Signals Inbox·July 28, 2026·Autonomous Systems

When will Tesla catch up with Waymo?

Tesla will probably catch Waymo around 2030. Its factories and cheaper vehicle design could close the gap quickly once the software is ready, but Waymo still has the passengers, the driverless mileage and the safety record Tesla has yet to match.

Q1What would Tesla catching Waymo actually mean?

Tesla will have caught Waymo when its Robotaxi service carries roughly as many people, works reliably across several cities and has enough driverless safety data to be judged on equal terms.

We are talking about commercial robotaxis in the United States. Tesla could surpass Waymo in vehicle production, training data or manufacturing cost without matching the service that passengers can actually use.

One successful driverless ride proves the technology can work. A large robotaxi business has to complete thousands of rides through rush hour, roadworks, poor visibility, unusual pickups and emergency situations every day. It also has to do this without quietly relying on one human operator for every few vehicles.

Tesla does not need to copy Waymo exactly. A smaller Tesla network could reach parity if it covers wider areas, expands faster or carries passengers at a clearly lower cost. City announcements and demonstration rides alone would not qualify.

Q2How far behind is Tesla’s Robotaxi service today?

Tesla’s Robotaxi service remains far behind Waymo today, with a gap still measured in dozens of times on the most useful operating measures.

Alphabet said in its latest earnings call that Waymo had passed 500,000 fully autonomous rides per week across eleven major US cities. Waymo’s safety dashboard also shows 220.6 million miles completed without a human driver through the end of the first quarter.

Tesla’s first-quarter report showed about 1.7 million cumulative paid Robotaxi miles. Paid mileage had nearly doubled from the previous quarter, which confirms that the service was growing quickly, but Tesla still does not publish a weekly ride count.

One comparison gives a sense of the distance between them. Waymo assumes an average journey of about 4.4 miles in its current sustainability calculations. At 500,000 weekly trips, that equals roughly 2.2 million passenger-trip miles in an ordinary week. Waymo was therefore completing more commercial mileage each week than Tesla had accumulated by the end of the first quarter.

The comparison is slightly imperfect because Tesla has continued adding miles since then. The scale difference is still unmistakable. Waymo has a mature service with hundreds of thousands of weekly customers; Tesla is running an early network in limited parts of Austin, Dallas, Houston and Miami.

Q3Has Tesla proved that its Robotaxi works without a driver?

Tesla has proved that its Robotaxi can carry paying passengers without anyone behind the wheel. It has not yet shown that the service can handle mass daily use.

This is a real break from the old Tesla autonomy story. For years, customers watched supervised demonstrations while remaining legally responsible for the car. Tesla now offers autonomous rides to the public through a dedicated app in four metropolitan areas.

Tesla also moved beyond its original Austin launch. Its first-quarter update said unsupervised operations were ramping in Austin and had begun in Dallas and Houston. Miami has since joined the public service page.

The experience remains restricted. Tesla says Robotaxi operates in limited areas, currently between 6 a.m. and 2 a.m. Central Time, and additional stops cannot yet be added to a trip. Fair enough for an early rollout, but four city names are not four mature citywide networks.

Tesla has crossed the first important threshold: real people can pay for rides without an in-car driver. The next step is harder. Tesla must keep thousands of vehicles busy, make waiting times predictable and show that performance holds up when ride volume rises sharply.

Q4Is Tesla closing the gap with Waymo right now?

Tesla is growing quickly from a small base, but Waymo is still increasing its absolute lead in rides, cities and driverless experience.

Waymo was completing about 150,000 weekly rides near the end of 2024. The figure rose above 400,000 by the end of 2025 and passed 500,000 shortly afterward. That is more than a threefold increase in roughly fifteen months.

The city rollout accelerated at the same time. Waymo opened Dallas, Houston, San Antonio and Orlando together, the first time it had launched four public markets at once. Alphabet later said six cities had been added during the year, taking the total to eleven.

Tesla’s paid Robotaxi mileage nearly doubled during the first quarter, so the company is clearly moving. But doubling a small network does not shrink the gap when the competitor is adding hundreds of thousands of weekly rides.

Waymo now expects to reach one million rides per week by the end of the year. Tesla would need to grow much faster than Waymo for several consecutive years, because matching Waymo’s current size will not be enough once Waymo itself has moved beyond it.

Q5Do Tesla’s 12 billion FSD miles give it better data than Waymo?

Tesla sees far more road situations through its customer fleet, while Waymo has far more experience where the software handled the whole journey without a responsible driver.

Tesla’s live safety counter has passed 12.1 billion miles driven with FSD Supervised, including about 4.56 billion city miles. Its fleet encounters different roads, driving cultures, weather conditions and rare events on a scale that a dedicated robotaxi fleet cannot easily reproduce.

That gives Tesla a powerful training system. The company can find difficult situations across millions of vehicles, collect examples and improve one shared driving model. Tesla says its fleet collectively experiences the equivalent of a lifetime of driving scenarios every ten minutes.

Every one of those FSD Supervised miles still has a human driver responsible for watching the road and intervening. A driver may brake before a collision, take control at a confusing junction or avoid using the system in difficult weather. The mileage tells us how much material Tesla can learn from, not how many miles its software can safely manage alone.

Waymo’s 220.6 million rider-only miles answer the second question more directly. Nobody was sitting behind the wheel during those journeys. The total is much smaller than Tesla’s supervised mileage, but it is a far cleaner test of driverless reliability.

Q6Is Waymo safer than Tesla’s Robotaxi today?

Waymo has a much stronger safety case today because it publishes a large driverless dataset that researchers and regulators can examine. Tesla has released very little comparable Robotaxi data.

Waymo’s latest dashboard covers 220.6 million rider-only miles across five operating areas. Across all locations, it recorded about 0.01 serious-injury-or-worse crashes per million miles, compared with a matched human benchmark of 0.23. Its rate of crashes involving any reported injury was about 82% lower than the human benchmark, as was its airbag-deployment crash rate.

A separate Waymo study released recently adjusted the comparison for location, day of the week and time of day. Across 127 million autonomous miles, the researchers estimated that Waymo had been involved in 359 fewer injury crashes than comparable human drivers would have experienced. More than half of those avoided crashes occurred between 8 p.m. and 4 a.m., when human crash risk normally rises.

Waymo produces the research itself, so the methodology deserves scrutiny. Still, the mileage, crash definitions and human benchmarks are published in enough detail for outsiders to challenge them.

Tesla says FSD Supervised produces seven times fewer major and minor collisions than normal human driving. That claim is useful for judging driver assistance, but it cannot establish Robotaxi safety. The Tesla figures involve a supervising driver, different road exposure and a benchmark designed by Tesla.

Regulators are also examining weaknesses in the supervised system that supplies Tesla’s Robotaxi technology. One federal investigation covers alleged traffic-light, lane-marking and wrong-way errors. Another was recently upgraded to an engineering analysis after nine crashes in reduced-visibility conditions, including a fatal incident.

Those investigations do not prove that Tesla Robotaxi is unsafe. They do show why Tesla needs a large, separate and clearly explained driverless safety report before the comparison with Waymo becomes serious.

Q7Can Tesla’s camera-only Robotaxi approach really work?

Tesla’s camera-only approach can work, but recent regulatory findings show that poor visibility remains a serious weakness rather than a theoretical concern.

Tesla uses cameras and neural networks to understand the road. The system is simpler and potentially much cheaper than Waymo’s combination of cameras, lidar, radar and external audio sensors. It also fits vehicles Tesla already manufactures in large numbers.

Tesla keeps improving that approach. Its latest FSD architecture included a sharper vision encoder for low-visibility situations and software changes that reduced inference delays. A faster system can spot a problem and react sooner.

The hard part comes when the image itself deteriorates. Glare, fog, heavy rain, dust, darkness or a dirty camera can remove information before the neural network gets a chance to interpret it.

The latest National Highway Traffic Safety Administration analysis is uncomfortable for Tesla. Investigators found cases where FSD failed to recognise that camera visibility had deteriorated or warned the driver too late. The agency also said limitations in Tesla’s internal data and labelling may have caused some crashes to be missed in earlier reporting.

Waymo tackles the same problem with overlapping sensors. Its sixth-generation system uses thirteen cameras, four lidar units, six radars and external audio receivers. The company has begun carrying public passengers with this hardware and says it can support expansion into snowier cities.

Tesla may eventually prove that better software can replace expensive sensor redundancy. Today, Waymo has far more evidence that its chosen hardware works without a driver in difficult real-world conditions.

Q8Is Waymo’s geofencing stopping it from scaling?

Waymo’s geofences make expansion slower, but they are no longer trapping the company inside a handful of demonstration zones.

Waymo approves specific roads and conditions before opening them to passengers. This creates more work in every new market, including testing, mapping, depot preparation and coordination with local emergency services.

Tesla’s ambition is broader. The company wants one general driving model that transfers easily between cities and eventually works in customer-owned vehicles. If Tesla reaches that goal, it could open new areas with much less preparation.

Waymo’s recent expansion shows that the slower approach can still produce a national service. Its coverage has grown to more than 1,400 square miles across eleven cities, and it opened four metropolitan markets on the same day. Waymo is also preparing for more than twenty cities while adding airport access and freeway driving.

The geofence debate looks different now. Waymo accepts geographic limits, then keeps moving them outward. Tesla promises a more general system, but its public service still operates inside limited zones in four metropolitan areas.

Waymo’s method looks slower in theory. In practice, it is expanding faster.

Q9Can Tesla open new Robotaxi cities faster than Waymo?

Tesla could eventually roll out Robotaxi cities faster because it uses a common vehicle and software platform. Waymo currently has the stronger expansion record.

A Tesla Model Y can arrive in a new city with the same cameras, computer and core driving model used elsewhere. Tesla can also reuse its service centres, charging network and existing local staff. That should reduce the physical work required to start a fleet.

Even so, Tesla cannot simply send cars to a new city and switch autonomy on. The company still needs to test local roads, confirm how the system handles unusual junctions, arrange cleaning and maintenance, train remote-support teams and meet state rules.

Waymo has lately shown that its city-by-city process can be repeated quickly. It opened four public markets together, added six cities during the year and now operates commercially in eleven major US cities.

Tesla serves four metropolitan areas, but its own support page describes the coverage as limited. The company has the potentially lighter expansion model. Waymo has already repeated its model at much greater scale.

Q10Could Tesla’s manufacturing scale erase Waymo’s lead?

Tesla can manufacture enough Robotaxis to erase Waymo’s fleet lead, provided its autonomous software becomes reliable enough to put those vehicles on public roads.

Tesla produced more than 408,000 vehicles during the first quarter alone. Its factories have more than two million units of installed annual vehicle capacity, excluding future expansion. Waymo could never match that output through its current dedicated fleet operation.

Cybercab strengthens the case. Tesla has started pilot production and says volume manufacturing will begin this year. The vehicle was designed without conventional driver controls, which should reduce the cost and complexity of building cars solely for passenger service.

Waymo is much smaller as a manufacturer, although its capacity is no longer negligible. Its Arizona operation is moving toward tens of thousands of Waymo-enabled vehicles per year. The first public riders are also beginning to use the Ojai vehicle equipped with Waymo’s cheaper sixth-generation hardware.

Money should not stop Waymo from building that fleet. The company raised $16 billion at a $126 billion post-money valuation, with Alphabet remaining the majority investor. That gives Waymo room to finance vehicles, depots and new markets before fare revenue covers the full cost.

Tesla retains the overwhelming advantage in potential vehicle output. Waymo has enough manufacturing capacity and capital to keep expanding while Tesla finishes proving the software.

Q11Will Cybercab make Tesla cheaper than Waymo?

Cybercab will probably cost less to manufacture than a Waymo vehicle, but Tesla has not shown that it can deliver a cheaper passenger mile.

Waymo adds cameras, lidar, radar, computing and specialised hardware to its vehicles. Tesla aims to use a simpler camera-based system inside a car designed from the beginning for autonomous service.

That should give Tesla a lower hardware bill. Tesla can also share batteries, electronics, factories and suppliers with its much larger consumer-vehicle business.

Vehicle cost is only part of robotaxi economics. Cars must be charged, cleaned, repositioned, repaired and insured. The operator also pays for depots, customer support and remote assistance. A cheap vehicle that sits unused for much of the day can lose money faster than an expensive vehicle completing frequent paid trips.

Waymo is steadily reducing its own cost. The sixth-generation Driver uses a more streamlined sensor configuration, and the company is designing its Arizona factory to produce vehicles faster and in much larger numbers.

Tesla’s introductory fares and low sensor costs do not prove superior economics. We would need to know how many paid miles each vehicle completes, how often people intervene remotely and how much Tesla spends maintaining the fleet.

Cybercab gives Tesla the better cost theory. Waymo has the better evidence that passengers will use the service frequently enough to support a business.

Q12Will regulation slow Tesla more than Waymo?

Regulation will hold Tesla back in California, but its small service in more permissive states suggests that engineering remains the larger constraint today.

California’s current permit list places Tesla Robotaxi among companies allowed to test with a safety driver. Tesla is not listed as holding the same driverless deployment status that allows Waymo to provide its established commercial service there.

Tesla must therefore submit additional evidence before it can remove drivers and charge passengers across California. The federal investigations into FSD’s behaviour around traffic controls and poor visibility could make regulators ask harder questions.

The situation is easier in Texas and Florida. Both states have allowed Tesla to start operating without completing California’s longer approval process.

Yet Tesla remains in limited parts of Austin, Dallas, Houston and Miami. If permits were the main obstacle, Tesla could already be deploying thousands of vehicles across Texas. Its cautious rollout suggests that the company is still building confidence in the software and the operating system behind it.

Regulation will decide where Tesla expands first. Technical performance will decide how many vehicles it dares to deploy.

Q13Can ordinary Tesla owners really join the Robotaxi network?

A network of customer-owned Tesla Robotaxis could change the race faster than Cybercab production, but today it remains a plan rather than a functioning business.

Tesla gathers driving data from a fleet of more than six million vehicles. If even a small percentage eventually became legally capable of unsupervised driving, Tesla could add cars without buying every vehicle itself.

That model would attack one of Waymo’s largest expenses. Waymo must finance and operate its fleet, while Tesla could share revenue with owners who provide the cars.

Several difficult questions remain unanswered. Tesla has not said how many existing vehicles will meet the final hardware requirements for unsupervised autonomy. Current FSD still requires an attentive driver, regardless of how well it performs on an individual journey.

Fleet operations also become messy when private owners are involved. Someone must inspect damaged or dirty cameras, clean the cabin, handle lost property, maintain the vehicle and make sure owners supply cars during the hours when passengers want them most.

Many owners may prefer to keep their cars during evenings and weekends, exactly when ride-hailing demand often rises. Insurance and responsibility after a collision will need clear rules too.

The owner network is Tesla’s most powerful possible shortcut. We would not include it in the base-case timeline until Tesla shows unsupervised customer vehicles operating publicly and explains how the fleet will work day to day.

Q14What must Tesla prove before it becomes a real Waymo rival?

Tesla becomes a genuine Waymo peer when it reaches hundreds of thousands of weekly driverless rides, releases serious safety data and repeats the service across several cities without constant special handling.

The ride threshold does not need to match Waymo exactly. Tesla could remain smaller if its coverage is broader, its waiting times are better or its cost per mile is much lower. A network completing only a few thousand weekly trips would still be too small to judge reliably.

Tesla should also publish Robotaxi crash rates separately from FSD Supervised. The data should show miles driven, crash severity, road type, weather, city and whether an onboard monitor or remote operator became involved.

Tens of millions of driverless miles would provide a reasonable starting point. Rare but serious failures are difficult to measure across only one or two million miles.

The service itself must become ordinary. Passengers should be able to open the app throughout the day, request a car without an unusually long wait and travel across useful parts of a city. Airport and freeway access would strengthen the case because both introduce difficult operating conditions.

Finally, Tesla must show that remote assistance scales efficiently. Human support is normal in autonomous fleets. It becomes a problem when too many workers are needed to keep a small number of cars moving.

Q15When will Tesla catch up with Waymo?

Tesla will probably catch Waymo around 2030, with 2028 as the aggressive case and 2031 or later becoming likely if safety validation continues slowly.

A simple mathematical forecast would create fake precision because Tesla does not publish weekly rides, fleet utilisation or Robotaxi intervention rates. We can still build a reasonable timeline from the milestones Tesla must complete.

A 2028 catch-up would require an almost perfect sequence. Cybercab would need to enter true volume production quickly. Tesla would have to move from four limited metropolitan services to several large networks, release convincing safety results and avoid a serious regulatory pause. At the same time, it would be chasing a Waymo network that expects to pass one million weekly rides well before then.

The case for 2030 is more believable. Tesla could spend the next two years improving its driverless system and expanding Cybercab production, then use 2028 and 2029 to add thousands of vehicles and accumulate enough autonomous mileage for a credible safety comparison. Its manufacturing machine could make the final part of the catch-up surprisingly fast once software stops limiting deployment.

A later result becomes likely if Tesla’s camera system continues struggling with poor visibility, unusual intersections or other rare situations. Waymo is already deploying new hardware, building capacity for tens of thousands of vehicles and entering additional cities. It will keep moving while Tesla works through those problems.

Waymo already handles more than 500,000 rides per week. Tesla had published about 1.7 million cumulative paid miles by the end of the first quarter and still withholds its weekly ride count. The gap is too large for parity within the next eighteen months.

Our direct judgment is 2030. Tesla has the data network, factories and potential vehicle cost to catch Waymo eventually. Waymo has the service people use, the safety record regulators can inspect and the rollout process that has already worked across many cities.

Tesla could pass Waymo in vehicles produced before it matches Waymo in passengers carried. Genuine catch-up arrives when those vehicles complete large numbers of safe, fully driverless rides every week. Based on the evidence available now, that is probably four years away.