, and against which limit state? Was that target checked by a probabilistic model or by a cover table? And who has committed to the overlay cycle that keeps the answer honest? The parallel argument about road surfaces, where the same gap between a design figure and a maintained outcome shows up in pavement, is taken up in how long a modern highway should last. Those four questions are answerable on any project, and none of them is how long the bridge will last.