Autonomous Vehicle Part III: The Race to Retrofit the Wheel

Moving away from the material transportation uses case discussed in Part II of our series, it’s time to take an even closer look at unregulated spaces where AV technology might find a winner: industrial machines and vehicles already in operation. 

Since customers will already own these machines (with plenty of life left in them), reinventing the wheel is not the answer. Retrofitting these machines with AV technology, however, does make sense on an economic and engineering level. 

If we’re looking for unregulated terrain that AVs can thrive in, we need not look any further than object manipulation, industrial machines, and a few key parameters:

Understanding industrial machines and their operators from an investment perspective

First, we look at the number of vehicles sold per annum. This serves as a high-level proxy to the number of machine operators and the corresponding market size for automation, which is largely driven by the amount of labor saved. 

Next, we look at the vehicle utilization rate to estimate the labor cost paid in association with the vehicle operations. It would not make sense to automate a vehicle that is used for only a couple of weeks or months in a year. 

Ideally, we will want to automate vehicles that are used and operated by humans on a daily basis. 

Finally, we examine the workflow to understand whether such vehicle operators need to perform adjacent tasks. We would prefer use cases in which the operator is solely dedicated to operating the vehicles. 

Construction and agriculture sectors dominate the mobile manipulation of autonomous vehicles. 

It is clear that the majority of the industrial vehicle market centers around agriculture and construction, making up approximately two-thirds of the market. Each machine is estimated to command $200-300B in labor value of vehicle operators.

Figure 1: US Industrial Machinery Market Size, by Application, 2013-2024 (USD Billion)

Harvesting is hard but worthwhile.

Agricultural vehicles account for the largest portion of the industrial vehicle market, with tractors, harvesters, and planting equipment alone forming more than half of the total market values. 

Figure 2: Asia Pacific agriculture equipment market size, by product 2014-2025 (USD Billion)

Looking at the top five products, we surmised the market size of the autonomous vehicle version of each product by estimating the labor cost associated with operating each type of vehicle, assuming a 1:1 vehicle-to-operator ratio, $11-13 per hour in labor cost, and five useful years, accounting for different utilization rate based on the workflow. 

Top 5 VehicleMarket Size by Unit SoldMarket Size by ValueAvg Selling PriceVehicle Operator Market Size 
Planting Equipment14MM$15.82B$5500 $46B
Tractor 7.8MM$66.5B$42,000$126B
Harvester2.3MM $28.8B$62,500$18.5B
Spraying Equipment1MM$2.2B $10250$3.4B
Irrigation and Crop Processing Equipment0.29MM$1.12B$24500$2.25B 

Unsurprisingly, tractors came out on top in terms of market size and potential. Their high unit price and high utilization rate make them a top target for creating an autonomous counterpart. 

From an early stage investment perspective, however, the space is highly competitive and is relatively matured, as evident by the recent acquisition of Bear Flag robotics by John Deere.

While also large in total market size, planting equipment tends to come in various shapes and sizes. The category is fragmented, and the machines tend to be versatile, making it difficult to automate around a few, large use-cases. 

Spraying is an interesting use-case: the automation benefit is not likely to result in substantial cost-saving, but rather saving in chemical nutrients. This may be highly beneficial for crops that see adverse yield from decades of fertilizer usage, especially if an automated solution can handle spot spraying. There is, however, a headwind around a declining cost of chemicals in recent years to be aware of. 

Figure 3: The Sum of Fertilizer, Seed, and Pesticide Cost in Central Illinois and the Heartland

Among these different use cases, harvesting remains the holy grail. 

The use case remains extremely technically challenging. The winner will likely have to create and commercialize a breakthrough gripping technology, much like one would need in the sorting and packaging use case within the eCommerce warehouse and distribution center

This technological barrier is ideally suited for early stage investment since the space is unlikely to be plagued by abundant competitors, as seen in the current autonomous tractor space. The key lies in the ability to discern the true capabilities of such novel gipping technology. 

Construction loader and concrete finishing are two perfect niches. 

The market for construction vehicles is just as large as agriculture vehicles, with the top five construction vehicle markets totalling more than $150 billion. Using similar methodology, we surmise the market size of the construction vehicle operator is materially larger than that of the agriculture vehicle operator.  

Top 5 Construction VehicleMarket Size by Unit SoldMarket Size by ValueAvg Selling PriceVehicle Operator Market Size 
Dump Truck1.8MM$44.3B$100 -150K$81B
Concrete Mixer1.1MM$13B$20 – 100K$42B
Excavator735K$44.12B$100 – 500K$33B
Loader988K$16.8B$40 – 130K$32B
Crane267K$48B$300K – 1.5MM$20B

Table 2: Top 5 construction vehicle by market size

The dump truck segment, despite being the largest in terms of market size, is unattractive to automate. Dump trucks traverse in and out of job sites, crossing into regulated roads. The vehicle also has to intricately dock to various kind of dumpsters, which is a very unconstrained and challenging technical problem to solve. 

The concrete mixer spans anywhere from a small piece of mixing equipment to a large mixer truck. The mixer itself does not appear to be an interesting automation opportunity due to the need for manual handling of concrete and concrete finishing. The process by which a concrete mason drives a Zamboni-like concrete finishing machine to smoothen the floor, however, appears to be much more attractive. While the market is much smaller—barely over $1 billion—there are opportunities to branch out and automate other concrete construction processes. 

Construction loaders are another attractive investment opportunity hidden in plain sight. Loaders transport materials (such as soil, rock, sand, and other demolition debris) from one area to another. From an engineering perspective, the crude nature of the construction process makes automating loaders compelling. 

An excavator, on the other hand, while looking very similar to a loader, is an unattractive vehicle to automate. Excavators are used to dig up soil and rocks from the earth in large-scale projects, however, an excavator operator needs to dig according to the marking left by a surveyor. These markings contain spatial relationship information, such as the pitch, depth, and parameter, and are done manually with no standard. Automating an excavator means that one will also need to create a computer vision technology sophisticated enough to interpret these readings, which is a set of problems unto itself. 

Finally, while the crane operator is expensive to hire, a crane is not a particularly attractive piece of equipment to automate due to the cost of failure. A malfunctioning crane might swing around the job site, damaging the building—or worse, severely injure workers. Given the heightened balance of safety and efficiency in the construction industry, it is unlikely that contractors are going to be willing to take on such risk. 

Dynamic, Dirt, and Damage are the Showstopper.

Autonomous industrial vehicles have been around for at least half a decade, but we have yet to see an inroad and widespread adoption of these technologies in the field. What exactly is preventing the adoption? 

Both agriculture and construction workplaces are extremely dynamic and filled with dozens—if not hundreds—of laborers, each doing their own tasks in a loosely prescriptive way, constricting autonomous vehicle movements. Additionally, dust, dirt, and other airborne contaminants are a major challenge. They can obscure sensors and damage componentry. 

But perhaps chief among these is the safety concern and the damage that might occur. 

The average autonomous car weighs around four tons. That’s enough to seriously hurt someone, even at low speeds. A bulldozer or excavator may weigh as much as ten times that figure. Any stray movement — even one that is just slightly off — could easily lead to damage and serious injury.

The cost of failure, both in terms of worker safety and the economic impact on the farm yield and construction project, has made farmers and general contractors extremely wary of widespread deployment of new technologies. 

The safety concern is real, and all vehicles need a communication protocol to communicate among themselves in order to be deployed seamlessly. Because of this, one of the most lucrative investment opportunities lies in solving vehicle-to-vehicle communication problems, which we will explore in Part IV of the unregulated autonomous vehicle series. 

The right answer is anything but obvious

While many of us can imagine the ultimate use case of AV —from robotaxis to drone delivery—the early opportunities are unlikely to be obvious. The most promising opportunities are more likely to be found within the unregulated terrains that allow for freedom to operate without mass regulatory hurdles and risks. 

Moreover, these terrains should hold semi-constrained environments, whereby successful technologies are likely to develop a strong technology moat. These environments are often harsh, unsafe, and undesirable to work, leading to labor shortages and demand drivers. 

To be cash efficient, companies may target existing vehicles with high utilization and required skilled labor and retrofit them into autonomous ones. 

Whatever the winner looks like, they are likely to be anything but obvious. But a careful eye and immense understanding is what will get us to spot one early and seize the opportunity. 

Leave a Reply