The robotaxi just got short enough to park in a garage

NHTSA just granted Zoox the first commercial exemption for a purpose-built robotaxi, and the design of the fleet changes everything.

In March we wrote that the AV bottleneck was moving from software to physical space. This is the first vehicle specification that changes which physical space an AV fleet can actually use.

What NHTSA actually granted

NHTSA announced the decision on July 30 and published it in the Federal Register on July 31, 2026. Temporary Exemption No. 2026-01 runs through July 31, 2028 and covers portions of eight Federal Motor Vehicle Safety Standards: windshield defrosting and defogging, windshield wiping and washing, lamps and reflective devices, rear visibility, light vehicle brake systems, occupant protection in interior impact, glazing materials, and occupant crash protection. Seven of the eight assume a human driver is present.

Zoox may introduce up to 2,500 exempted vehicles into commerce for commercial deployment in any 12-month period. It's the first passenger-carrying ADS-equipped vehicle ever granted an exemption under NHTSA's general exemption authority.

Two conditions matter more to our industry than the safety analysis does.

First, Zoox told NHTSA it will not sell the robotaxi. It keeps ownership and takes responsibility for daily operations, fleet management, maintenance and repair, routing and optimization, remote operations support, and customer service. NHTSA then made that a condition: Zoox must maintain operational control of every exempted vehicle for the entire life of the vehicle and may not sell or transfer possession.

Second, NHTSA said it intends to update FMVSS for controls and displays (101), vehicle lighting (108), mirrors and rearview display (111), electronic stability control (126), and sun visors and warning labels (201/208), and that once that work is complete, individual vehicle exemptions may no longer be necessary. It also streamlined both the Automated Vehicle Exemption Program and the Part 555 process.

Read that second one twice. Zoox walked through the door first and left it open behind them.

The number in the spec sheet

Zoox's published dimensions for the robotaxi are 142.9 inches long and 76.2 inches tall. Eleven feet eleven by six foot four. NHTSA's grant notice puts its GVWR at 3,000 kilograms, or 6,614 pounds, with a top speed of 75 mph and seating for four.

For comparison, a Jaguar I-PACE is 184.3 inches long, and once the roof sensor stack goes on, our site spec for a covered location has been eight feet of vertical clearance. We have walked away from a lot of well-located buildings over that number.

Most parking structures in America post a clearance bar at six-eight or seven feet.

A vehicle that stands six foot four clears a six-eight bar. Purpose-built robotaxis can park indoors. Retrofitted production cars cannot. That is the whole story, and almost nobody covering this ruling has said it out loud.

Where the deadhead math actually gets solved

Every AV fleet we talk to is running the same trade. Stage cheap and far, and pay for it in deadhead miles twice a day. Stage close and expensive, and pay for it in rent that scales with every vehicle added.

Until now that trade had a hard floor, because staging inside the service area meant surface land in the urban core. The most expensive dirt in any market, and the least available. Which is why AV depots keep landing in industrial submarkets well outside the service area, and why every fleet operating model we see carries a deadhead line item nobody is happy about.

The clearance bar is what kept fleets out of the cheapest well-located real estate in urban America. Downtown decks built for a five-day office population that never fully came back, sitting on soft occupancy, with power already in the building and a location by definition inside the densest part of the service area.

The bar didn't move. The vehicle did.

The charging that's already in the ground

Here's the part fleets should be looking at hardest, and it has nothing to do with the ruling.

Cities and states spent the last five years mandating charging capacity into buildings. California's CalGreen code requires new multifamily projects to make a substantial share of parking EV-capable, and the 2026 update pushes further on installed chargers in shared parking. Austin's 2024 energy code, effective July 2025, requires a set percentage of EV-capable and EV-ready spaces in new multifamily. Dozens of jurisdictions have a version of this.

Resident adoption did not keep pace with the conduit. NREL's EVGrid Assist modeling projects that by 2030, roughly 64% of EV charging energy will be delivered at single-family homes against about 4% in multifamily private parking. Property managers have a name for a bank of Level 2 chargers that four residents use, and it's the same name they use for the broken treadmill in the gym.

So there is wired, energized, already-permitted charging capacity sitting at low utilization in buildings that fall inside AV service areas. Nobody built it for fleets. It's there anyway.

Multifamily was never available to fleets. Now it is.

We have never been able to put a fleet client in a multifamily garage at any scale. The reasons stack up fast. Podium decks are built to a six-eight or seven-foot bar. Ramps are tight. Stalls are sized for residents. Property managers won't hand over forty contiguous spaces. And no owner wants sensor-covered SUVs cycling through a garage where people live.

Change the vehicle and most of that list resolves.

A six-foot-four pod clears the deck. An eleven-foot-eleven vehicle fits stalls striped compact. A fleet taking twelve stalls instead of forty is a conversation a property manager will actually have, because most multifamily is overparked against its zoning minimum and carrying stalls that generate nothing. And the daily curves line up better than anyone expects. Resident parking empties out on weekday mornings and refills at night. Robotaxi demand peaks at commute and evening and troughs midday. The vehicles want the deck exactly when the residents don't.

For a fleet, that reframes what market entry looks like. Instead of one hundred-stall depot in an industrial park with a long interconnection queue in front of it, you assemble ten to fifteen stalls across six or eight buildings inside the service area, on power that is already energized.

The juicer model

This is where it gets interesting operationally, and where we think most fleets are still thinking in the wrong shape.

Level 2 is slow. Zoox's published pack capacity is 133 kWh, which on a typical Level 2 circuit is most of a night. That's a real constraint if you're trying to turn a vehicle at peak and irrelevant if you're covering an overnight or a four-hour midday trough. Distributed Level 2 inside the service area is not a replacement for depot charging. It's the dwell layer.

Which means somebody has to plug the cars in. The micromobility industry already solved this with juicers, contractors paid per vehicle to charge and stage scooters overnight. The same labor model maps onto distributed AV parking: a contractor covering six buildings on a route, plugging and unplugging, doing an interior wipe-down, flagging damage, moving on. No fixed site staff. No depot overhead at every node.

Hub and spoke, in other words, which is what we argued for service vans in a different context. The central depot keeps the work that has to be centralized: sensor calibration, maintenance bays, deep cleaning, inspections. The spokes are cheap, close, wired, and already built.

Four things to check before any fleet chases this:

Access control is the hard part, not the parking. A driverless vehicle cannot badge through a gate arm or scan a fob at a resident garage entrance. Getting a fleet into a controlled deck means transponders, credential integration with the property's access system, and someone accountable when it fails at 2 a.m. This is the piece that kills most of these deals, and the piece nobody scopes in the LOI.

Existing circuits are shared and usually load-managed to protect residents. Adding a fleet to that panel is a load calculation and probably a utility conversation, even when no new service is required.

Weight is worth a look, though it's rarely a blocker. At a 6,614-pound GVWR the pod is in the range of a large SUV or half-ton pickup, which most decks were designed to carry. Concentrating twelve of them in one bay is still a question for the structural drawings.

Leases and HOA documents frequently prohibit commercial use of resident parking. That's a document problem, it's solvable, and it takes longer than the fleet's launch date assumes.

What this doesn't fix

Density is still a power problem wearing a parking problem's clothes.

The pod's other dimension helps on paper. A standard double-loaded bay is 18 feet of stall, a 24-foot aisle, and another 18 feet of stall. Sixty feet. Size the stall to a 12-foot vehicle and the module comes in near 50. Roughly 17% more rows per acre. The bigger gain is that the pod is symmetrical end to end with four-wheel steering and no three-point turn, so a lane can be stacked deep and drawn from either end, and stall width can drop because the nine-foot stall exists so a human can open a door and nobody opens a door at a depot.

None of which creates a single kilowatt. Doubling vehicles per acre gets a fleet to the transformer faster. Every depot deal we work still lives or dies on the load study and the timeline to energize, and the interconnection queue does not care how clever the striping is. That's precisely why already-energized capacity inside the core is worth more to a fleet right now than raw land is.

The company that made this happen will not be calling us

Zoox runs its own depots. New facilities in Phoenix and Dallas, a Fusion Center in Scottsdale, a testing fleet across ten markets, Amazon's balance sheet behind all of it. Paid rides start in Las Vegas on August 10, after nearly a million free rides across Las Vegas, San Francisco, Austin, and Miami. They are not shopping and we would not expect them to be.

This is the same spectrum we described in March. On one end, vertically integrated operators like Waymo and Zoox that control everything from the AI stack to the parking lot. On the other, lean AV companies that need someone else to handle the physical world. The exemption doesn't change the spectrum. It changes how many companies get to stand on it, and it changes what the cheap end of the real estate looks like for all of them.

Prediction: The first AV fleet to run a distributed urban-core charging network off existing multifamily and garage infrastructure will beat its competitors to profitability in that market, and it won't be because of the rent. It will be the deadhead miles it stops paying for on day one.

What we'd do Monday if we ran a fleet

Rewrite the site criteria. Every covered building screened out on clearance since 2023 goes back in the funnel, and the screen changes from eight feet to six-eight the day a purpose-built vehicle enters the fleet.

Stop underwriting markets as one depot. Underwrite them as one service hub plus eight dwell nodes, and price the deadhead you avoid against the premium you pay to sit inside the core.

Ask about power that already exists before asking about power you'd have to build. In a garage or a multifamily deck, the question is what got installed to satisfy code and how little of it is being used.

Everyone spent last week talking about a vehicle with no steering wheel.

Six foot four is the number that moves the ground game.

About Mobility Places: Mobility Places sources and manages parking for fleets. We've spent years translating between two industries that rarely speak the same language: fleet operators who need physical space, and property owners who don't understand fleet operations. Our clients today are microtransit companies and mobile service fleets. The next generation looks like AV depot operators, robotaxi fleet managers, and the third-party companies being stood up to run autonomous vehicle operations in new markets.