McLane Company, the Berkshire Hathaway subsidiary that distributes food to restaurants and retailers, ran a three-year pilot with Aurora Innovation that wrapped in 2023 after logging more than 280,000 autonomous miles and 1,400 loads with a 100% on-time delivery record. The pilot has since converted to fully driverless commercial runs between Dallas and Houston. By any measure, that’s a real product, not a demo.
It’s also a hybrid product, and the hybrid part rarely makes the headline. Aurora’s trucks haul the middle-mile distance between McLane’s distribution centers. Human drivers still handle the local deliveries on each end. Strip away the press release language and the actual claim is narrower than “driverless trucking”: a truck can now drive itself across Texas highways without anyone in the cab, and a person still has to get the freight the rest of the way.
That gap between the highway and everything around it defines where autonomous trucking actually stands. The highway portion is scaling fast, with real mileage, real customers, and real regulatory momentum behind it. The portion on either side of the highway, loading the truck and getting cargo to its final destination, has barely moved. Treating those as one story is why “driverless” coverage keeps overpromising, and untangling them is the only way to read the technology’s timeline accurately.
The Highway Side Is Scaling Quickly
Aurora currently runs roughly 109 trucks and plans to reach 200 by the end of 2026, with thousands projected within five years. A second route, built with Volvo Autonomous Solutions, now connects Dallas directly to customer facilities in Oklahoma City.
That route runs on Volvo VNL Autonomous trucks paired with the Aurora Driver system, operating under supervised autonomy five days a week while the partnership works through what the companies describe as the final validation phase before fully driverless runs. Volvo has committed to manufacturing hundreds of autonomous trucks starting in 2027, which says more about where the capital is headed than any press release could.
The regulatory side is moving in step. Aurora has asked the Federal Motor Carrier Safety Administration for a five-year exemption from the rule requiring drivers to manually place warning triangles around a stopped truck, arguing that a vehicle with no human in the cab has no one to place one.
The agency’s public comment period on that request closes May 15, 2026. Halfway around the world, Japan’s Ministry of Land, Infrastructure, Transport and Tourism is directing a separate effort with Yamato Transport, Mitsubishi Fuso Truck and Bus Corp, and the software firm Tier IV to put Level 4 autonomous semitrailers on long-haul routes by 2026, a government deadline rather than a corporate forecast.
What unites all four of these, McLane’s pilot, the Volvo route, the FMCSA exemption, and Japan’s MLIT project, is that every one of them operates in a constrained, predictable environment: fixed highway corridors between two known points, mapped in advance, with no requirement to interact with anything outside the road itself. That’s the version of autonomous trucking that’s ready. It’s also the only version most coverage of the industry bothers to describe.
The Part Nobody’s Automated Yet
A truck that arrives at a distribution center still needs something or someone to get the cargo off it, and that turns out to be a harder problem than driving in a straight line at highway speed. Udelv builds electric delivery vehicles with no driver cabin at all, converting that space into extra cargo capacity, and controls loading through an adaptable shelving system and a patented curtain mechanism that reveals only the packages meant for a given stop. It’s a genuine engineering answer to the loading problem, not a workaround dressed up as one.
It also reveals the limits of that answer. Udelv targets business-to-business routes first, specifically because professionals on the receiving end can load and unload the vehicle themselves.
The autonomy stops at the loading dock; someone on the other end still does the physical work, which is exactly the same arrangement McLane and Aurora settled on for the Dallas-Houston route. Strip the humans out of either end of that chain and the system stops functioning, no matter how many highway miles the truck in the middle can cover on its own.
Residential delivery removes even that workaround. A large truck can’t pull into most driveways, can’t navigate a typical residential street built for passenger cars, and can’t carry a package up to a front door. Door-to-door delivery requires a vehicle small enough to access tight spaces and some method for transferring goods directly into a customer’s hands, and neither of those is something a highway-grade autonomous truck was built to do.
Designing one vehicle that can handle both an interstate at 65 miles per hour and a cul-de-sac at walking speed is a different engineering problem than the one Aurora and Volvo have spent years solving, not a smaller version of the same problem.
Why the Industry Keeps Conflating the Two
Part of the confusion is structural. McLane’s own description of its operating model, Aurora running the middle-mile hauls while human drivers manage local last-mile deliveries, is the clearest summary available of how every one of these deployments actually works, yet it rarely survives the trip into a headline. “Fully driverless” gets applied to the truck on the highway, and the human standing at the loading dock or driving the final five miles quietly disappears from the story.
The same gap shows up outside freight trucking entirely. Coverage of robot car shipping, using autonomous technology to relocate vehicles between dealers, auctions, and buyers rather than pallets of cargo, has reached an identical conclusion.
Even citing the existing partnership between Daimler Trucks and Waymo to run SAE Level 4 autonomous trucks, that coverage found that humans are likely to remain part of the shipping and delivery process for years to come, for the same reasons trucking faces: a truck that can’t independently load or unload a vehicle, and can’t navigate a residential street to complete a door-to-door handoff, still needs a person somewhere in the chain.
COVID-19 accelerated investment in contactless shipping by turning minimal human contact into a selling point rather than a cost-saving measure, and that demand pulled money toward the highway-automation side of the industry, where the engineering problem was already closer to solved.
It didn’t pull equivalent investment into the loading-dock problem, because that problem was never about reducing contact between two people. It’s about replacing physical labor that machines still can’t reliably perform outside a handful of pilot programs like Udelv’s.
Where the Money Actually Goes
These are genuinely two different markets, and the data on each one explains why the highway side is years ahead. Freight automation, the Aurora and Volvo side of the business, draws capital because the return scales with miles: a truck running thousands of highway miles a month produces savings fast enough to justify the sensor packages, route mapping, and regulatory lobbying required to get it approved.
The market built to solve the other half of the problem, getting a package or vehicle the rest of the way to a door, tells a different story. The global autonomous last-mile delivery market, valued at roughly $1.3 billion in 2025, is projected to reach $11.5 billion by 2035, a more than tenfold increase. However, within that market, trucks and vans hold only a 9% share.
Delivery drones account for 49%, and ground-based robots hold another 42%, both platforms built specifically to navigate the tight, irregular spaces a truck can’t reach. Vehicle-scale automation is the smallest, slowest-growing piece of the one market explicitly designed to solve the last-mile problem, because the door-to-door obstacle that defeats a freight truck doesn’t disappear just because the truck gets smaller.
That split tells the real story better than any single headline could. Highway freight automation and last-mile delivery automation aren’t two stages of the same technology maturing on the same timeline. They’re two separate engineering problems with two separate capital markets, and the one with the faster, more predictable return is years ahead of the one that actually touches a customer’s door.
What Actually Changes Next
The next several months will mostly settle questions about the highway side. The FMCSA’s comment period on Aurora’s exemption request closes May 15, 2026. Japan’s MLIT-led project has its own 2026 target for Level 4 deployment. Volvo’s manufacturing commitment kicks in starting 2027. Each of these will say something real about how far highway automation can scale on infrastructure that already exists.
None of them will touch the loading dock or the last mile, because none of them were built to. Closing that gap requires investment aimed specifically at the physical handoff, robotic loading systems built for irregular cargo, vehicle designs that don’t assume a person is standing inside the trailer, or new platforms purpose-built for tight urban access instead of highway miles. \
Until that kind of investment reaches the scale of what’s currently funding highway freight, the pattern McLane and Aurora already settled on will keep repeating: a machine handles the easy, well-funded majority of the trip, and a person handles the hard, underfunded part that actually gets the cargo into someone’s hands.
Matthew Wilde
Matthew Wilde is an automotive journalist with experience contributing to leading publications. He focuses on delivering clear, well-researched analysis of automotive industry news and vehicles. Growing up surrounded by a variety of cars, Matthew developed a strong foundation in automotive technology and design. His work emphasizes accuracy and depth, aimed at informing both enthusiasts and industry professionals with straightforward, precise reporting.
https://theweeklydriver.com/