Texas robotaxis, one trip at a time

Four Waymo cities, two booking arrangements

Dallas opened to all Waymo riders on August 4, Houston on August 20, and San Antonio on October 14, including SAT. Austin uses Uber: request an eligible product, receive a Waymo offer if matched, and accept it. The AV preference setting improves the chance of a match. It leaves the vehicle choice with matching. The dated Dallas, Houston, and Austin instructions describe those booking arrangements.

I would treat Austin’s Waymo option as a possible way to complete the trip. Declining the offer returns the rider to a human driver. A travel manager can approve that transport product; an organizer selling an autonomous experience needs a commitment to provide it. Uber explains the matched offer.

The following comparison puts each program beside the trip condition a planner still needs to resolve.

The airport is another journey

San Antonio’s opening covers SAT. Earlier guidance places pickup and dropoff at Terminal A/B Departures and the Arrivals outer curb near Rideshare Zone 2; follow the app’s instructions for the booked trip. Dallas’s August announcement described autonomous testing at Love Field and future freeway testing. I would book another option for Love Field, DFW, Houston’s airports, or Austin-Bergstrom unless the app offers the required journey. The opening announcement, terminal instructions, and Dallas notice cover different stages of access.

Because construction, traffic, and pickup restrictions can require a walk at either end, I would start the business-travel clock at the hotel door and stop it when the passenger is ready for the meeting, including the time and help needed to board at the assigned pickup point. Measuring only time inside the car rewards a service for moving inconvenience outside the measured trip. Pickup guidance.

A Waymo business profile can carry its own payment method; Austin’s Uber request uses the corresponding Uber arrangement. Texas wasn’t among the markets named on the checked Waymo for Business page. I would include an employer subsidy only after confirming the traveler’s eligibility through the profile and business-program terms.

The October 12 freeway guidance described limited access. I would preview the offered route before budgeting a suburban meeting. Freeway guidance.

Zoox can have a customer before it has a passenger

Zoox’s Austin page directed readers to a waitlist at the October 12 check. Dallas’s March announcement and Houston’s September announcement described mapping and testing. A supplier or depot landlord can have a customer before a hotel can offer guests a ride. The Austin page, Dallas announcement, and Houston announcement describe those stages.

Zoox’s retrofitted SUVs use a safety driver for mapping, testing, and validation. Its purpose-built robotaxi follows a different deployment path. Adding the two fleets together conceals what they can sell. Testing-fleet description.

The federal commercial exemption runs from July 31, 2026, through July 31, 2028. It allows up to 2,500 exempted vehicles to enter interstate commerce for commercial deployment in a twelve-month period, with local operation subject to Operational Authorizations. Zoox must also keep lifetime operational control on the grant’s terms. I would underwrite the vehicles against the applicable authorization and purchase commitment, because the federal ceiling assigns neither a Texas fleet nor a fare. Grant, sections X.A and X.G.

A depot landlord can have a sound investment before a passenger can buy a ride. Testing needs space, cleaning, equipment, and support. I would follow the signed order or lease: who pays during the ramp, when payment begins, and who bears specialized works if the rollout changes. The passenger forecast is worth financing only to the extent the contract puts someone behind the obligation.

A passenger has to fit the product

Tesla’s rider guidance listed limited Austin, Dallas, and Houston areas, with Model Y and Cybercab rules differing by vehicle. It requires self-booking and excludes added stops. Accompanied riders aged 8–17 can use Model Y; Cybercab’s range starts at 13. Wheelchair-accessible links refer riders to city providers. I would check the guest’s account, age, and boarding needs before offering the ride. Tesla rider guidance.

Avride’s Dallas service covers eligible Uber requests in Central Dallas north of the Trinity River. The optional AV offer includes an onboard specialist who can take control. Three passengers fit in the rear seats, and the boarding window is seven minutes. I would plan party size and pickup around those rider instructions.

May Mobility’s May 20 announcement described Arlington on Uber as upcoming, and the checked active-locations page had no Arlington booking entry. I would confirm the program with the operator before adding it to an employee’s itinerary. The launch plan and active locations answer that availability question together.

Authorization gets the operator to the booking question

Texas began enforcing its commercial driverless authorization requirement on May 28, 2026. I would check the authorization holder and vehicle listing, then the supported route and platform’s responder plan. The statutory test concerns commercial transport without a human driver, so an onboard person’s driving role matters. The TxDMV program implements the Transportation Code.

Authorization has no fee or expiry. Material information must be updated within 30 days of a change, a document duty with its own deadline. I would also require notice when the purchased service pauses and an alternative that can carry the passenger. A contractual credit arrives too late to make the meeting. Continuing duties and the operator lookup identify the operating party.

NTSB’s preliminary account of the January 12 Austin school-bus event describes an incorrect remote-assistance response before a Waymo passed the bus. No crash occurred, and the investigation was ongoing at the October 13 check. NHTSA opened a Cybercab self-certification inquiry on September 4. I would ask what the operator corrected and how assistance works for the service I’m buying. Different reporting rules and exposure prevent a valid safety ranking from raw incident totals. The Austin investigation, Cybercab inquiry, and reporting methodology set the scope of those questions.

What remains of the fare saving?

Assume 200 company trips a month. Geography leaves 160 eligible, rider and account rules leave 140, and pickups and hours leave 120 usable AV trips. At $20 for those trips and $28 for the other 80, the mix costs $4,640 against $5,600 for conventional rides. The apparent saving is $960.

Each failed AV attempt adds a $28 replacement fare and 20 minutes of staff time valued at $60 an hour, while the original AV charge stays paid. Each exception costs $48. Ten consume half the saving. Twenty consume all of it.

A busy city needs spare cars

Twenty failures among 120 attempted AV trips is a 16.7% exception rate. Dividing by all 200 requests flatters the rate while leaving every dollar unchanged. I would keep attempted journeys, eventual outcomes, and replacement charges together, because failed requests are precisely the experience the travel manager needs to price.

In a 2017 study, Castillo, Knoepfle, and Weyl model and examine in Manhattan Uber data a feedback loop in which scarce idle cars travel farther to collect passengers, consuming the vehicle time that could have served other trips and making nominal availability a poor guide to completion. The study concerns human drivers, including labor-supply responses; I apply its vehicle-time mechanism here. A car traveling empty to one pickup can’t collect somebody else. Matching research.

An autonomous fleet could improve the journey through more available cars, reliable operation, and better matching. I would adopt it early if those gains reduce walking or waiting, even at the same fare. I would compare completed trips at the times employees need them, including the conventional alternative.

Austin’s Uber Waymo match carries the same price as the selected eligible product with a driver. In a direct-booking comparison, an assumed $5 discount disappears with six extra minutes when staff time is valued at $50 an hour, so I would replace both the price and time assumptions with the offered quote and measured journey before budgeting a saving. Uber’s price parity leaves no automatic discount to spend on waiting.

I would begin with optional local trips and test the busiest departure windows. If a deadline matters, I would buy committed capacity and a timed alternative. The service earns its place when it improves the completed journey at the time the passenger needs it.