Aurora Driver: the first truly empty cab
Nobody has driven further with nobody aboard. Aurora's driverless trucks run twelve Sun Belt routes at night and in the rain — but the story of the observer PACCAR put back in the seat tells you where the real bottleneck lies.
On 1 May 2025 an Aurora-equipped Peterbilt ran freight from Dallas to Houston with nobody in the driver's seat, and American trucking changed. Sixteen months on, the numbers are no longer a demo: roughly 440,000 driverless miles, twelve routes across the Sun Belt, a 1,000-mile Fort Worth–Phoenix run that no single human driver is legally allowed to complete in a day, and freight moving at night and in the rain. Aurora is further along the hardest version of this problem — public highways, long haul, nobody aboard — than anyone else on Earth.
Getting there was messier than the press releases suggest. Three weeks after that first run, Aurora moved its observer out of the sleeper berth and back into the driver's seat — at the request of PACCAR, which built the truck and cited "certain prototype parts" in its platform. Aurora insisted the seat was unnecessary and that the Aurora Driver remained fully responsible; critics noted PACCAR had never publicly celebrated the launch and that its logos were blacked out in Aurora's own launch imagery. The requirement was never lifted. Aurora solved it by changing trucks: in July 2026 it launched Aurora Driver 2 on the International LT, observer-free, and quietly repositioned PACCAR to a 2027-and-beyond hardware partner.
Range of the FirstLight lidar on Aurora Driver 2 — double the first generation, and about 34 seconds of reaction time at highway speed. The hardware costs roughly half as much and is designed to last a million miles.
The autonomy
Aurora's philosophy is verification, not vibes. Chris Urmson — who ran Google's self-driving programme before Waymo existed — argues that pure end-to-end AI leaves you playing whack-a-mole, and hard-codes non-negotiable rules while applying machine learning where rules genuinely fail. The company claims some 10,000 requirements and four million validation tests behind its safety case, and in June 2026 it did something no rival has: paid an outside firm, Edge Case, to audit that safety case against UL 4600 and ISO 26262 and publish the result. Aurora reports zero collisions attributable to the Aurora Driver — a careful phrasing worth noticing, since it says nothing about crashes caused by other road users.
"Aurora has the miles, the money and the audit. What it doesn't have is a truck maker willing to say, out loud and without hedging, that the seat can stay empty."
What we're watching
- The 200-truck target for the end of 2026, with Roush upfitting toward a thousand a year — the gap between five trucks and two hundred is where this thesis is won or lost.
- Hirschbach's 500-truck memorandum converting into a binding order, and Volvo's own driverless launch in early 2027.
- The arithmetic: about $2 million of quarterly revenue against a $270 million quarterly loss, with $1.2 billion in the bank and free cash flow promised for 2028.
Fact sheet — Aurora Innovation (2026)
The ai.cars verdict
- The only company running long-haul freight on public highways with nobody in the cab, at genuine commercial scale.
- Aurora Driver 2 halved the hardware cost, doubled the lidar range and finally removed the observer PACCAR insisted on.
- Commissioning an independent audit of its safety case is the most grown-up thing anyone in this sector has done.
- The PACCAR episode cost fourteen months of clean narrative and revealed the real bottleneck: OEM liability, not software.
- Verdict: the leader on the metric that matters, burning $270 million a quarter to stay there. Watch the truck count, not the miles.
How does it stack up? Compare the Aurora Driver with the Kodiak Driver, or build your own comparison.
