The News
PepsiCo has signed a multi-year agreement to scale fully driverless middle-mile freight with Gatik, putting roughly 41 driver-out box trucks into daily operation across Arizona, Texas, and Arkansas, the largest commercial driverless freight deployment in the US to date. The SAE Level 4 Isuzu medium-duty trucks haul Frito-Lay snacks and soft drinks on plant-to-DC and DC-to-store lanes, per the joint Gatik/PepsiCo release. PepsiCo ran the lanes with safety drivers from 2022 and went driver-out in mid-2025; the companies report 98%-plus on-time delivery.
For an operator, the move is simple to read. PepsiCo took its most repeatable, hardest-to-staff short-haul lanes and turned them into fixed, contracted capacity that runs nearly around the clock, and it did it without buying a single truck.
Know More
Gatik owns and operates the trucks; PepsiCo pays only for the throughput. That keeps it off PepsiCo's books as an operating expense instead of a capital purchase of a fleet that loses value. PepsiCo ran the lanes with a safety driver from 2022 and took the driver out in mid-2025, the same path every Gatik customer follows.
Gatik has $600 million in committed, multi-year, noncancelable revenue, more than 60,000 driverless orders run incident-free, and a fleet scaling from 10 to 60 trucks toward "hundreds" by year-end, per its January milestone. That book roughly doubled in January on a single consumer-goods contract, Bloomberg and Axios reported.
PepsiCo is the only Fortune-50 consumer-goods company in Gatik's disclosed customer list, which makes this the likely deal. PepsiCo's is the largest deployment running today; Loblaw's 50-truck Canadian agreement, signed in 2025, is the largest still planned.
Step Back
PepsiCo is one instance of a pattern worth naming: the repeatable middle mile. One vendor, Gatik, has now put driver-out trucks inside the short-haul lanes of half the Fortune 50, all the same lane shape, fixed, high-frequency, low-variance, and hard to staff:
Walmart ran the first driver-out middle-mile route in Bentonville in 2021.
Tyson Foods put refrigerated AV box trucks on northwest Arkansas lanes in 2023.
Kroger moved freight from an automated Dallas facility the same year.
Loblaw signed a 50-truck Toronto deal in 2025 and took an equity stake in the vendor.
Georgia-Pacific and Pitney Bowes round out the book, and now PepsiCo.
McLane, the Berkshire-owned distributor, automated freight the same quarter on the opposite bet. It put Aurora on long-haul I-45 between Dallas and Houston, a deployment still ramping toward commercial scale. Both target the same driver shortage, at opposite ends of the network. Gatik's middle-mile trucks already run empty at commercial scale. Aurora's long-haul trucks are still in planning.
Gowtham's notes
Automation arrives one lane at a time, and the first lanes to go are the structured, repeatable ones. PepsiCo's driverless middle mile is the most advanced example, but not the only one. A year ago I sat with EASE Logistics, a carrier running human-guided truck platoons on I-70: a lead driver, an autonomous follower with a safety rider, more than 100 deliveries across state lines with no incidents. EASE keeps two drivers in the convoy; Gatik takes the driver out entirely. Both started in the same place, a fixed and familiar corridor. EASE's first test for the pilot was one question: did the fleet run structured, repeatable lanes?
What makes a lane automate first is repeatability. The lanes that go first share four traits:
Fixed and repeated — the same route run thousands of times, until the truck "learns as it goes," in Jim Farrell's words.
Short-haul middle mile — plant-to-DC and DC-to-store legs.
High-frequency, low-variance — predictable volume and timing.
Hard to staff — the lanes with the worst driver retention.
So the question for an operator is simple: how many lanes like this do I already run? For most networks it is a short list. Those are the ones to automate first.
Room for Disagreement
The economics are still unproven. Gatik's $600 million is contracted revenue, not profit, and its fleet is still scaling from 60 trucks toward "hundreds." Those are signed commitments. Whether the model makes money at scale is the thing to watch.
PepsiCo is not doing this to cut costs now. EASE found the same thing in its own pilot: the numbers did not add up at first, and the payoff was strategic: experience, credibility with regulators, and a say in the rules. "We're not making money on this pilot. But we're building capability and trust that will matter later," EASE's Justin Arciaga told us. PepsiCo is further along, already driver-out at scale, but the goal is the same: secure capacity on its hardest-to-staff lanes and get ahead of what is coming.
Two other things could keep it from spreading:
Regulation. The trucks run in a few permissive states. One bad public-road crash could reset the rules across all of them.
Labor. The Teamsters came out against the deal directly. "Driverless trucks are a clear and present danger to good-paying jobs and the safety and infrastructure of our local communities," the union said, arguing PepsiCo's delivery drivers rely on union wages that Gatik is "trying to destroy."
Today this works on a narrow set of lanes in a few states. Whether it spreads depends on the economics holding up as the fleet grows.
Notable
Loblaw did not just buy from Gatik; it took an equity stake, a customer becoming an owner.
The trucks are Isuzu medium-duty box trucks, not AV-native vehicles; routes run up to 400 miles; and the permissive-regulation states, Arizona, Texas, and Arkansas, are doing the selecting.
Dig deeper:
How EASE Logistics deployed automated truck platooning on I-70 — The Conveyor [Explainer] — our own field report on a carrier running the same "structured, repeatable lane" playbook, with a 10-step pilot framework you can steal.
SAE J3016 User Guide — Philip Koopman, Carnegie Mellon [Explainer] — the definitive, vendor-neutral primer on what "Level 4," ODD, and the dynamic-driving-task fallback actually mean.
