National Bridge Inventory (NBI) ASCII Files

It is FHWA's census of American bridges, held as typed fixed-width rows rather than a map layer or a dashboard. States, federal agencies and Tribal governments report their structures under the National Bridge Inspection Standards, and the outcome is one record per public highway bridge or culvert.

The scale is easy to underestimate. The 2025 edition counts 624,193 highway bridges across the 50 states, DC, Puerto Rico and the US Virgin Islands - Guam sat out the 2025 cycle - and Texas anchors it with 56,951 structures. Because annual editions run from 1992 through 2025, the archive holds 34 vintages, enough to watch any individual span age across three decades of inspections.

In Datadory's catalog the record scores 9/10 on our quality rubric, with the twenty-six-field core dictionary verified during the August 2026 research pass. It sits on the highways & railtracks data hub beside the rest of the publisher's estate, covered on the FHWA source profile. Get a sample of this dataset cut to your states and edition years before anything else.

One bridge, one line, 445 characters. Nothing about that shape has moved between editions, which is why the anatomy below - taken straight from the documented record format - is worth reading closely:

ONE RECORD = ONE PUBLIC HIGHWAY BRIDGE, POSITIONS FIXED
------------------------------------------------------
pos 001-003  ITEM 1    State Code
pos 004-018  ITEM 8    Structure Number      <- the join spine
pos 019-027  ITEM 5    Inventory Route
pos 038-062  ITEM 6    Features Intersected
pos 063-080  ITEM 7    Facility Carried
pos 130-137  ITEM 16   Latitude              pos 138-146  ITEM 17  Longitude
pos 157-160  ITEM 27   Year Built            pos 165-170  ITEM 29  Avg Daily Traffic
pos 259      ITEM 58   Deck condition        pos 260      ITEM 59  Superstructure
pos 261      ITEM 60   Substructure          pos 263      ITEM 62  Culverts
pos 265-267  ITEM 64   Operating Rating      pos 269-271  ITEM 66  Inventory Rating
pos 287-290  ITEM 90   Inspection Date       pos 291-292  ITEM 91  Frequency (months)
pos 375      ITEM 113  Scour Critical        pos 434      CAT10    Bridge Condition
pos 436-445  CAT29     Deck Area (sq m)

Decoded, one line reads as: a structure pinned by coordinates, carrying route X over feature Y since year Z, rated deck/superstructure/substructure/culvert on single-character inspection codes, rolled up by FHWA into the federal good/fair/poor class (CAT10), with operating and inventory loads in metric tons and a scour-vulnerability flag at the foundations. Values arrive with your sample, cut to the states and edition years you name - the anatomy is the constant.

Twenty-six fields make up the verified core, doing four jobs. Identity: the FIPS State Code and the ITEM 8 Structure Number, the agency-assigned key that tracks one bridge across all 34 editions. Place and route: inventory route, features intersected, facility carried, location text and decimal coordinates. Physical and service description: year built and reconstructed, average daily car and truck traffic, main structure type and length in meters. Condition, loads and inspection: the four single-character condition ratings, operating and inventory ratings in metric tons, inspection date and required frequency, scour criticality - plus FHWA's download-only federal roll-ups CAT10, CAT23 and CAT29.

Definitions were verified against the published record-format documentation during the August 2026 research pass; nothing here was inferred from headers. If you want the narrative tour first, see what fields the National Bridge Inventory includes.

Geography - all 50 states, DC, Puerto Rico and the US Virgin Islands. Guam did not submit 2025 data. Coverage is a census rather than a sample, so state-level counts reconcile without estimation: Texas leads with 56,951 structures in the current edition.

Temporal - 34 annual editions, 1992 through 2025, with 624,193 highway bridges in the 2025 vintage. Two coding breaks matter for trend work: STRUCTURALLY DEFICIENT was redefined starting with the 2018 data, and FUNCTIONALLY OBSOLETE was discontinued after 2016. Panels that cross those years simply need the break treated explicitly - the items involved are identified in the dictionary.

Granularity - one record per bridge or culvert structure, 445 characters wide in the fixed-width layout or delimited with identical fields. Load ratings deliver in metric tons regardless of any stated rating factor, and the combined all-states edition compresses to roughly 51 MB (about 56 MB including non-highway structures and routes carried under bridges).

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel that fits the workload: files or warehouse load for the historical backfill - thirty-four editions reward loading once - then a rotation cadence that suits the decision you are feeding. Deliveries keep the field dictionary above unchanged, preserve the fixed-width anatomy or flatten it to typed columns, and flag the 2018 and 2016 coding breaks so time-series work starts clean. Start smaller than a feed: get a sample scoped to your states and years, validate it against your actual question, then set the schedule.

Six jobs the record settles outright:

1. Bridge deterioration modeling - condition ratings and inspections spanning 1992 through 2025 for roughly 624,000 structures, joined across vintages by one structure number; see quant backtesting for the modeling frame.
2. Deficiency screening at network scale - sort the whole population into good, fair and poor with CAT10 in one pass, instead of reconciling fifty state spreadsheets.
3. Deferred-maintenance exposure - infrastructure investors read poor-condition concentration and load ratings across the geographies they hold.
4. Bridge-stock profiling - age, type, material and condition distributions feed construction-demand forecasts and infrastructure studies.
5. Logistics-corridor risk flags - scour-critical and deficient spans near competitor corridors fall out of a mechanical screen.
6. Citation-grade sourcing - a federal inventory with a published format dictionary stands up in articles, filings and testimony; see citation-grade research.

Data scientists and ML engineers lead the fit: parse the fixed-width records once and a 34-edition deterioration panel falls out; see data scientists working in highways & railtracks. Market researchers and consultants profile the national stock's age and condition with official figures; see market researchers. Journalists, academics and students cite an annual federal inventory whose provenance is built in; see journalists and academics. Investors and quants price deferred maintenance for infrastructure diligence; see investors and quants. Developers and builders round it out - ingestion is configuration, because the character windows never move.

Within highways-and-railtracks data, each neighboring record answers a different question. Highway Performance Monitoring System (HPMS) owns roads rather than structures - route-level traffic and pavement instead of bridge condition. FHWA Highway Statistics Series (1945-present) is the finance and mileage ledger, not an asset-level inventory. The USDOT Open Data Portal is the breadth play across many agencies' catalogs. Overture Maps Transportation Theme supplies global geometry but no condition codes - the head-to-head runs on the NBI vs Overture comparison.

If the question is 'what condition is America's bridge stock in, structure by structure, over time', this is the record; everything else on the shelf sits beside it. See where it lands on the best highways & railtracks datasets list.

Three things, stated plainly. First, inspection recency is a property of the data: routine inspections occur roughly every 24 months, which makes the inventory excellent for network-scale questions and explicitly unsuitable for route clearance decisions about any single crossing today.

Second, trend work needs the coding breaks handled: STRUCTURALLY DEFICIENT changed definition with the 2018 data and FUNCTIONALLY OBSOLETE ended after 2016, so multi-decade series should treat those years as structural breaks rather than real deterioration. Your sample flags the affected items.

Third, scope is yours to define: name the states and edition years, and the sample returns in exactly the schema shown above - fixed-width anatomy intact or flattened to typed columns - with the deeper items (improvement-cost estimates, non-highway structures) confirmed alongside it. The practical walk-through lives on National Bridge Inventory Download.