A truck driver passing on Interstate 45 looked twice. He wasn't looking at the freight. He was looking at the absence of a person in the cab of the Class 8 semi that had just overtaken him. That double-take, reported by CNBC after a three-hour ride from Houston to Dallas, is the most honest metric we have for where autonomous trucking sits today. The technology works well enough to be boring. The cultural shock hasn't worn off.
Last week, Kodiak Robotics was posting about a "final check" — a full run of a lane, start to finish, without the safety observer taking over, and expressing "confidence in the target we've set: driverless long-haul operations before the end of the year." That was a readiness statement. This week, Aurora Innovation put a CNBC crew and its CEO inside a driverless truck on a live commercial route and let the cameras roll. The delta is not incremental. Readiness metrics are inputs. Revenue runs are outputs. The industry just crossed from one to the other.
Aurora's numbers tell the story of a company that has stopped proving and started operating. The company currently has 20 driverless trucks on public roads and plans to grow that fleet to 200 by the end of this year. CEO Chris Urmson told CNBC the company is targeting positive free cash flow by 2028, after losing more than $800 million in 2025. That is not a research roadmap. That is a business plan with a P&L attached.
That is not a research roadmap. That is a business plan with a P&L attached.
The economics are the engine. Bank of America estimates Aurora's services will cost roughly $0.85 per mile versus approximately $1.30 per mile for human driver wages and benefits — before considering indirect labor costs. Morgan Stanley's Ravi Shanker estimates an autonomous fleet should be nearly 7.5 times as profitable as a human-driven fleet today. Those are not marginal improvements. They are the kind of gap that reorganizes an industry's cost structure.
The operating model is "driver as a service" — customers pay by the mile driven. Urmson framed the pitch plainly: "We can help them save on fuel economy. We can move goods more quickly and then we can fill the need when they can't hire amazing drivers for their teams." The driver shortage is real. Autonomous trucks can roll up to 20 hours a day. Federal law limits human drivers' hours. The math does the selling.
But the ride itself matters. Aurora has logged more than 440,000 miles since it began testing AV technology on public roads in 2022 — most of those with a human safety operator behind the wheel. The Houston-to-Dallas run was different. No one was behind the wheel. The truck pulled off the highway once, for less than a minute, to reset hazard lights that had come on for an unknown reason. Otherwise, three hours on I-45, uneventful. That is the goal. Uneventful is the product.
The unglamorous questions are where the real engineering lives. How does a driverless truck stop for fuel? Aurora's answer, posted to X: the company is "piloting a process with gas stations on our routes where we'll call ahead so an attendant is ready to fuel it up and send it on its way." That is a human-in-the-loop system, just not one sitting in the cab. It is also a reminder that autonomy doesn't eliminate labor. It redistributes it.
Weigh stations presented a similar problem. Trucks pull into them so states can confirm they're safe to be on the road. Kodiak and PrePass have built a way for a truck with no one in the cab to satisfy that requirement. Inspection status moves digitally through the PrePass network. Testing started in Texas and Louisiana. This is the invisible scaffolding that has to exist before driverless trucking scales. Nobody posts a press release when a weigh station integration works. But without it, the trucks don't move.
Gatik has taken a different route to the same destination. The company focuses on the "middle mile" — the short-haul routes between distribution centers and retail stores — rather than long-haul highway driving. In January, Gatik announced it had become the first company in North America to deploy fully driverless trucks in commercial operations at scale. The company reported $600 million in contracted revenue, with $400 million of that booked in the second half of 2025 alone. It has completed 60,000 fully driverless orders without incident, logged more than 2,000 hours of driverless operation, and covered over 10,000 driverless miles on public roads. Some routes extend up to 400 miles.
Gatik's fleet consists of 26- and 30-foot box trucks operating nearly 24 hours a day in the Dallas-Fort Worth region, the Phoenix metro area, and Northwest Arkansas. They move ambient, refrigerated, and frozen goods. The company's current driverless fleet is 10 revenue-generating trucks, with plans to increase to 60 in the coming weeks and expand to hundreds by the end of the year. CEO Gautam Narang was direct in the announcement: "Autonomous trucking is no longer a promise. It's a business."
That sentence is the thesis. The question the industry faced for years was whether the technology would ever work reliably enough to remove the safety driver. The question now is whether the unit economics will hold at scale. The early data says yes. Gatik's 60,000 incident-free orders and $600 million in contracted revenue are not pilot-program numbers. They are commercial-operations numbers. Aurora's 440,000 logged miles and 20-truck fleet are not test-track numbers. They are public-road numbers.
The skepticism hasn't disappeared. Sean Wu, CEO of the freight firm uShip, told CNBC he understands the cost appeal but questions how many shipping firms will want to ditch their drivers. "The drivers do more than just move the rig. There is judgement. There is customer service. There is protection," Wu said. "There is just a lot more to think about than 'Hey they are moving a truck from point A to point B.'" That is a fair point, and it is not a technology objection. It is a business-process objection. The answer will come from customers, not engineers.
The regulatory environment is watching closely. NHTSA leadership and the Pittsburgh mayor stopped by Aurora's booth at a robotics event to see the Aurora Driver up close. The conversation, per Aurora's post, was about "the jobs the industry is creating and the impact that this technology will have on safety." That is a different tone than the adversarial posture that defined AV regulation five years ago. The agency is engaging, not resisting.
Urmson's response to skeptics is a track record claim: "We've been able to put out timelines and we've basically hit those timelines." He also cited "millions of miles of experience" and "15 million tests before we put it out on the road." That is a defensible position. The question is whether it survives the first serious incident. It probably will, if the incident rate stays where it is. It probably won't, if it doesn't. That is the risk profile of every safety-critical system.
Kodiak's readiness metric from last week — a full lane run without observer takeover — was the setup. This week's revenue runs are the payoff. The company said it was confident in driverless long-haul operations before the end of the year. Aurora is already there. Gatik is already there, at scale, with Fortune 50 retailers and daily deliveries. The industry's center of gravity has shifted from "can it work" to "how fast can we deploy."
The truck driver on I-45 who looked twice probably won't look twice next year. That is the transition. When the double-take stops, the technology has won.