It wasn’t too many years ago when you couldn’t read an article about broadband infrastructure without hearing about the need for smart highway infrastructure that was going to enable self-driving cars. There were various versions of how this would happen, but the predominant concept was that 5G networks along roads would communicate with cars and would enable efficient and safe travel by eliminating driver error by taking the driver out of the equation. This was one of the primary business cases for 5G promoted by the cellular carriers.

The idea went quiet for a variety of reasons, and I thought this idea was dead. I was surprised to recently hear about a $130 million project in Michigan to create a trial project for smart roads. This project will be for a 25-mile stretch of I-94 between Ann Arbor and Detroit. The project is described as creating the world’s most advanced road network for connected and automated vehicles.

There are a number of reasons that the industry has migrated almost all research into developing smart cars rather than smart roads.

First is the basic discussion of whether computing brains should be provided by centralized infrastructure or moved to the edge -in the self-driving car arena, the edge is the smart car. Most of the effort by car manufacturers has been to make cars smarter. Google and other pioneers in the field decided that cars needed to be able to deal with all driving conditions and all roads rather than relying somehow on smart roads. This approach has taken a lot longer than first predicted, but cars are getting better at this every year as manufacturers introduce new features.

The best arguments against the smart road are practical. First is the classic chicken and egg issue. Do we really need to wait until there are smart roads everywhere (or at least in a large percentage of places) before the self-driving car industry makes any economic sense? Let’s say this corridor in the project works as promised. What benefits or incentives does this one stretch of road provide until there are hundreds of times more smart roads with the same features? Are any car manufacturers going to develop features that rely on smart roads until there are enough smart roads for this to make sense? Will people be willing to pay more for a car with smart road features if it can only be used on limited stretches of smart roads?

And then there is the cost issue. This project costs over $5 million per mile of roadway. I assume that construction costs will drop if the technology is expanded. Let’s assume this might cost $4 million per mile. There are over 430,000 miles of major roads in the country, including interstate highways and other major divided arterial highways. It would cost over $1.7 trillion to bring this technology to just those major highways. Who is going to pay for that? There are 276 million cars in the country, and that equates to almost $6,300 per vehicle. I’ve thought about this several times over the last decade, and I can’t envision this being a priority compared to how $1.7 billion could be used elsewhere. Worse yet, that huge cost only brings the solution to major highways. Is that enough miles for this to be worth it? How do driverless trucks navigate when they have to get off the major highways? There are over 2 additional million miles of paved roadways in the country.

But just suppose the country decides that smart roads are an important national priority, and we spend the trillions to get a smart highway system that revolutionizes product delivery nationwide. There are some obvious advantages to that vision of having driverless vehicles shuttling goods across the country 24/7, eliminating truckers. What about the cost of maintaining this smart road network? This network will consist of sensors along and embedded in roads and 5G and other wireless technologies to communicate with vehicles. We know that many of these technologies only have an effective life of around ten years. That means a perpetual huge annual budget to keep the electronics network upgraded and functioning. There also would be a lot of people needed to keep the smart roads operating properly. I can see why companies like Cavnue and the cellular carriers love the idea of the smart road—it’s a perpetual, guaranteed revenue stream for them. But I ask again about who would be willing to pay for this huge ongoing expense?

When I first heard about smart roads, my first thought was to ask what happens when the smart road networks break down, as they inevitably will do. What happens on a smart road during the time when the road isn’t smart? Does traffic stop or slow to a crawl? I really thought this was an idea that had been put to bed. But I guess the lure of making money from building, operating, and updating a new kind of infrastructure is just too lucrative to let die—even if it is impractical.

The project is being described as a public-private partnership, and I assume that means that public grants are helping to fund this, perhaps out of ARPA money. I’ve been doing a lot of work in Michigan, and I know how far $130 million would go to bring broadband to unserved homes—probably including some near this smart road. I am never opposed to projects that push innovation, and if this is intended as a test bed to explore a lot of new ideas I would think this is a worthwhile idea. But Cavnue is touting this as the first of many such projects that will be replicated across Michigan and the rest of the country. I honestly don’t get it, and I invite anybody to tell me why this is a good idea.