International and PlusAI pick NVIDIA DRIVE AGX Hyperion and Thor for Level 4 trucks

International Motors and PlusAI have settled on the computer their driverless trucks will run on. The pair are building factory-installed Level 4 trucks around NVIDIA's DRIVE AGX Hyperion platform, with the DRIVE AGX Thor system-on-chip running PlusAI's SuperDrive software, according to an Oct. 28, 2025 International announcement that Fleet Owner covered in its Trucking Tech Today roundup that November.

Heck argues the industry's biggest challenge is something less flashy than sensors or chips: validating safety in the real world. Nauto CEO Stefan Heck told Heavy Duty Trucking executive editor Jack Roberts in May 2026 that the industry's real bottleneck is proving driverless trucks safe in the rare, odd situations a human driver handles without thinking, and that doing so takes real-world data on the order of hundreds of billions of miles.

That leaves a fleet buyer in a strange spot. The hardware is defined and the manufacturing partners are named, yet the thing that decides when a truck can run without a driver is a validation record no vendor can ship in a box.

Three companies, one truck

Fleet Owner described the arrangement as a three-way split: International supplies the vehicle and the factory, PlusAI supplies the SuperDrive autonomous driving software, and NVIDIA supplies the AI compute. The trucks are to carry lidar, radar and cameras for 360-degree awareness, with sensor fusion and redundancy intended to cover long-haul highway environments.

Tobias Glitterstam, International's SVP and chief strategy and transformation officer, said in the announcement that the NVIDIA and PlusAI work builds on fleet trials in Texas and is a step on the company's path to production. Production is the word to notice. A truck that leaves the plant with autonomy designed in is a different purchase from a retrofit, though the material Fleet Owner cited gives no price or delivery date.

For a fleet that runs International tractors today, the practical implication is narrow but real: the autonomy stack being specified is tied to a chassis the shop already knows, and the compute inside it is a named NVIDIA part rather than a proprietary black box.

The last few percent, in Heck's telling

Heck's argument, as HDT reported it, is that the final sliver of driving competence is vastly harder to achieve than the first half. He pointed to aircraft and nuclear plants, which run at what he called seven nines of reliability, as the bar that safety-critical systems actually meet, against a public conversation that tends to stop at 99%.

Simulation cannot close that gap alone, Heck told HDT, because engineers can only simulate situations they have already imagined or observed. Heavy rain, flooded roads, construction zones, pedestrians and other drivers behaving unexpectedly are the cases that need actual miles, and HDT reported that these everyday conditions still create situations autonomous systems cannot yet handle safely.

Heck also expects the rollout to stay tightly controlled. HDT reported that early deployments will concentrate on corridors such as Interstate 10 through the Sun Belt, where weather, traffic and regulation are predictable. For a carrier whose freight runs those lanes, the technology could arrive well before it reaches an operator running northern routes in January; the corridor-first pattern HDT describes suggests a much longer wait for the latter.

200,000 vehicles against 12 million trucks

Heck told HDT he expects tens of thousands of self-driving vehicles, collectively perhaps 100,000 to 200,000, on the road within two years, helped by manufacturing partnerships such as Aurora's with Continental and Waymo's expansion onto Hyundai platforms. HDT set that projection against roughly 12 million trucks operating in the US.

Even at the top of Heck's range, autonomous systems will still represent only a small fraction of the market for years compared to the roughly 12 million trucks operating in the US, according to HDT. Heck himself cautioned against mistaking early growth for widespread adoption. HDT has separately reported Hirschbach's plan to deploy 500 Aurora autonomous trucks, which gives a sense of what one large early order looks like.

The sharper question a fleet can now put to any Level 4 vendor is which specific corridors it can document as validated, and with how many real miles behind them. A demo on the show floor answers a different question than that.

Grote's trailer warnings arrive for the fleet that still has drivers

The same Fleet Owner roundup covered two Grote Industries products shown at the ATA 2025 Management Conference & Exhibition. H.E.L.P. DeliverSAFE adds digital and lighting alerts to make a trailer more visible at the roadside. The 4SEE Rear-Collision Warning System uses radar to detect an approaching vehicle and strobe lamps to warn the motorist behind.

The detail worth a second look is who each warning is aimed at. Grote president and CEO Dominic Grote framed both products as tools that help protect drivers, according to Fleet Owner, yet the 4SEE radar-and-strobe alert is directed at following motorists. It works by changing the behavior of the car behind the trailer.

Both systems run on Grote's 4SEE Digital Harness, which Fleet Owner said handles secure data transfer, lamp-specific light-out detection and over-the-air updates. For a maintenance director, a trailer harness that reports exactly which lamp failed and accepts a software update is a different maintenance object from a wiring loom, and it belongs on the trailer spec now, whatever the timeline for the tractor pulling it.

J. J. Keller folds ELD and GPS into one box

Fleet Owner's third item was the VT100 Vehicle Telematics Device from J. J. Keller & Associates, which combines ELD compliance with GPS tracking in a single unit for Class 1 through Class 8 vehicles. One device instead of two is the pitch. For a mixed fleet running pickups alongside tractors, a single device family that spans every weight class is the operational detail that matters more than either function on its own.

The 2015 forecast and the 2026 timetable

In December 2015, EHS Today's Ginger Christ reported that Ted Scott, then director of engineering at the American Trucking Associations, regarded autonomous trucks as a matter of when rather than if, but called fully automated trucks ruling the roads a long way off, with public and government acceptance as the gate. Cozen O'Connor attorney Julia Molander, an adviser to insurers, expected driver-assist features to grow gradually into full autonomy, Christ reported.

A decade on, the shape has changed. HDT's account describes named manufacturing partnerships and six-figure deployment projections rather than organic creep. Yet HDT also notes that in the mid-2010s some industry observers predicted fully autonomous trucks would be commonplace within five years, and that the industry instead discovered achieving high reliability in difficult operating conditions is exponentially harder than handling straightforward highway. That is the reason the gate Scott described in 2015 is still the gate.

The US trucking industry has the potential to be an early adopter of autonomous vehicles, according to a paper in Transportation Research Record: Journal of the Transportation Research Board, because it hauls the majority of US freight by weight and trucking companies may be able to reduce costs through labor savings and increased utilization of trucks. The paper lists three preconditions: interest from technology developers, truck manufacturers and trucking companies; government support to allow testing and deployment on public roads; and acceptance from the public who will share the road. The International, PlusAI and NVIDIA arrangement is the first of those made visible. The hundreds of billions of miles Heck describes are what the other two will cost.

International has published no production start date in the material Fleet Owner cited. The Texas trials Glitterstam pointed to are the thing to watch, and the signal that would change the buyer's calculus is a published count of validated miles, corridor by corridor.