Zoox Dissected — How It Became the First Company to Charge Fares in a Car With No Steering Wheel

Jesse Levinson had spent nearly a decade at Stanford researching autonomous driving. He wrote the navigation algorithms for "Junior," the modified Volkswagen Passat that took second place at the 2007 DARPA Urban Challenge, and his doctoral advisor was Sebastian Thrun, who later led Google's self-driving program. Tim Kentley-Klay was a designer from Melbourne, Australia, whose career had been in ad directing and running an animation studio — no automotive or robotics background at all. He first incorporated Zoox in Melbourne in 2013, then moved to Silicon Valley to recruit engineers and raise venture capital.

The company's name comes from zooxanthellae, algae that live symbiotically inside coral. As the company explains it, zooxanthellae generate energy through photosynthesis and maintain a mutually beneficial relationship with the coral they inhabit — and Zoox wanted to build a mobility service that had that same kind of relationship with cities.

One decision made in that founding year would shape the next twelve. Instead of bolting sensors onto existing cars, Zoox chose to build a car from scratch. The company repeated that same point in a blog post on July 30, 2026: "Since founding the company in 2014, we have strongly believed that the best way to solve the challenges of autonomous driving is with a purpose-built robotaxi." The first prototype was built in 2015 inside a converted firehouse on a university campus, and the vehicle's first autonomous drive happened in a private parking lot.

The now-signature four-seat, face-to-face cabin layout traces back to a single 2014 photo — Levinson, Kentley-Klay, industrial design lead Nahuel Battaglia, and an early employee sitting in four lawn chairs arranged to face one another, just to see how it felt.

A Board Fires Its Founder, Then Amazon Steps In

The funding came fast. Zoox raised $200 million at a $1 billion valuation in June 2016, then $500 million at a $3.2 billion valuation in July 2018. Series B was led by Atlassian co-CEO Mike Cannon-Brookes. According to Form D filings with the SEC, the amounts Zoox actually received were $103.05 million and $50 million in 2016, $462.39 million in 2018, and $200 million in October 2019.

Then, just a month after Series B closed, on August 22, 2018, the board fired founding CEO Kentley-Klay. He tweeted: "I was fired by my board with no warning, no reason, and no chance to respond." The company has never issued an official explanation to this day. Reporting at the time cited investor concerns about his personality and leadership style, not fraud or misconduct — but all of that reporting relied on anonymous sources. Carl Bass, a board member and former Autodesk CEO, stepped in as executive chairman, while Levinson moved up from CTO to president. Kentley-Klay's name disappears from the list of related persons in the October 2019 Form D filing.

The company named its permanent CEO on January 14, 2019: Aicha Evans, who had spent 12 years at Intel, rising to chief strategy officer. At Intel, she'd run the communications and devices group, overseeing a 7,000-person organization spread across multiple continents. She started February 26. At the time, the company said it had more than 700 employees and had raised more than $750 million cumulatively.

On March 20, 2019, Tesla sued Zoox and four former Tesla employees over trade secrets. The dispute wasn't about self-driving technology, though — it involved documents on warehouse, logistics, and inbound/outbound procedures. The case settled in April 2020, with Zoox stating: "We acknowledge that some new hires from Tesla were in possession of Tesla logistics-related documents when they joined, and we regret the actions of those employees." Settlement terms weren't disclosed.

Amazon announced its acquisition on June 26, 2020. Amazon has never disclosed the price. The Financial Times reported it exceeded $1.2 billion, and outlets that reviewed the deal documents described a structure combining $1.3 billion in cash with more than $100 million in Amazon stock for employee retention. Amazon's 2020 annual report simply described the year's acquisitions as totaling "$1.2 billion net of cash acquired, of which $1.1 billion was capitalized as in-process research and development intangible assets" — without naming Zoox at all. In fact, the word "Zoox" doesn't appear anywhere in the body of Amazon's 10-K or 10-Q filings; the company's name only shows up in an earnings press release attachment (EX-99.1). It was a down round from the $3.2 billion valuation, and by early 2020 Zoox was burning through nearly $30 million a month.

Amazon's acquisition announcement never mentioned logistics or delivery — only ride-hailing. Levinson put it more bluntly in an August 2026 interview: "They liked the purpose-built robotaxi. If we hadn't had that vehicle, they wouldn't have bought us." After the acquisition, Evans and Levinson continued running the company, and Zoox remained an independent Amazon subsidiary. As of August 2026, Evans is still CEO and Levinson remains co-founder and CTO. Company materials put headcount at more than 3,000.

What It Actually Means to Have No Steering Wheel

According to company specs, the Zoox robotaxi measures 3,630mm long and 1,936mm tall, with an 8.4-meter turning circle. Its designed top speed is 75 mph (about 121 km/h) in either direction, and it can run for up to 16 hours on a single charge. Federal regulatory filings list its gross vehicle weight rating (GVWR) at 3,000 kg, classified as a passenger car seating up to four. When the vehicle was unveiled in 2020, its battery was listed at 133 kWh, but since 2025 the company's materials have dropped the kWh figure entirely, listing only hours of range. When Panasonic Energy announced in November 2025 that it would supply 2170 cylindrical cells starting in early 2026, that older figure was never updated.

That 75 mph figure is a design spec, not an operating condition. At the point its federal exemption took effect, Zoox's declared Operational Design Domain (ODD) capped speeds at 45 mph.

Removing the steering wheel isn't just dropping one component — it meant redesigning the entire safety architecture from the ground up.

Airbags are the clearest example. Without a dashboard, there's nowhere to deploy a traditional frontal airbag. The body of Zoox's federal petition describes four airbag types. A "horseshoe" airbag deploys from the ceiling in each direction of travel, forming a U-shape that prevents occupant ejection during side impacts and rollovers — and doubles as the reaction surface the frontal airbags push against. Frontal airbags drop from the ceiling at each seat; side airbags at each seat protect the pelvis; and airbags built into the seat cushions lift the front of the cushion during high-speed frontal collisions to keep occupants from sliding forward. That last type only deploys in forward-facing seats. But the vehicle manual and occupant manual attached to the same petition list a fifth type — rear airbags that deploy from behind the headrest in each direction of travel. Since the company's own documents don't agree on the count, the total number can't be pinned down.

The crumple zone was another problem. The car is too short to leave room for one to fold. Zoox's solution packs the drive module and motors tightly together so that the structure itself absorbs and dissipates crash energy before it reaches the passenger compartment. The company ran thousands of virtual crash tests before building a physical prototype in 2018, and conducted its first physical crash test in 2019. For the face-to-face seating configuration, Zoox says it applied a higher internal standard than federal rules require for rear seats. However, no third-party crash rating — from IIHS, NCAP, or similar bodies — exists for this design. Every piece of verification available so far comes from the company itself.

Redundancy borrows concepts from aviation. There are two batteries, primary and secondary steering systems, and simultaneous use of three cellular modems across AT&T, Verizon, and T-Mobile. Zoox describes this as "fail-operational" design — the vehicle keeps operating safely even after a failure — rather than "fail-safe," where it simply stops. Per the petition, if all three modems lose connection for an extended period, the vehicle automatically pulls over.

The Company Won't Say How Many Sensors It Uses

There are five sensor types: high-resolution cameras, lidar, radar, long-wave infrared cameras, and microphones. Four "sensor pods" mounted at the roof's corners form the core of the system; radar is distributed around the body, and microphones are embedded separately inside it. Despite there being four pods, they actually use only two part numbers, and they're designed to be upgraded independently of the vehicle body. Because the car has no hood or trunk, the corner pods have a direct, unobstructed line of sight down to the ground — which is the basis for its close-range vision.

The exact sensor count isn't disclosed by the company. Even the sensor layout diagram in the federal petition is redacted as "Confidential Appendix 1." Outside reporting gives conflicting numbers — EE Times said 64 in 2020, Forbes said 48 in 2025, and TechCrunch said around 40 in 2026 — likely reflecting different vehicle generations and different counting methods (whether microphones and interior cameras are included). Detection range figures don't line up across the company's own materials either: a July 2025 vehicle brochure says "150m or more," while the website's safety FAQ says "up to 200m." No evidence anywhere suggests Zoox designs its own lidar. Nelson Pedreiro, senior vice president of hardware, has said: "We have several manufacturing facilities working on our behalf, and more than 130 suppliers worldwide."

The most detailed picture of the software comes from a "Safety Case Framework" document released August 12, 2026. The self-driving system is built around five interlocking functions — localization, perception, prediction, planning, and control — sitting beneath an independent collision-checking layer. That layer uses its own perception and its own algorithms to judge whether the trajectory the main system has plotted is safe, and intervenes if not. The petition calls this layer the Collision Avoidance System (CAS), describing it as responsible for both longitudinal and lateral interventions and for deciding when an intervention should end. The safety metric is CIF — miles driven between crash, injury, or fatality events. The human-driver baseline was built from NHTSA's CRSS and FARS data along with the Federal Highway Administration's SHRP2 and mileage estimates.

The prediction models have named generations. Zoox originally used a graph neural network called UAP (Unified Active Prediction), then moved to QTP (Query-centric Trajectory Prediction) after concluding UAP couldn't handle unexpected behaviors like jaywalking or illegal U-turns. QTP pulls behavior directly from data rather than assuming human intent or destination.

Simulation runs on more than 10 million real-world driven miles as raw material. Synthetic simulation searches for conditions likely to produce a collision, while log-based simulation replays actual driving logs through updated software, using machine learning to identify and prioritize rare, safety-critical events for replay. Results are weighted by real-world exposure to produce risk estimates. The company also transfers scenes captured in San Francisco into an Austin environment as a form of cross-city simulation.

But Zoox has never disclosed its cumulative simulation mileage — a contrast to Waymo, which cites 20 billion miles. The federal petition redacts not only the number of scenario variants run per software release but also cumulative training miles and driverless miles; the category "cumulative simulation miles" doesn't appear at all.

One more detail: at a December 2025 AWS re:Invent session (AMZ304), Zoox disclosed that it trains a multimodal foundation model — one that processes camera, lidar, and radar together — on Amazon SageMaker HyperPod. According to AWS materials, this runs on 64-plus GPUs at 95% utilization, and vehicle data collection peaks at 4 TB per hour. This work has never been mentioned through any of Zoox's own channels. There's a gap between the stack the company describes in its public documents and the stack its own engineers describe at conferences.

Four Years of Self-Certification, Then an Exemption

The longest-running problem in Zoox's history wasn't technical — it was paperwork. U.S. Federal Motor Vehicle Safety Standards (FMVSS) are written assuming a car has a steering wheel and pedals. A car without them has two options: self-certify that it meets the standards, or apply for an exemption.

On July 11, 2022, Zoox announced it had chosen the first path, calling itself "the first company to self-certify a purpose-built vehicle that doesn't rely on traditional driving controls like a steering wheel, without seeking regulatory change or an exemption."

Here's how the following four years played out:

Date	Event
2022-07-11	Zoox announces FMVSS self-certification
2022-09-06	NHTSA issues special order demanding basis for certification
2023-03-03	Audit query AQ23001 opens, covering 64 certified Zoox vehicles
2024-12	Following field inspection, agency issues inspection report citing "multiple apparent nonconformances with applicable FMVSS"
2025-06-05	Zoox applies for demonstration exemption
2025-08-04	Exemption approved, AQ23001 closed. Condition: Zoox must delete/redact all prior FMVSS compliance claims
2026-07-30	Part 555 commercial exemption approved — paid rides allowed
2026-08-10	First paid ride begins in Las Vegas

NHTSA's closing report explained why the inquiry was opened: Zoox had chosen self-certification instead of applying for an available exemption, making it an outlier in the industry, and public information suggested it might not meet certain standards. The report added: "As a result of discussions with the agency during the investigation, Zoox decided to apply for an FMVSS exemption."

In other words, the 2022 self-certification wasn't validated — it ended with Zoox switching to an exemption and being required to delete its compliance claims. And the exemption doesn't even cover every vehicle in question. The closing report states: "This exemption covers all Zoox vehicles that were the subject of this investigation, except for certain 2023 model-year vehicles that Zoox has excluded from public road operation." Zoox marked its original 2022 announcement post with a note reading "Updated August 6, 2025," and in an August 15, 2025 post explained that its vehicles would for now be labeled "demonstration vehicles," meaning they "do not meet all FMVSS requirements related to manual driving controls and driver equipment." A July 2026 post describes those same four years as "thoughtful conversations with regulators around outdated FMVSS requirements."

The commercial exemption, announced July 30, 2026 and effective July 31, covers eight standards in part: windshield defrosting (103), wipers and washers (104), turn signal controls and headlamp switching (108), rear visibility and mirrors (111), foot-operated brake controls (135), sun visors (201), the requirement that windshields use AS1-rated glass (205), and airbag warning labels on sun visors (208). None of these are full exemptions from an entire standard — each covers a specific provision. Occupant crash protection performance and braking/lighting performance are not exempted, and glass safety performance testing requirements remain in place. The exemption runs two years, through July 31, 2028, with a cap of 2,500 commercially deployed vehicles per 12-month period.

The structure of the exemption's conditions matters here. NHTSA attached a standing oversight mechanism called "Operational Authorization" to this exemption — a tool that sets conditions by operating environment and can be modified in subsequent approvals. The first operational authorization was issued simultaneously with the exemption decision. In effect, the oversight that the now-withdrawn AV STEP program was meant to provide has been folded into Part 555. Separate from the annual manufacturing cap of 2,500, the number of vehicles that can operate simultaneously is set independently through this operational authorization.

NHTSA was explicit that its safety-equivalence finding applies only to the specific items Zoox sought exemption for, and that it reached no conclusion about compliance with the remaining standards. But the same notice states, in the very next paragraph, that in deciding whether granting the exemption served the public interest, the agency "considered the safety and maturity of Zoox's automated driving system." In short: driving performance was excluded from the compliance review, but factored into the public-interest determination.

July 2028 isn't a hard cliff. Federal regulation 555.8(e) states that if a renewal application is filed at least 60 days before expiration, the exemption remains in effect until the agency approves or denies it. The real risk isn't expiration itself — it's the possibility of denial, and the fact that the operational authorization can be scaled back or modified at any time. In the meantime, one standing program has already disappeared: NHTSA withdrew the AV STEP proposed rule on June 26, 2026. Industry groups called it too heavy a burden for an autonomous vehicle program; safety advocates said it lacked mandatory oversight.

Five Recalls — All Software, All Voluntary

Zoox doesn't sell its vehicles — it owns and operates them directly — so there's no owner-notification process for recalls. All five recalls to date involve software defects, and all were self-reported.

Recall	Filed	Vehicles	Defect
25E-019	2025-03-13	258	Excessive hard braking in response to crosswalk cyclists or fast-approaching vulnerable road users from the rear → risk of rear-end collision
25E-029	2025-05-01	270	Above 40 mph, system misjudged whether vehicles entering from side roads would stop
25E-037	2025-05-22	270	While stopped or moving slowly, system failed to continue recognizing a road user who had fallen right next to the vehicle
25E-090	2025-12-22	332	Unnecessary lane incursions near intersections (62 observed instances, no collisions)
26E-044	2026-07-16	105	Failure to detect and respond to heavy smoke, particularly at active emergency scenes

The first three recalls each stem from an actual incident. 25E-019 was triggered when a modified test vehicle with a safety operator aboard braked hard and was then rear-ended twice by a motorcycle, which led to NHTSA opening preliminary investigation PE24-015 in May 2024. 25E-029 followed an April 8, 2025 crash in Las Vegas, where a driverless Zoox vehicle collided with a passenger car. The company identified the triggering condition within two days of the crash, finalized the cause and a fix within three days, and deployed it fleet-wide within eight days — suspending driverless operation entirely in the meantime.

25E-037 followed a May 8, 2025 incident in San Francisco, in which an electric scooter rider collided with a stopped Zoox vehicle making a low-speed turn. The rider fell right next to the car, and the car then moved again. The injury was minor, but the incident exposed a gap: the system could miss a fallen person pressed against the body of the vehicle. The Ultra-Short-Range Detection (USRD) system Zoox describes in its federal petition is specifically designed to close that gap — it blocks the vehicle from moving if any object, regardless of classification, is detected immediately around or above the body. That said, the company has never formally identified USRD as the corrective action for this recall.

The 2026 incident is different in kind. On June 20, a driverless Zoox vehicle entered thick smoke at an active fire scene that had not been cordoned off with cones. The vehicle braked hard, changed direction, and stopped; a remote operator then reversed it so firefighters could set up cones. No one was injured. Zoox notified regulators on June 25 and pushed a fleet-wide software update on July 15. The recall report doesn't name the city where it happened, and the company has declined to disclose it.

What's notable is that Zoox's own defined ODD includes "light to moderate fog, haze, or smoke." Thick smoke falls just outside that boundary — one notch beyond it. This wasn't a failure outside the operating domain; it was a failure right at its edge. The recall was also filed during a window that NHTSA Administrator Jonathan Morrison had flagged in early July, when he told autonomous vehicle developers he'd observed "a clear pattern of driverless vehicles interfering with law enforcement and emergency response," and gave them until the end of July to propose fixes. The agency has not disclosed which companies received that letter.

Zoox's regulatory record is otherwise clean. Across NHTSA's entire investigation database, only two formal investigations have ever targeted Zoox — AQ23-001 and PE24-015 — and both are closed. As of the August 21, 2026 update, no investigations remain open. But much of that clean record comes from the company filing recalls voluntarily before regulators act. And notably, Zoox wrote blog posts about each of its March and May 2025 recalls, but wrote nothing about the December 2025 or July 2026 recalls.

A Plant That Can Build 10,000, Permits for 2,500

The Hayward plant opened June 18, 2025. It covers 220,000 square feet (about 20,000 m²) — roughly the size of three-and-a-half football fields, according to the company. At full capacity, Zoox says it can produce more than 10,000 vehicles a year.

What's unusual is how little of the plant is automated. Robots are only used for tasks like applying glass adhesive and moving the body down the line — everything else is done by people. There's no welding, stamping, or paint shop at all, which also means the plant uses less power than a typical automaker's factory. Suppliers pre-assemble major components and ship them in, and Zoox's quadrant-symmetric design lets it assemble the vehicle in a modular fashion.

End-of-line inspection is split into eight stations. At the sensor calibration bay, halogen lighting calibrates the visible-light cameras while simultaneously heating dots on a board to calibrate the infrared cameras from the same surface. During wheel and headlight alignment, since there's no steering wheel, the system electronically sets the steering angle to zero. On the dynamometer, the vehicle runs itself up to 75 mph with no driver, while a laser-based system detects lateral drift and keeps it centered on the rollers. From there it goes through a leak test, a light-tunnel visual inspection, a static test (FST), and a buzz-squeak-rattle test (BSR). The final dynamic test (FDT) is the vehicle's first autonomous drive — and because it's bidirectional, it runs for hours in both clockwise and counterclockwise directions.

Mass production began in July 2026. When the company unveiled its "production-intent vehicle" on June 24, it said it could ramp up to 100 vehicles a week, "pending regulatory approval." But what actually changed in that June reveal was cosmetic: interior color (aloe green seats, stone gray flooring), seat cushions and headrests, the touchscreen, a charging pad, cup holders, the positioning of color-changing reflectors used to indicate direction, and the door speakers and microphones. No changes to sensors, compute, batteries, or the drivetrain were announced.

This is where the numbers stop lining up.

Metric	Figure
Hayward production capacity	10,000+ vehicles/year
Federal exemption cap	2,500/year (5,000 total over two years)
Simultaneous operating fleet	Set separately by operational authorization
Nevada permit	100 vehicles
Actual cumulative production	~100–105 vehicles

If the plant ran at full speed, it would burn through the federal government's entire annual allotment in about 13 weeks. The bottleneck isn't manufacturing capacity — it's regulation and operations, a point the company itself acknowledged by attaching the caveat "pending regulatory approval."

Where Things Stand Right Now

As of August 24, 2026, Las Vegas is the only city where Zoox charges fares.

City	Status	Timeline
Las Vegas	Public, paid	Free public launch 2025-09-10 → paid 2026-08-10
San Francisco	Free, waitlist-invite only	Launched 2025-11-18, service area expanded 2026-03-24
Austin & Miami	Early access limited to employees and friends	2026-03-24
Atlanta, Los Angeles, Seattle, Washington D.C., Phoenix, Dallas	Test driving and mapping	Phoenix and Dallas added 2026-03-09

Cumulative autonomous mileage exceeds 3 million miles. Ridership figures vary by source. Zoox itself said on July 30, 2026 that it had served "more than 500,000 riders, with more than 500,000 on the waitlist." In early August, Axios reported figures combining Las Vegas, San Francisco, Austin, and Miami at "close to 1 million" — roughly double the company's own number.

Fares combine a base charge with distance and time, locked in at the time of booking, so riders don't pay extra if the vehicle circles back. Destination fees apply separately at Harry Reid Airport, The Sphere, and T-Mobile Arena. Pricing is positioned in line with the "comfort" tier of standard rideshare apps. But Zoox doesn't disclose its base fare or per-mile/per-minute rates — any specific dollar figures reported in the media are estimates.

Zoox still can't charge fares in California. It holds a driverless testing permit from the California DMV and a passenger-carrier pilot permit (with and without a driver) from the California Public Utilities Commission (CPUC), but it lacks the deployment permit required to charge fares from either agency. The CPUC permit explicitly states: "This permit does not authorize [the company] to receive compensation for providing transportation service in autonomous vehicles operated for testing purposes." Waymo, by contrast, holds a deployment permit and on August 14, 2026 won approval to expand in the Bay Area and Los Angeles while entering Sacramento and San Diego for the first time. Zoox can't generate revenue in the state where it's headquartered.

Zoox partnered with Uber on March 11, 2026. Under the deal, riders will be able to hail Zoox through the Uber app in Las Vegas starting summer 2026 and in Los Angeles by mid-2027, while Zoox keeps its own app running in parallel. As of August 24, though, there's no confirmation that Zoox rides have actually gone live inside the Uber app in Las Vegas.

Winning on Regulation, Losing on Scale

One thing is clear about Zoox: it's the only company in the U.S. charging fares in a vehicle with no steering wheel or pedals. Waymo's Jaguars, Zeekrs, and Ioniq 5s are all vehicles a human can drive; Tesla's Cybercab hasn't been deployed yet.

Every other metric tells the opposite story.

Metric	Zoox	Waymo
Vehicles on the road	~100–105	~3,500 (est. July 2026)
Cumulative autonomous miles	~3 million	220+ million (as of end of March 2026)
Weekly paid rides	Not disclosed (second week of paid service)	~500,000
Paid cities	1	11 metro areas
Third-party safety verification	None	IIHS study (2026-07-23)

Zoox's total ridership to date is roughly what Waymo handles in a week or two. The gap with Chinese competitors is just as wide. Pony.ai reported 1,975 robotaxis as of late June 2026 with plans to reach 3,500-plus by year's end, and WeRide was running more than 1,800 robotaxis as of late July. Baidu Apollo Go's Q2 2026 disclosure cited operations across 28 cities worldwide, with cumulative autonomous mileage of 350 million km.

Just ten days after Zoox launched paid service, on August 20, Nevada transportation regulators approved commercial robotaxi permits in Clark County for 5,000 Tesla vehicles, 1,000 Waymo vehicles, and 1,000 Uber-affiliated vehicles. Zoox's permit allows just 100 vehicles. Some of that 1,000-vehicle Uber allocation is earmarked for partners like Motional and Zoox, so it's not entirely competing volume — but it's still clear that the window in which Zoox was effectively alone in the only market where it could charge fares lasted less than two weeks.

What about vehicle cost? Zoox has never disclosed it. Here's what's known about comparable vehicles: Waymo's Jaguar I-Pace reportedly costs upward of $200,000 per unit. Waymo's imported Zeekr has a declared U.S. import value of about $38,000, but after the 127.5% tariff on Chinese-made EVs, that climbs to roughly $86,500 — and adding roughly $25,000 for its sixth-generation self-driving hardware pushes the total past $100,000. Tesla claims a manufacturing cost of $30,000 for its Cybercab, and Baidu Apollo's RT6 is reportedly priced at $28,000.

There's a counterargument that vehicle cost might not even be the decisive factor. According to analysis by Brad Templeton, vehicle depreciation accounts for only about 25% of per-mile robotaxi costs. Even if a vehicle costs 25% more, total per-mile cost only rises by 7–10% — a negligible difference at current rideshare price points of roughly $2.50 per mile. If that math holds, Zoox's real constraints aren't vehicle price at all — they're the 2,500-vehicle annual cap, the simultaneous-fleet limits set by operational authorization, Nevada's 100-vehicle permit, the lack of a California deployment permit, and a production rate that started at roughly one vehicle a day.

Editor's Take

There are two ways to read the Zoox story.

One is about the value of time. The 2014 decision to build a car from scratch paid off exactly as designed, twelve years later — Zoox is now the only company charging fares in a vehicle with no steering wheel, and while Waymo pays more than $48,000 in tariffs just to import a single Zeekr, Zoox is stamping out its own vehicles at a California plant. But that same decision also created twelve years of delay. Because Waymo modified off-the-shelf cars, it was able to scale in the early 2020s, and the 220 million miles it accumulated in the meantime is now feeding third-party studies like the recent IIHS research. Zoox has only 3 million miles, which meant it had no choice but to substitute simulation and analysis for real-world mileage in its safety case. When VP of Safety Qi Holmes said the company "had to lean on analytical simulation and data-driven approaches rather than accumulating miles," that sounds less like a methodological preference and more like a consequence forced by circumstance.

The other read is that regulation and scale are entirely different games. Zoox won the regulatory game. It spent three years on the self-certification path, got flagged for apparent noncompliance, changed course — and still ended up becoming the first purpose-built robotaxi to secure a commercial exemption. But the prize for that win was an annual cap of 2,500 vehicles. Ten days later, the same county approved 7,000 vehicles for competitors. Clearing regulatory hurdles first and capturing the market first turned out not to be the same thing.

Looking ahead, there are three things worth watching over the next year. First: will Zoox secure a driverless deployment permit from California's CPUC? Not being able to charge fares in the state where it's headquartered isn't a situation that can persist indefinitely. Second: can Hayward actually hit 100 vehicles a week? The gap between one vehicle a day at launch and that target is the company's real bottleneck — and on top of it sits the separate ceiling set by operational authorization on simultaneous fleet size. Third: the renewal review in July 2028. Under the rules, filing a renewal application on time keeps the exemption in force during the review period, but the outcome of that review is entirely out of Zoox's hands. What the company says it's hoping for, in its own petition, is a scenario where FMVSS itself gets revised by then, making its vehicle simply legal on its own terms — which is an even bigger variable outside Zoox's control.

No one knows exactly how much money Amazon has put into Zoox. Amazon has never disclosed it, and the company's name doesn't even appear in the body of its 10-K filings. The only mention Andy Jassy made of Zoox in his 2025 shareholder letter was a single line about it "just beginning commercial service." Levinson has said the company will need "double-digit billions" to scale the technology going forward. That statement — not a disclosure of past spending, but a forecast of future spending — may be the most accurate summary of where Zoox stands right now.