Freight/Signal Subscribe
Issue 5 · AI in Freight ·

Driver-out went live this month — here's the operator map

Two fully driverless freight deployments crossed the line in two weeks. The question isn't whether it's real — it's which of your lanes it touches, and what to do this quarter.

For a decade, the honest version of “self-driving trucks” carried an asterisk: there was still a human in the seat. The truck steered, braked, and held its lane, but a safety driver sat behind the wheel, hands hovering, ready to take over. The asterisk was the whole story — it meant the technology wasn’t trusted to run a load alone, and it meant none of the labor math actually changed.

In the last two weeks, two companies removed the asterisk on real freight.

On June 9, PepsiCo and Gatik announced a multi-year expansion of fully driverless operations — no safety driver, no in-cab observer — running medium-duty box trucks on short, repeatable runs to roughly 250 Walmart and Dollar General stores across Texas, Arizona, and Arkansas. Gatik puts on-time performance at 99%. The same month, Aurora tripled its driverless network to ten routes and opened a roughly 1,000-mile Fort Worth–Phoenix lane — the first commercial driverless route long enough that the truck legally outruns what a single human driver is allowed to do in a day.

Gatik @Gatik_AI · Jun 8, 2026

@PepsiCo and Gatik are bringing driverless trucks to North America at scale. Today, we announced a multi-year partnership with @PepsiCo to deploy driverless trucks across the company’s supply chain. Gatik is already operating over 40 driverless trucks for the food and beverage giant.

View the post on X →

That’s the line being crossed. Not “a truck drove itself in a demo.” Freight moved, for paying shippers, with nobody in the cab — on both ends of the length spectrum, middle-mile and long-haul, in the same fortnight.

The wrong response is to panic, and the second-wrong response is to wave it off as more AV hype. The useful response, if you run freight, is to figure out exactly which of your lanes this touches, why, and what you do about it. Here’s the map.

What actually got automated

Strip away the Doritos headline and the deployments have a shared shape. The freight that went driver-out is fixed, repeatable, and heavily mapped:

  • Gatik’s lanes are middle-mile, regional, and short — distribution center to store, the same route over and over, on roads the company has mapped in detail and validated for years. Gatik has run driver-out in Arkansas since the early 2020s; the PepsiCo news is scaling a proven pattern, not flipping a switch on something brand new.
  • Aurora’s lanes are long-haul but dedicated and corridor-bound — Dallas–Houston, Fort Worth–El Paso, and now Fort Worth–Phoenix, hub-to-hub on interstates Aurora has mapped and is comfortable operating, in fair-weather Sun Belt geography.

What did not get automated is everything that varies: irregular routes, multi-stop runs, unmapped lanes, dense or unpredictable urban delivery, anything that changes load to load or depends on a human reading an ambiguous situation. That’s not a temporary gap. Mapping, validating, and insuring a lane for driver-out operation is expensive and slow, so autonomy spreads lane by lane, not all at once. The network grows as a set of specific, named corridors — which is exactly why Aurora’s announcement counts routes, not territory.

The economics that should get your attention

The cost story everyone leads with — no driver wage — is real but it’s not the part that reshapes a lane. The part that reshapes a lane is hours of service.

A human driver on Fort Worth–Phoenix burns most of an 11-hour driving limit and then, by law, stops. The trip takes the better part of two days with mandated rest. A driverless truck has no clock to run out. Aurora’s pitch on that lane is roughly halving transit time, because the truck simply keeps moving. To match that with humans you need a team — two drivers, two wages, more complexity — or you accept the slower single-driver schedule.

On dedicated, predictable freight, that isn’t a marginal cost improvement. It’s a different service level: faster, more consistent, available overnight without a premium. For a shipper running steady volume on a fixed lane, that’s a genuinely better product. For a fleet that makes its margin running that same lane with human teams, it’s the competitive threat to take seriously — earlier than the driver-wage math alone would suggest.

What’s exposed first — and what isn’t

Put the two together and the exposure map is specific:

Most exposed: dedicated, high-frequency, point-to-point freight on mapped corridors — regional DC-to-store middle-mile and steady Sun Belt long-haul lanes. If a meaningful share of your book is one shipper, one lane, run the same way every day, that’s precisely the freight autonomy is being engineered to take, and the HOS math makes the long version of it attractive fastest.

Least exposed, for years: irregular-route truckload, multi-stop and final-mile-with-complexity, anything in hard weather or dense cities, specialized and high-touch freight (white-glove, hazmat nuance, live-load coordination), and the long tail of lanes too low-volume to justify mapping. The economics of mapping protect this freight more durably than any regulation.

The mistake is treating “autonomous trucking” as one wave that either does or doesn’t hit you. It’s a tide coming in on specific beaches. Know which beach you’re standing on.

Keep the hype honest

A few caveats worth holding onto, because the vendors won’t volunteer them:

  • “Tens of thousands of trucks” is ambition, not inventory. Today the reality is hundreds of medium-duty box trucks on narrow lanes (Gatik) and a network measured in single-digit-to-ten routes with a stated target of 200-plus driverless trucks by end of 2026 (Aurora). Real, growing, and still small.
  • Both companies have a financial stake in the narrative. Gatik and Aurora benefit from “driver-out is here” being the headline. The deployments are verified; the framing is theirs.
  • Driver-out doesn’t mean human-out. These operations still need remote oversight, maintenance, terminal labor, mapping teams, and recovery crews. The job mix shifts; it doesn’t vanish.

None of that makes the threshold-crossing less real. It just means the right planning horizon is “this changes my dedicated-lane economics over the next few years,” not “my OTR business evaporates next quarter.”

The regulatory runway

The reason any of this can scale is partly legal, and that’s moving too. The SELF DRIVE Act of 2026 (H.R. 7390, from Reps. Bob Latta and Debbie Dingell) would let AV makers self-certify their systems through a “safety case” and would preempt the patchwork of state restrictions that has kept autonomous trucking in semi-permanent pilot status. If it advances, the deployments above stop being state-by-state special cases and start being a national operating model.

It’s a House discussion draft, not law — say “would,” not “does.” And it’s contested: critics point out it would permit 80,000-pound driverless trucks to operate on safety cases the companies write themselves, with no requirement that the federal government verify them. Watch it as a leading indicator. If self-certification and state preemption survive into a passed bill, the scaling curve steepens.

Four moves to make this quarter

  1. Map your own book the way the AV companies map roads. Sort your revenue by lane: which is fixed, repeatable, point-to-point on major corridors, and which is irregular, multi-stop, or specialized? The first bucket is your exposed freight. You can’t reposition what you haven’t measured.
  2. Don’t panic-sell, but don’t re-sign blind either. The mapped network is still tiny. You have time. But on multi-year dedicated contracts in Sun Belt corridors, price in the possibility that an autonomous competitor undercuts the lane before the contract ends.
  3. Reposition toward what autonomy can’t take. The durable freight is the complex kind — irregular routes, high-touch service, lanes too varied or low-volume to map. If your differentiation is “we run a clean dedicated lane cheaply,” that’s the differentiation most under pressure. Build toward service and complexity, not raw lane efficiency.
  4. Track the SELF DRIVE Act and your state’s AV stance. The speed of all of this is gated by the regulatory runway. A passed federal self-certification regime with state preemption is the signal that the timeline just compressed.

Driver-out freight is no longer a question of whether. It’s a question of where and how fast — and for once, both are answerable. The operators who do well out of this won’t be the ones who saw it coming; everyone sees it coming. They’ll be the ones who knew which of their own lanes it was coming for, and moved first.

Sources