Let me concede something upfront. Waymo got there first. Cruise got there second, then unwound itself in a single quarter of GM board meetings. By any standard reading of the autonomous-vehicle race, Nuro — the delivery-pod company that pivoted toward passenger robotaxis — is showing up to a party where the open bar already closed. Founder Jiajun Zhu's argument is that this timing is the asset, not the liability. The companies that paid to discover what the regulators, the rider behavior data, and the sensor stack actually require have left a trail. Nuro plans to walk it.
What does Nuro mean by "second mover" in the robotaxi race?
Zhu's framing is narrow and you have to read it precisely. He is not claiming Nuro is the second autonomous vehicle company to ship a passenger product — Waymo, Cruise, Zoox, Motional and a handful of Chinese players were all ahead on that count years ago. The "second mover" he is describing is more specific: the second cohort of operators, the ones who enter after the first cohort has paid the cost of discovering what the operational design domain actually looks like once a paying rider is in the back seat.
The distinction matters. First movers in autonomy were not really competing on rider experience. They were competing on whether the underlying system could be made to function at all without a safety driver in the front. That problem is now answered — Waymo answered it in Phoenix, then San Francisco, then Los Angeles, and is now operating commercially across multiple metros. The unsolved problem has shifted. It is no longer "does the car drive?" It is "does the unit economic model close?" Nuro is positioning to enter that second question without having paid for the first.
Who already burned the cash Nuro now gets to learn from?
Two cautionary tales sit on the desk and a third is still being written. Cruise is the most expensive of them. Under GM's ownership, the company absorbed billions in losses building out a San Francisco rideshare operation, only to have the entire program suspended in late 2023 after a serious pedestrian incident exposed both engineering and disclosure failures. GM eventually folded the consumer robotaxi program entirely, redirecting whatever salvageable IP toward driver-assistance for its own vehicle lineup. The money is gone. The regulatory lessons remain.
Waymo, by contrast, is the survivor. Alphabet has reportedly poured something on the order of $11 billion into the program across its lifetime — a figure that has been widely cited in coverage of Waymo's funding rounds and Google's segment disclosures. The company now operates commercially in multiple US metros and is the de facto reference point for what production robotaxi service looks like. Nuro does not need to replicate Waymo's existence. It needs to learn from where Waymo over-spent.
The third cautionary tale is harder to score because it is mid-arc. Zoox, owned by Amazon, has been operating in test geometry for years and has produced a custom-built robotaxi vehicle, but commercial scale-out has been measured. The capital required to keep all three of these programs running has materially exceeded what any of them projected at founding. That gap between projected and actual is the central document Nuro is reading.
How does Nuro's delivery-bot history change the autonomy stack it brings to passengers?
Nuro spent the better part of a decade not building passenger vehicles. It built sidewalk-and-low-speed-road delivery pods — the R1, R2 and R3 platforms — operated in pilot zones in Arizona, Texas and California. From a hardware standpoint, the delivery business looks like the wrong training ground for a robotaxi. The vehicles are smaller, slower, and carry no humans. The argument cuts the other way once you look at the software layer.
The autonomy stack does not particularly care whether the payload is a passenger or a bag of groceries. What it cares about is the operational design domain — the streets, intersections, weather conditions and traffic patterns the system has been trained on. Nuro's pods accumulated road miles in residential neighborhoods and arterial roads under public-road regulatory exemptions, in some of the same metros where Waymo and Cruise were running passenger pilots. The perception, planning and prediction modules were learning from the same urban substrate.
The constraint Nuro is now relaxing is the speed envelope and the legal-occupancy framework. Adding a passenger to a vehicle that already navigates city traffic is a different category of engineering work than building the navigation itself from scratch. That is the leverage Zhu is claiming.
What regulatory groundwork has been laid that Nuro inherits for free?
The free inheritance is meaningful and quantifiable. California's DMV created an autonomous vehicle deployment permit pathway that did not exist a decade ago — both the testing-with-driver and the driverless-deployment categories now have published criteria, application forms, and a working pipeline of incumbents who have completed it. The NHTSA's exemption process for vehicles not meeting traditional Federal Motor Vehicle Safety Standards was reshaped under pressure from petitions from both Nuro and GM. The state-by-state patchwork that defined the 2018 era has consolidated into a smaller set of well-understood frameworks.
San Francisco's CPUC approvals for fared driverless service, which Waymo and Cruise both pursued, established the precedent that a state utilities commission can authorize a robotaxi to charge a fare on the same regulatory footing as a TNC. That precedent now exists as case law and procedural template. A second-cohort operator filing today is not asking regulators to invent a category — it is filing into a category that already has incumbents in good standing.
The cost of that regulatory pioneering to the first cohort was real. Lawyer hours, lobbyist retainers, agency engagement headcount — Cruise alone disclosed a meaningful policy and regulatory affairs operation in its corporate filings before the shutdown. Nuro can run a smaller policy team because the substantive battles are largely settled. That is one of the line items where the second-mover thesis actually has cash value attached.
Where does the second-mover argument break down for Nuro?
It breaks down in two specific places and a third more diffuse one. The first is the network-effect concern. Robotaxi services may turn out to behave more like ride-hail than like wireless carriers — meaning the first operator into a metro captures a rider habit, an app install, and a baseline of trust that the second operator has to displace at a marketing cost the spreadsheet did not anticipate. Waymo's brand recognition in Phoenix and San Francisco is now substantial. A second mover entering Phoenix in 2026 is not selling robotaxi-as-novelty; it is selling robotaxi-from-someone-not-named-Waymo, which is a different and harder pitch.
The second is the data-flywheel concern. Autonomous systems improve in proportion to the miles they see — and edge cases are sub-linear in miles, meaning each additional rare event costs more miles to encounter than the last. Waymo has accumulated tens of millions of driverless miles in its operational metros. A second entrant cannot buy that fleet history. It can only build it forward.
The third is the OEM-relationship concern. Vehicle platforms for robotaxi service have to be sourced from somebody. The companies that locked in long-term partnerships with Jaguar Land Rover, Hyundai, or Geely while those OEMs were still enthusiastic about driverless platforms got terms a 2026 entrant will not. That is a soft cost but a persistent one.
Which cities are the natural early markets for a Nuro robotaxi rollout?
The shortlist writes itself from the regulatory and operational record. Phoenix is on it — the climate is favorable, the road grid is regular, the Arizona Department of Transportation has the longest institutional memory of autonomous vehicle operation in the United States, and Nuro already has delivery-bot experience in the metro. Houston, similarly, sits on the operational maps of multiple AV operators thanks to flat geometry and a regulatory posture from the Texas legislature that has been deliberately accommodating.
Mountain View and the broader Bay Area are obvious as Nuro's home market — the engineering team is local, the political relationships are local, and the CPUC framework is now well-worn. The harder calls are the secondary metros where Waymo has not yet entered but where the second-mover thesis would actually be tested. Austin, Las Vegas, San Diego, and Miami all have municipal and state policy environments that have signaled openness without yet hosting a commercial driverless rideshare.
For a traveler trying to forecast where you might actually hail a Nuro vehicle, watch the city-by-city CPUC and DMV filings. Permit applications are public. The companies file their geographic intent before they file their press releases.
How do unit economics shift when you skip the R&D loss-leader years?
Here is where the math gets interesting. The reported cumulative spend on the leading first-cohort programs runs into the multi-billion-dollar range — Cruise had absorbed something north of $10 billion at the point GM pulled the plug, and Waymo's lifetime burn has been similarly large. The bulk of those dollars went into solving foundational problems: building the sensor stack, building the simulation infrastructure, building the safety case, building the regulatory relationships. Those costs are largely fixed and largely behind us.
A second-cohort entrant inherits an off-the-shelf sensor cost curve that has come down materially since 2020. The cost of a long-range lidar unit is now an order of magnitude below what it was when Waymo first built its custom Laser Bear hardware in-house. Compute costs per autonomous-mile-of-inference have dropped on the standard ML hardware deflation curve. Simulation environments are now partially available as commercial product, not bespoke engineering.
The variable-cost-per-mile question — the one that decides whether a robotaxi service operates above gross margin once you net out depreciation, energy, insurance, teleoperation labor and cleaning — is the one Nuro actually has to solve. The fixed-cost question is, to a meaningful degree, already paid. Whether that translates into a faster path to operating profit depends on rider-per-vehicle-hour, which is the metric to watch in any disclosures Nuro eventually makes.
What should a traveler watch on the calendar to test this thesis?
Three dates and one rolling watch item. The first is the next round of California DMV driverless deployment permit grants — the agency publishes its approved-operator list with each update, and Nuro's presence or absence on that list will be the first concrete signal of commercial intent versus pilot operation. Watch the DMV's autonomous vehicle deployment permit page through the back half of 2026.
The second is the CPUC's quarterly TNC and AV passenger service filings. New operator applications, geographic expansions and fare-collection authorizations are docketed publicly. Nuro filing for fared driverless service in a specific metro is the moment "robotaxi pivot" becomes "robotaxi operator."
The third is any NHTSA exemption petition or FMVSS amendment that touches custom-built robotaxi vehicle designs — that is the federal gate for a vehicle that lacks traditional driver controls, and the timing of any such petition will tell you whether Nuro plans to operate retrofitted standard vehicles or a purpose-built platform.
The rolling watch item is GM, Ford, and Volkswagen — the legacy OEMs who collectively wrote down billions on their own autonomous programs and have been quietly re-evaluating. If any of them announces a partnership with a second-cohort autonomy company, the second-mover thesis goes from interesting to consensus. That announcement is the one to watch for between now and end of 2027.
FAQ
Is Nuro actually launching a robotaxi service or still in the pivot stage?
As of the most recent public statements from founder Jiajun Zhu, Nuro has signaled a strategic pivot toward passenger autonomy after years of building delivery pods, but the company has not yet launched a commercial fared robotaxi service. The operational ramp for any robotaxi entrant typically runs 18 to 36 months from announced intent to public driverless rides, based on the Waymo and Cruise precedents. Watch DMV and CPUC filings for the concrete signals.
How much money has been spent in total on US robotaxi development?
Cumulative industry spend on US robotaxi programs is in the high-tens-of-billions range when you sum Waymo (widely reported at around $11 billion lifetime), Cruise (north of $10 billion before GM unwound the consumer program), Zoox (Amazon's acquisition price plus ongoing investment), and a long tail of smaller programs. Exact figures vary by source because not all spend is disclosed, but the order of magnitude is publicly defensible.
Are there robotaxis I can actually ride as a traveler today?
Yes — Waymo One operates commercial driverless service in Phoenix, San Francisco, Los Angeles and is expanding to additional metros. As of 2026 it remains the most accessible public-facing robotaxi service in the United States. Service availability is geofenced and app-gated, so you need the Waymo app and you need to be inside the operational area. Other operators run more limited pilots, often by invitation or waitlist.
Does Nuro's delivery-bot experience actually transfer to passenger autonomy?
Partially. The software layers that handle perception, prediction and motion planning are largely payload-agnostic — the system does not care if the cargo is a person or a package once you account for the speed envelope and the passenger-safety case. What does not transfer cleanly is the human-factors engineering: cabin design, ride comfort tuning, in-vehicle emergency response, and the entire interface layer a rider interacts with. That part Nuro has to build.
What is the main risk to the "second mover" thesis?
The data flywheel. Autonomous driving systems improve with miles, and edge cases scale sub-linearly with miles driven — meaning each additional rare event takes disproportionately more miles to encounter. Waymo's accumulated driverless mileage represents a real advantage that cannot be purchased off-the-shelf. A second-cohort entrant has to either accept slower edge-case learning or invest heavily in simulation infrastructure to compensate, and simulation is not a perfect substitute for road exposure.
When could a Nuro robotaxi realistically appear in a US city?
Based on the typical timeline from public pivot announcement to revenue-generating service in the AV sector, a realistic earliest window is late 2026 to 2027 for limited pilot operation, with commercial fared service likely in the 2027 to 2028 range depending on regulatory pace and capital availability. These are pattern-matched estimates, not company guidance. The actual schedule will be set by DMV and CPUC permit cycles, which run on their own clock.