I was in West Texas recently to witness firsthand the emerging practical applications of artificial intelligence. What I saw bolstered my conviction about the technology’s progress and the need to mold it rather than resist the change.
At the intersection of one of those dusty, gravel-covered roads that pepper the Permian Basin oil fields, a truck that would not be out of place on any American highway made a wide turn to account for the two trailers it was hauling. That’s a normal sight in these parts, but what wasn’t normal was that nobody was in the cab.
Here, Atlas Energy Solutions Inc. has teamed up with Kodiak AI Inc. to become one of the first to commercially deploy driverless trucks, which are being integrated into the steady flow of haulers that carry sand for use in hydraulic fracking to well sites along private bumpy roads. In West Texas and southeastern New Mexico, it can be difficult to recruit drivers. They often stay in barracks-type housing called man camps and work two weeks on and a week off to return home, which is typically in another state.
Driverless trucks are graduating this year from a long testing phase and becoming one of the first physical AI tools to break out of the confines of a factory or warehouse and interact with the public. Robotaxis got a head start and more of the headlines, but freight’s larger market size and more predictable routes could make the autonomous trucks roll out faster and with a larger impact on the economy than driverless ride-sharing. A Goldman Sachs report late last month forecast that the market for autonomous trucking in the US will reach $105 billion in 2035, more than double the $48 billion projected for robotaxis.
The business case for driverless trucks is clear. The all-in cost per mile for an autonomous truck in the US is projected to drop from about $8.56 in 2025 to $2.03 in 2035, according to the Goldman report. Meanwhile, the comparable cost for a truck with a driver is expected to rise from $2.55 to $2.84 over the same period, driven partly by rising wages. Driverless trucks also have a key advantage: They are not subject to the hours-of-service limits that restrict humans to 11 hours of driving each day in the US.
The opposition comes mainly from two camps. The first is those who worry about the safety of massive vehicles operating without any human guidance — though driverless trucks would most likely improve safety on US highways. These trucks have more situational awareness than humans, and the software that guides them is trained with live operations using a safety driver and with sophisticated simulation that can test even the most unusual safety challenges. That means they could reduce accidents caused by all motorists, not only truck drivers.
The other camp comprises those who are concerned about sweeping job losses, like the Teamsters union.
These arguments against technological advances haven’t changed much since the Industrial Revolution: The machines are unsafe and will throw countless out of work. Part of the problem is that it’s much easier to identify the jobs that a machine can replace than it is to predict the future jobs that will be created by technology. AI-enabled automation promises to increase productivity and raise the overall standard of living, and if US companies don’t lead the charge, they will cede the advantage to Chinese ones, which are pursuing the same technology full bore with the enthusiastic support of President Xi Jinping.
Although the Trump administration has resisted shackling AI, autonomous trucking may be the first use case that forces Congress to lay down some rules. That’s because freight crosses state lines and can’t be governed by a patchwork of state regulations forever. Enough states, mostly in a wide swath across the South, have approved driverless regulations to get the industry started. Even California recently approved a pathway — although a bit onerous — for obtaining permission for autonomous trucks.
Bills are winding their way through the House and Senate, but it will take time to gain the momentum needed for passage. It’s important the legislation support the technology instead of allowing special interests to put up roadblocks. These autonomous-vehicle companies will be successful only if they prove much safer than human drivers and turn a profit.
The starting point for that is in the oil fields of West Texas.
When fracking first started, the required sand was being shipped by rail from as far away as Wisconsin to obtain the shape and size of grains that producers wanted. Bud Brigham, a native of Midland, Texas, who made a fortune in the early fracking boom, set about searching for that sand among the dunes that he used to slide down on cardboard as a kid. He found the right grains near Kermit, Texas. He started Atlas in 2017 to develop a sand mine in the desert, and it has become perhaps one of the unlikeliest proving grounds for the rollout of physical AI.
Kodiak is operating 28 driverless trucks to haul sand, and the startup, which began trading publicly in September, is under contract to deliver an additional 72 to Atlas. These trucks use cameras, radar and lidar — which bounces light off objects — to gather information from all directions and farther out than humans can perceive. An onboard computer crunches the data to guide the truck.
The trucks pick up the sand at the end of a $400 million, 42-mile conveyor belt — aptly named the Dune Express — that Atlas built to ship sand from the mine to a staging area deep in the oil patch, allowing the company to bypass several crowded public roads.
Although many people think of chatbots when it comes to artificial intelligence, the conveyor belt, like its cousin the driverless truck, has the components of physical AI. Sand whirls along on a belt with almost 69,000 rollers that have sensors that monitor speed, vibration and other variables and flag the need for preventative maintenance. Another sensor detects any tear of steel cables in the belt. A slew of cameras and fiber optic cable allow the conveyor to be monitored remotely. The system has acoustic sensing that uses machine learning to identify whether an intruder or disruptive wildlife has approached the system, and it uses facial recognition and reads license plates to tell workers from trespassers.
At the mine, there are 10 massive silos where about 1,000 trucks a day used to load up with sand and haul it on public highways to reach those private roads that connect to the well sites. Now, just a handful of trucks show up at the expansive, paved staging area to pick up sand. The silos now feed the Dune Express.
Being able to avoid those public highways let Kodiak reduce the risks of having to compete with other motorists at high speeds. Ahead of Kodiak’s plan to begin driverless service on highways at the end of this year, the company has gained experience using its trucks in a 24/7 logistics operation along with the conventional ones.
The challenge for Kodiak is to reduce the number of workers it needs on site to shepherd the driverless operation and eventually hand over most of the work, including maintaining the trucks, to Atlas. That effort includes validating the safety case for loading sand on the trucks while in driverless mode at the end-of-line silos.
Like other driverless freight startups, Kodiak is in a race to deploy its autonomous trucks to earn revenue while reducing the costs of both the hardware and the need to babysit the system. A sign that this automation technology will take hold is the military’s keen interest. Kodiak has partnered with General Dynamics Corp. to build an autonomous counter-drone vehicle based on a large Ford commercial truck. Don Burnette, the chief executive officer of Kodiak, has discussed how its driverless software will be able to power other physical AI, including robots.
The physical AI revolution is clearly taking its first significant strides. Soon it will expand beyond the dusty backroads of the Permian Basin and have a greater presence in the general public. It should be welcomed with greater support.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Thomas Black
