Take one row of one federal table, hold everything constant except location, and the price of building a lane of Interstate on new alignment runs from $3.55 million per lane mile to $85.85 million. Same agency, same line item, same dollar year. The low figure is rural flat terrain. The high figure is a major urbanized area where conventional widening will not work. Twenty-four times, inside a single published estimate.
That table is Exhibit A-1 of Appendix A to the 23rd edition of FHWA’s Conditions and Performance report to Congress, and the spread in it is the whole answer to the question in the headline. There is no cost per mile of Interstate. There is a cost per lane mile, for a scope, in a terrain, in a market, in a year, and every one of those five qualifiers moves the number more than the choice of pavement material does.
What the federal numbers are, and what they are for
Exhibit A-1 exists to feed a model, not to price a project. The Highway Economic Requirements System evaluates the current state of the network from Highway Performance Monitoring System data, forecasts conditions across four consecutive five-year periods, and tests candidate improvements against deficiencies in eight section characteristics. To do that it needs a unit cost for each improvement type, and the exhibit supplies one.
The provenance matters because it bounds how much weight the figures can carry. FHWA restructured the improvement cost matrix for the 2004 edition using project data from six states, disaggregating urban values by functional class and by three population groupings. The 2006 edition added data for large urbanized areas, rural mountainous regions and high-cost capacity improvements, and introduced a category for urbanized areas above one million people. The 2008 edition adjusted the matrix again. The values published in the 23rd edition, which was submitted to Congress in November 2019, are stated in thousands of 2014 dollars.
FHWA then attaches its own warning, and it is worth quoting rather than paraphrasing because it disposes of most cost-per-mile journalism in one clause: “These costs are intended to reflect the typical values for these types of projects in 2014, and thus do not reflect the large variation in cost among projects of the same type, even in a given year.” The report names the drivers of that variation. A widening estimate assumes a section carries a number of bridges typical for its length, and some sections carry more. Complicated interchanges, major environmental issues and extreme engineering issues push costs to the unusual end.
The exhibit also splits capacity work into two cost regimes, which is the most useful structural feature in it. Values for adding a lane at “normal cost” apply where sufficient right-of-way is available or readily obtained. Values for adding lane equivalents at “high cost” cover situations in which conventional widening is not feasible.
Lane miles are not miles
The unit trap catches almost every casual comparison. Exhibit A-1 is denominated per lane mile, and Interstate mainline rarely has fewer than four lanes.
Multiply the rural flat new-alignment figure by four and a four-lane rural Interstate on level ground works out to roughly $14.2 million per centerline mile in 2014 dollars, before interchanges, before structures, before the right-of-way that a new alignment necessarily requires. That arithmetic is a derivation from FHWA’s table rather than a figure FHWA publishes, and it should be treated as an order of magnitude rather than an estimate. What it establishes is the size of the unit error: quoting the per-lane-mile number as a per-mile number understates a four-lane facility by a factor of four.
Scale gives the same point from the other direction. FHWA’s public road length table for 2023 credits the Interstate system with 29,109 rural miles and 19,306 urban ones, 48,415 altogether, inside a national network of 4,199,209 miles. Those are centerline figures. The lane-mile count behind them is several times larger, and it is lane miles that get paved.
Terrain multiplies harder than road class does
Within rural Interstate, the terrain columns move the number more than moving down two functional classes does. Reconstructing and widening a lane runs $1.993 million on flat terrain, $2.234 million on rolling, and $4.235 million in mountainous country. Adding a lane runs $2.561 million flat, $2.777 million rolling, and $8.646 million mountainous, so mountainous capacity is roughly 3.4 times flat capacity for the identical scope. By comparison, dropping from rural Interstate to rural minor arterial on flat terrain takes reconstruct-and-widen from $1.993 million to $1.423 million, a difference of about 29 percent.
The engineering behind that asymmetry is earthwork and structure, not surfacing. Cut and fill quantities scale with the ground, retaining walls appear where the cross-section will not fit, and drainage gets harder. That relationship is developed in earthwork and grading.
Resurfacing sits in a different order of magnitude entirely. An existing rural flat Interstate lane resurfaces for $462,000 per lane mile in the same table, against $1.302 million to reconstruct it. Reconstruction costs about 2.8 times a resurfacing, which is the arithmetic that makes preservation timing a financial question rather than a maintenance preference.
Urban cost is not paving
The urban rows are where the table stops behaving like an engineering estimate and starts describing a real estate market. Adding a lane equivalent at high cost in a major urbanized area is $64.219 million per lane mile against $2.561 million for a normal-cost rural flat lane, a ratio of about 25. New alignment at high cost in the same major urbanized category reaches $85.845 million per lane mile.
Nothing about asphalt or concrete changes across that range. What changes is what the lane displaces: buildings, businesses, utilities, existing structures, and the traffic that has to keep moving through the work. The same table prices a normal-cost added lane in a major urbanized area at $15.4 million, so even within one population category the normal-to-high ratio is better than four to one. When a project’s cost per mile looks implausible, the first question is which of those two regimes it sits in.
What the Interstate itself actually cost, by category
The one place the federal government did assemble a complete cost breakdown for Interstate construction is the series of Interstate Cost Estimates submitted to Congress between 1958 and 1991. Fifteen legislatively mandated estimates were prepared over that period, and the final one is the closest thing to an audited total for the system.
The 1991 estimate put the cost at $128.9 billion, of which the federal share was $114.3 billion. Within the $124.256 billion subtotal that excludes legislated takedowns and contingencies, preliminary engineering accounted for 4.5 percent, right of way 13.1 percent, and construction 82.4 percent.
Those proportions are the corrective to two opposite errors. Right of way was not a rounding error, and it was not the dominant line either. On the system as built, across flat farmland and dense cities together, land was about an eighth of the bill. In a specific built-up corridor it can invert entirely, which is precisely what Exhibit A-1’s high-cost urban columns encode.
The same series recorded why the estimates kept rising, and the distribution is the single most useful thing in the record. Of the increases FHWA attributed to identified causes, unit price changes accounted for 52.2 percent. Right-of-way increases were 10.2 percent, roadway design changes 8.5 percent, bridge design changes 5.9 percent, increases on added-lane projects 4.1 percent, and construction cost overruns 3.3 percent.
Of every cause of increase FHWA named, contract overruns were the smallest. The Interstate did not end up costing three and a half times its first estimate because contractors missed their bids. It cost that because prices moved underneath a program that ran for decades and because the specification kept changing. The 1981 revision makes the second half of that visible: the estimate fell from $133.932 billion to $119.000 billion when the Highway Act of 1981 redefined what completion meant. Scope is a variable, and a cost figure that does not state its scope is not comparable to anything.
The price index exists, and it is severe
Any cost figure older than a couple of years needs repricing, and FHWA maintains the instrument for it. The National Highway Construction Cost Index is a chained Fisher Ideal index, the formula Irving Fisher proposed in 1922, applied to bid quantity and estimated bid price data at the individual cost item level from a commercial bid-tabulations database. FHWA’s methodology paper is explicit about why that formula: the Fisher index takes weights from both the base and the current period, so it accommodates substitution between cost items, and chaining adjacent periods avoids rewriting the whole series every time a base year is moved.
The published quarterly series starts at 1.000 in the first quarter of 2003. In the fourth quarter of 2025 it stood at 3.234. Highway construction prices are about 223 percent above the 2003 base. No dollar figure in more on highway design means anything without the year it was quoted in and a run through that index.
The recent path is steeper than the long average. The index was 1.860 in the fourth quarter of 2020 and 3.116 three years later, an increase of 67.5 percent. Its largest single quarterly jump in the whole series is the second quarter of 2022, up 11.9 percent in three months. It also falls, which is the part that gets forgotten: the index reached 1.618 in the third quarter of 2006 and had dropped to 1.442 by the first quarter of 2010, about 11 percent lower. Its high point so far is 3.363 in the third quarter of 2024, slightly above where it sat at the end of 2025.
Repricing Exhibit A-1 is then arithmetic. The index averaged 1.682 across the four quarters of 2014 and stood at 3.234 in the fourth quarter of 2025, a ratio of 1.92. On that basis the rural flat new-alignment lane mile moves from $3.55 million to roughly $6.8 million, the mountainous figure from $10.12 million to roughly $19.5 million, and the four-lane rural centerline mile from about $14.2 million to about $27 million.
That calculation is a derivation, not a published FHWA figure, and the gap between the two does not stop there. More substantively, the index is built from construction bid items, so it prices concrete, asphalt, grading and structures. It does not price right-of-way acquisition, preliminary engineering, environmental mitigation or utility relocation, which were 17.6 percent of the Interstate’s own cost. Applying a construction index to a total that includes those categories under-corrects them.
How to get a real number for a real corridor
The published national values are for modeling. For an actual road, four documents will answer the question, and none of them requires a subscription.
Start with the state’s unit prices. Most state transportation departments publish tabulations of awarded bid items, and the reports are more granular than any national average. Florida’s statewide item average unit cost report covers a rolling twelve-month window, currently July 1, 2025 through June 30, 2026, and lists for each pay item the number of contracts it appeared in, the weighted average unit price, the total dollars and the total quantity. In that window, Superpave asphaltic concrete for Traffic C with a PG76-22 binder averaged $149.93 per ton across 80 contracts and 394,334 tons. Regular excavation averaged $7.11 per cubic yard across 146 contracts and 4.3 million cubic yards. Maintenance of traffic averaged $1,055.37 per day across 285 contracts, totaling $115.8 million, which is a useful reminder that keeping traffic moving through the work is a major pay item in its own right and not overhead.
Then read the contract count before trusting the average, because thin samples produce nonsense that looks authoritative. The same Florida report prices plain cement concrete pavement 8 inches thick at $264.68 per square yard, drawn from 3 contracts covering 435 square yards, while the 11-inch item averages $129.00 from a single contract covering 21,836 square yards. Thicker pavement at half the unit price is not a discovery about concrete. It is what happens when an average rests on three small quantities. Any per-mile figure assembled from pay items inherits every sample problem in the items it uses.
Third, find the corridor in the programming documents. Under 23 CFR 450.218 a state’s transportation improvement program covers at least four years and must be fiscally constrained, and each project listing carries an estimated total cost or cost range that may extend beyond the program’s four years. That is the agency’s own number for that project, attached to a year of money. Fourth, read the project’s own estimate. A project developed through the process described in how highway construction projects are planned generates a documented estimate at several stages, and the difference between the planning estimate and the engineer’s estimate at letting is often the most informative number available.
A cost per mile is a statement about a place, a scope, a unit and a year. Every comparison that holds up names those four before it names a dollar figure. Detached from all of them the number is not a fact about highways, and the people who quote it that way are usually arguing about something else.