The holiday mood has barely faded, yet the global autonomous driving industry has already hit the fast-forward button.

On October 8, Waymo announced the completion of a $5 billion term loan, its first-ever debt financing, secured against intellectual property, with lenders including Blackstone, PIMCO and Sixth Street.

On the same day, Texas Department of Motor Vehicles records showed that Tesla Motors (TSLA) added 150 newly registered Cybercabs in a single day, bringing the total to 319 units, the largest single-day increase since the model entered the registration system.

At the same time, XPENG officially launched the Chinese name "XPENG YOYO" for its Robotaxi business, and the XPENG YOYO autonomous ride-hailing mini-program went live simultaneously, inviting beta testers.

While some players are raising capital and expanding their fleets in the capital markets, others are accelerating production on the manufacturing side, and new entrants are forcefully joining the fray.

Behind these moves, a noteworthy shift is taking place: competition in Robotaxi is moving beyond validating whether technology can achieve driverless operation and extending to validating whether a commercial system can run sustainably.

Technology still determines whether vehicles can safely hit the road, but it no longer single-handedly determines victory.

The proliferation of end-to-end and VLA large models is turning technical capability into an "entry ticket" for scaled operations, rather than the only advantage that can create separation.

Going forward, scaled vehicle supply, sustained order growth and continuously declining operating costs will jointly determine whether technological advantages can be converted into commercial advantages.

Robotaxi competition has evolved from a contest of isolated technical capabilities into a battle of entire commercial systems.

At the data level, the safety baseline for scaled operations has already been validated.

An independent study by the U.S. Insurance Institute for Highway Safety (IIHS) covering approximately 50 million miles of autonomous driving operations showed that Waymo had 1.28 reportable crashes per million miles, compared with 4.06 for human drivers.

As of October 2026, Waymo had entered 15 U.S. cities, with a fleet of more than 4,000 vehicles, completing over 500,000 paid rides per week.

To understand this Robotaxi competition, it can be broken down along four dimensions: the differences among players lie not in "who does the operations," but in which links they control and whom they depend on.

Autonomous Driving Layouts of Major Global Mobility Platforms

Who Is Accelerating, Who Is Blocked by Regulation?

Global autonomous driving is forming a "three-way race" among China, the U.S. and Europe, with major markets almost simultaneously entering a critical window for operational deployment.

The Middle East and Europe are accelerating deployment. WeRide Inc. (WRD) has achieved fully driverless commercial operations in Abu Dhabi and Dubai, covering more than 70% of core urban areas, with a local fleet of about 400 vehicles. In Europe, it partnered with Uber to launch in Madrid, Spain, and Zurich, Switzerland.

Waymo, with its base of 15 U.S. cities, has simultaneously launched international expansion, starting its first overseas test in Tokyo in 2026 and announcing plans to enter more than 20 cities including London and New York in the future.

Chinese companies are also accelerating their overseas pace. In July 2026, Baidu's Apollo Go, together with Uber and Lyft's Freenow, launched public road testing in London, with plans to offer driverless mobility services to the public starting in 2027. This was the first on-road test of Chinese autonomous vehicles under complex road conditions in London.

Pony.ai, meanwhile, expanded its global strategic cooperation with Uber, planning to deploy more than 2,000 Robotaxis across five European cities based on its existing fleet in Zagreb, Croatia.

Tesla Motors's Cybercab simultaneously launched static tour exhibitions in Beijing and Shanghai as well as Hong Kong, but explicitly stated it would not involve commercial operations in the Chinese market.

However, policy loosening does not mean regulatory relaxation. In September, shortly after the Cybercab launched, the U.S. National Highway Traffic Safety Administration (NHTSA) announced a review of approximately 1,000 Cybercabs, assessing how Tesla Motors determined that this vehicle — without a steering wheel or pedals — meets federal safety standards through self-certification.

This review demonstrates that autonomous driving safety remains under intense regulatory scrutiny. Waymo's experience in New York is even more telling: despite holding a testing permit, New York State requires a safety driver inside the vehicle, effectively prohibiting fully driverless commercial service.

Among the more than 20 cities Waymo plans to enter, regulatory barriers in key markets such as New York and London remain the most uncertain variable.

Robotaxi Deployment Scale by Region

Who Controls Key Links, Who Depends on Partners

Since 2025, competition has extended from algorithmic rivalry to the combination of technology, vehicles, traffic and urban resources.

Mobility operators' advantage lies in demand pools and ground-level relationships, while technology companies' advantage lies in deep hardware-software integration. There is currently no consensus on which is better suited to integrate the industrial chain.

Most recently, Uber simultaneously connected with more than six technology providers including WeRide Inc. (WRD), Pony.ai, Apollo Go, Waymo, Zoox and Wayve. Its autonomous driving mobility services grew more than 10-fold year-on-year and are currently live in 8 cities, with a year-end target of 15 cities covering Hong Kong, Madrid, Zurich, Houston, London, Munich, Los Angeles and the San Francisco Bay Area.

In Europe, Bolt partnered with Lucid to launch a deployment plan for 25,000 Robotaxis, an important step toward Bolt's goal of deploying 100,000 autonomous vehicles on its platform by 2035.

Southeast Asia's Grab entered through a tens-of-millions-of-dollars equity investment in WeRide Inc. (WRD).

Currently, three typical paths can be observed:

Technology company vertical extension (Waymo): self-developed driving systems, partnership with Hyundai Motor to produce IONIQ 5 Robotaxis at its Savannah, Georgia plant, and self-built operations. High hardware-software integration, but heavy capital investment and long cycles.

Platform aggregation (Uber, Grab, Bolt): leveraging existing demand pools and dispatch capabilities to connect multiple technology providers in an asset-light manner. Fast expansion, but requires coordinating standardized integration across different suppliers, with dependence on external partners for technology and vehicle costs.

Automaker full-chain (Tesla Motors, XPENG): integrated manufacturing, technology and operations. Tesla Motors's Cybercab began operations in Texas in early September and plans to support all-weather operations through FSD iteration; domestically, XPENG launched its Robotaxi business "XPENG YOYO." Manufacturing capability is an advantage, but faces uncertainty in regulatory certification.

The difference among the three paths lies in varying degrees of control over four links — vehicles, algorithms, demand and urban operations — which in turn determines per-vehicle cost and expansion speed.

WeRide Inc. (WRD) founder Han Xu offered another perspective: "The Robotaxi track is very crowded. Ultimately, the companies that can be profitable will be those that solidly solve users' real needs."

How Does Scale Convert into Efficiency?

When technical capability becomes a necessary condition for entering scaled operations, competition extends further to fleet utilization, order density, vehicle depreciation, dispatch efficiency and cross-city replication capability.

This barrier consists of three layers.

The first layer is the complexity of ground operations. City deployment goes far beyond technology installation; it involves regulatory communication, insurance system construction, parking dispatch, ground operations team building and a series of "hard chores." Each new city entry means building a ground ecosystem from scratch.

This is precisely the core capability that mobility operators have accumulated over years.

The second layer is the flywheel effect of safety mileage. Pony.ai's Wang Haojun articulated this incisively: "Full driverless is not about deploying first and supplementing data later, but about using scaled operations to continuously accumulate safety mileage and earn public and regulatory trust." He drew an analogy to commercial aviation: safety records take a long time to accumulate, and ultimately very few players will be able to achieve reliable and reasonably cost-effective operations.

This "flywheel effect" means first movers' advantages will continuously amplify over time. Latecomers, even with comparable technology, need to re-accumulate mileage data to prove L4 safety.

Apollo Go's data provides another observation sample. As of June 2026, it had cumulatively provided more than 23 million autonomous driving ride-hailing orders to the public, with cumulative autonomous driving mileage exceeding 350 million kilometers, of which fully driverless mileage exceeded 240 million kilometers.

This shows that scaled operations not only mean more vehicles in service, but also mean companies are accumulating longer actual road operation mileage, larger-scale mobility service experience, and opportunities to continuously validate system safety.

The third layer is achieving a viable profit model. The premise of scaling is a viable per-vehicle economic model, which tests operational efficiency — whether algorithmic capability can be converted into stable orders, lower empty-running rates and controllable unit costs.

WeRide Inc. (WRD) provides an industry reference in this regard: over the past six months, average daily orders per vehicle reached 15, rising to 26 during peak periods, average passenger wait time dropped to under 10 minutes, and registered users grew more than 900% year-on-year in Q4 2025.

Its total cost of ownership per Robotaxi vehicle in the Chinese market has decreased by approximately 38%, and its Middle East subsidiary's Robotaxi business has achieved operating profitability, making it one of the few autonomous driving companies globally to achieve profitability in a regional Robotaxi market.

But it is important to see clearly what this data proves and what it does not: order volume, TCO reduction and regional operating profitability are indicators at different levels and cannot substitute for each other.

To truly judge Robotaxi profitability, these questions still need to be answered: how much time per day does each vehicle actually spend carrying passengers, rather than empty-running, charging or waiting; what are the revenue and costs per order; how are vehicle depreciation, insurance, remote assistance and ground operations allocated; and how many vehicles and orders are needed to amortize upfront investment in new cities.

Most of this data is currently undisclosed, so at this stage it can only be said that improved operational efficiency is a necessary condition for profitability, not profitability itself.

The Scale Dilemma: Why Efficiency Does Not Equal Profitability

Even as efficiency continues to improve, large-scale deployment still faces multiple bottlenecks, and the "big exam" for autonomous driving is far from over.

First, emergency response capability when urban infrastructure fails. The large-scale power outage in San Francisco in December 2025 remains an industry wake-up call. After traffic signals went dark, multiple Waymo vehicles stalled at intersections, causing citywide traffic paralysis, with even emergency vehicles unable to pass.

The Mayor of San Francisco subsequently wrote to regulators requesting a pause on Waymo's expansion until it meets stricter safety conditions. On July 4, 2026, during Independence Day fireworks celebrations, a similar scene played out again, with dozens of Waymo vehicles causing massive congestion near the Golden Gate Bridge.

This series of incidents exposed more than technical problems. When urban infrastructure fails, autonomous driving systems lack emergency degradation plans, and it also shows that mature collaboration mechanisms between autonomous driving companies and city managers — in emergency response plans, data transparency and other areas — have yet to be established.

Second, the dual test of per-vehicle economics and operational stability. Achieving a viable per-vehicle economic model requires simultaneous optimization across fleet size, dispatch efficiency, vehicle depreciation, energy costs and multiple other dimensions.

But growth in order volume does not mean operations have entered a stable state. Apollo Go's operational data provides a noteworthy sample. In Q1 2026, its fully driverless orders reached 3.2 million; in Q2, approximately 1 million.

Baidu attributed part of this order fluctuation to operational adjustments made in some Chinese cities due to regulatory factors. This shows that Robotaxi operational scale depends not only on vehicles and user demand, but is also affected by regulatory arrangements, urban operating conditions and other factors.

Even after accumulating substantial orders and road mileage, companies still need to continuously respond to changes in operational tempo and validate the stability of per-vehicle economics under different cities and regulatory conditions.

Therefore, measuring a Robotaxi company's commercialization progress cannot rely solely on cumulative orders or single-quarter peaks; it also requires observing whether order growth can be sustained, whether vehicle utilization can improve, and whether unit costs can decline as scale expands.

Third, the fragmentation of regulation and liability systems. From NHTSA's review of the Cybercab to New York State's de facto ban on fully driverless operations, fragmented regulatory frameworks are the biggest obstacle to global deployment.

The establishment of insurance systems is similarly lagging. When accident liability is allocated among algorithms, vehicle manufacturers and operators, traditional auto insurance models face fundamental challenges. IIHS's 68% safety data is a positive signal, but changing public perception requires time and continuous safety records to validate.

Every occasional accident can cause a disproportionate impact on industry trust.

Technological Advantages Must Be Tested by Operating Systems

Han Xu once used Churchill's words after the Dunkirk evacuation to position the current Robotaxi process: "This is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning."

This autonomous driving marathon is already past the halfway mark. The first half was about whose technology could get on the road; the second half is about whose operating system can survive, converting technological advantages into advantages tested by operating systems and thereby into commercial advantages.

Looking at the operating conditions of major global players, fleet size does not equal operational efficiency, and operational efficiency does not equal profitability.

Waymo is a benchmark case of "no vehicle additions, more orders," with about 3,000 vehicles pushing weekly orders from 50,000 to 500,000. Chinese fleets exceed 4,000 vehicles, yet utilization lags relatively behind.

But since 2026, Pony.ai has achieved per-vehicle profitability in Guangzhou and Shenzhen, and WeRide Inc. (WRD) is approaching break-even in the Middle East, both exploring possibilities for the next step in improving efficiency and profitability.

Why are fleet size and order volume not proportional? The answer lies in the relationship among fleet utilization, order density and cost structure, which is also the most important core indicator to track going forward.

It can be said that scale can be bought, efficiency can be ground out, but profitability is the true tipping point.

The next phase of Robotaxi depends on who can organize vehicles, orders, urban resources and safety systems into a sustainable, replicable commercial system.

Only when technological advantages are tested by operating systems will they become true commercial advantages.