California Wrote 100 Pages of Rules for Ticketing Robots. The Same Document Lets 80,000-Pound Trucks Drive Themselves.
One hundred pages. That is how long it took California's Department of Motor Vehicles to produce what Director Steve Gordon called "the most comprehensive AV regulations in the nation," adopted April 28, 2026, covering everything from how a cop tickets a car with no driver to how quickly a remote operator must pick up the phone when a fire truck pulls up behind a frozen robotaxi. Buried in the same regulatory package, past the geofencing mandates and the data reporting requirements and the first responder communication protocols, sits a provision that does something California has refused to do for years: it opens the state's highways to 80,000-pound autonomous semi trucks.
One hand tightens. The other opens wide.
Both sets of rules split into two articles. Article 3.7 covers testing requirements. Article 3.8 covers deployment. Together they create a two-track system: light-duty vehicles, meaning anything under 10,001 pounds, which is every robotaxi currently operating, and heavy-duty vehicles, meaning the Class 7 and Class 8 trucks that haul freight on interstates. Mileage requirements between the two tracks differ by a factor of ten, and the practical implications of that gap are far larger than the number suggests.
The 10× Mileage Gap and What It Actually Means
Light-duty AV manufacturers must complete 50,000 miles of supervised testing before applying for driverless testing permits, then additional driverless miles before commercial deployment. Heavy-duty manufacturers face a 500,000-mile supervised testing requirement at each phase.
Here is the math nobody has run. A typical autonomous truck testing program operates one to three trucks on public roads, each logging roughly 500 miles per day across an 8-to-10-hour shift, with testing days limited to around 300 per year after accounting for maintenance, weather, regulatory pauses, and route restrictions. That yields approximately 150,000 miles per year per company assuming a three-truck fleet. At that rate, clearing the 500,000-mile supervised testing threshold alone takes 3.3 years, and that is before the driverless testing phase even begins.
| Metric | Light-Duty (Robotaxi) | Heavy-Duty (Semi Truck) |
|---|---|---|
| Testing threshold | 50,000 miles | 500,000 miles |
| Typical fleet size (testing) | 20 vehicles | 3 vehicles |
| Daily miles per vehicle | ~100 | ~500 |
| Time to threshold | ~25 days | ~3.3 years |
| Effective time asymmetry | ~48× | |
For light-duty, a company running 20 test vehicles at 100 miles per day clears 50,000 miles in roughly 25 days. Waymo, with its existing fleet generating an estimated 357,000 fleet miles per day across its operating cities, could satisfy the threshold in a matter of hours from a pure mileage standpoint, though the structured safety case documentation would take longer. In practice, the time gap between the two tracks is not 10×, the ratio of the mileage requirements. It is closer to 48×, because test fleet sizes for trucks are dramatically smaller and because long-haul routes accumulate miles differently than urban ride-hailing loops.
Kodiak VP of External Affairs Daniel Goff told TechCrunch Mobility the company is "already working on the required documentation to apply for a permit," but Kodiak has not disclosed total fleet mileage, and neither has Aurora, which is already operating driverless trucks in Texas under that state's more permissive framework. Realistically, no autonomous truck company could reach commercial deployment in California before roughly 2029 to 2030, given the 500,000-mile supervised plus driverless testing progression.
How Do You Ticket a Car That Cannot Receive a Ticket?
On the robotaxi side, the regulation introduces a mechanism called the "Notice of Autonomous Vehicle Noncompliance." When a driverless car runs a red light or fails to yield, law enforcement issues the notice not to the vehicle but to the manufacturer. Manufacturers then have 72 hours to report the violation to the DMV, per the implementation of AB 1777.
There is no monetary fine attached.
Industry insiders told TechCrunch that the data trail is more valuable than a fine would be, and they are not wrong in a narrow sense. Accumulated noncompliance notices give the DMV a documented basis for permit suspension, fleet size restrictions, or geofencing limitations. Ahmed Banafa, a technology professor at San Jose State University, put it more directly: "This is the only way you can force a company to improve, through accountability." But the question TechCrunch's Kirsten Korosec raised after her reporting deserves repeating: why not both data collection and fines? A company generating hundreds of millions in ride revenue can absorb operational friction from paperwork far more easily than it can absorb per-incident financial penalties that scale with violation volume.
The Compliance Cost Nobody Has Calculated
Beyond the mileage gates, these regulations impose a standing operational cost that scales inversely with fleet size, and the math reveals a structural advantage for incumbents that the regulations never explicitly acknowledge.
Consider what 24/7 compliance actually requires. Staffing the 30-second first responder response mandate means running a teleoperations center around the clock, which typically requires five rotating shifts of at least eight operators per shift, totaling roughly 40 full-time employees at an average annual cost of approximately $85,000 per operator, or $3.4 million per year. Data collection and 72-hour reporting systems run approximately $2 million annually for the telemetry infrastructure, storage, and compliance team. A dedicated regulatory compliance unit of five full-time employees at around $120,000 each adds another $600,000. All told, baseline compliance infrastructure comes to roughly $6 million per year before a single vehicle turns a wheel.
For Waymo, operating approximately 2,500 vehicles across 10 cities and generating an estimated 130 million fleet miles per year, that $6 million amortizes to approximately 4.6 cents per mile. For a startup running a 50-vehicle test fleet logging 3.6 million miles annually, the same compliance infrastructure costs $1.64 per mile, making the per-mile compliance burden roughly 35 times heavier for the smaller operator.
| Fleet Profile | Annual Miles | Compliance Cost/Mile | Relative Burden |
|---|---|---|---|
| Waymo (~2,500 vehicles) | ~130M | $0.046 | 1× (baseline) |
| Mid-size (200 vehicles) | ~14.6M | $0.41 | ~9× |
| Startup (50 vehicles) | ~3.6M | $1.64 | ~35× |
| Truck startup (10 vehicles) | ~1.5M | $4.00 | ~87× |
None of these calculations appear in the DMV's regulatory impact assessment, and that is the sort of omission that shapes markets. Nothing in the regulatory language explicitly blocks small players. But the fixed-cost compliance floor creates a natural moat around companies that already have scale, and the companies that already have scale are exactly the ones that lobbied hardest during the public comment period.
The China Divergence
While California was opening its doors to autonomous trucks, China was slamming them shut on robotaxis. On April 29, 2026, Chinese regulators suspended all new autonomous vehicle permits nationwide following a Baidu Apollo Go outage in Wuhan that left more than 100 vehicles frozen mid-route, blocking intersections and requiring manual retrieval by emergency crews. The two largest markets for autonomous vehicles are now moving in opposite directions: California expanding its regulatory framework to accommodate heavier, faster, and more dangerous vehicle classes, while China, having previously led the world in robotaxi deployment scale, contracts after a single high-visibility failure.
Strategically, the timing is brutal. AV companies operating in China now face an indefinite permitting freeze. Companies like Kodiak, Aurora, and Gatik that have been developing autonomous trucking technology predominantly in Texas and Arizona suddenly have a path in California, the state that blocked them longest, at precisely the moment their potential Chinese competitors are grounded. Whether this timing was deliberate or coincidental, the effect is the same: California's 500,000-mile testing requirement is a barrier, but it is a barrier that domestic companies can begin climbing while their international competitors cannot.
What the Strongest Critics Get Right
Senator Dave Cortese, who introduced SB 1246 calling for additional human guardrails on autonomous vehicles, makes the strongest case against the heavy-duty opening. His argument is straightforward: California has never permitted commercial driverless operation of vehicles over 10,001 pounds. The physics of an 80,000-pound truck losing its automation at highway speed are categorically different from a 4,000-pound robotaxi doing the same thing at 25 miles per hour in a residential neighborhood. The kinetic energy scales with mass and with the square of velocity. A fully loaded semi at 65 mph carries roughly 130 times the kinetic energy of a Waymo Jaguar at 25 mph. The failure modes are not just worse in degree; they involve fundamentally different crash dynamics, braking distances, jackknife risks, and cargo hazards that the light-duty regulatory framework was never designed to address.
Five hundred thousand miles of testing does not directly mitigate this concern, because testing miles measure software maturity, not the physical consequences of a failure when that software encounters a situation it has never seen.
What We Did Not Prove
This analysis relies on publicly available data and reasonable estimates where companies have not disclosed specifics. Kodiak, Aurora, Gatik, and other autonomous trucking companies have not published total fleet mileage or daily testing rates, so our 500-mile-per-day estimate and 3-truck fleet assumptions are based on industry reporting and may be conservative for some operators and generous for others. Our $6 million annual compliance estimate is a bottom-up calculation using published salary data and typical enterprise infrastructure costs; actual costs will vary based on existing infrastructure, geographic distribution, and whether companies can repurpose compliance systems already built for other jurisdictions. Our 48× time asymmetry calculation assumes new market entrants; companies with years of testing in Texas or Arizona cannot transfer those miles to California but may be able to compress their California testing timeline through operational experience. Waymo's fleet miles are estimated from its reported 500,000 weekly rides at an assumed 5 miles per ride, and the company does not disclose this figure directly.
The Bottom Line
California built a regulatory machine that does two things at once: it creates accountability infrastructure for robotaxis that is genuinely novel (the noncompliance notice system, the 30-second response mandate, the emergency geofencing), and it opens the state to a vehicle class that carries 130 times the kinetic energy of the cars the framework was originally written for. Compliance costs embed a structural advantage for large operators, the testing timelines mean the first autonomous truck will not deploy in California until roughly 2029, and the absence of monetary fines for traffic violations means the primary enforcement mechanism is institutional data collection rather than financial deterrence.
If you are an AV startup: model the compliance cost per mile before you commit to a California launch. That $6 million annual floor is real and it does not shrink with a smaller fleet. If you are a trucking company considering autonomy: Texas remains years ahead in operational permitting while California's 500,000-mile requirement is a marathon, not a sprint, so plan your testing infrastructure accordingly. If you are a resident of a city where Waymo or Cruise operates: your city's police can now issue citations, but the citations carry no fines and enforcement depends on accumulated DMV data, so watch for whether the DMV actually uses its new suspension authority or treats the noncompliance system as a filing cabinet. If you are a state legislator in any other state: California just set the regulatory floor. Your constituents will ask why your state has not done the same, and "we are still studying the issue" stopped being a viable answer about 500,000 miles ago.