Deep Dive: Chinese AV Player Q3/2025 Earnings Analysis
(I analyzed the earnings calls, so you don't have to)
Welcome back,
Both WeRide and Pony.ai reported strong revenue growth, improving unit economics, and accelerating international expansion. But beneath similar headlines sit two different strategies.
This edition breaks down what actually matters:
how each company is scaling, where the economics are improving, and why their global footprints are starting to shape the competitive landscape from Abu Dhabi to Europe.
We also cover the underlying shift that becomes clearer with every earnings call and every new deployment: the AV race is no longer defined by who has the best autonomy stack, but by who can build and deploy fleets at scale, across multiple markets, with sustainable economics.
Let’s get into it.
⏱️ ~2,800 words, 12 minute read
🇨🇳 China’s AV Leaders Hit Milestones, Diverge on Strategy, and Accelerate Global Expansion
Two of China’s leading autonomous-vehicle companies reported Q3 results within 48 hours of each other. Together, WeRide and Pony.ai gave the clearest view yet of how China’s AV sector is evolving: more global, more capital-efficient, more diversified, and closer than ever to achieving scalable unit economics. But beneath the shared momentum sit two very different strategies.
WeRide Q3/2025 Earnings Analysis
Revenue in Q3 grew 144% YoY to $24 million, driven by major increases in robotaxi deployments, robobus sales, and intelligent data services. Product revenue rose 428% YoY, while service revenue grew 67% YoY.
The standout:
Robotaxi revenue surged 761% YoY to $5 million
Share of total revenue from robotaxis climbed to 20,7%, from only 5,8% last year
Gross margin increased from 6.5% → 32.9%
Operating expenses dropped sharply (-51% YoY), mostly due to the unusually high share-based compensation in 2024. WeRide still grew R&D significantly (+39% YoY excluding SBC), indicating continued investment in L4 and L2+ systems.
And with US$764M in liquidity, plus proceeds from its Hong Kong IPO, WeRide ended Q3 with a strong cash position.
Source: WeRide Investor Presentation
WeRide’s operational footprint
WeRide entered the quarter with one of the broadest global footprints in the industry. The company is now active across 11 countries and more than 30 cities, and it has become the first AV firm globally to secure fully driverless permits in eight different countries. This is not a symbolic statistic. It reflects a multi-product deployment strategy that spans robotaxis, robobuses, and robosweepers across markets as diverse as China, the UAE, Belgium, Singapore, and Switzerland. Few AV players operate across so many use cases and regulatory regimes simultaneously, and this geographic diversification has become the backbone of WeRide’s global positioning.
Source: WeRide Investor Presentation
Sitting underneath this footprint is a fleet that has grown into one of the largest L4 deployments worldwide. WeRide now operates over 1,600 autonomous vehicles.
The fleet number, however, becomes more meaningful when broken down. Only about 750 of these vehicles are robotaxis. For comparison, Waymo currently operates roughly 2,500 robotaxis. WeRide’s footprint is broader, but its robotaxi concentration in any single geography is lower, which is why its international scaling is wide rather than deep.
That said, WeRide’s ability to operate multiple product classes—robotaxis, robobuses, robosweepers—gives it a multi-revenue-stream structure that differentiates it from pure-play robotaxi competitors.
Nowhere is this more visible than in the Middle East, which has emerged as WeRide’s most strategically important international region. The company’s business model differs sharply between China and overseas markets:
In China, WeRide owns and operates most of the fleet through its WeRide Go platform. Abroad, it pursues a platform-partnership model, selling vehicles and licensing its virtual driver while ride-hailing players like Uber provide demand aggregation and fleet operations.
So WeRide’s international revenue streams consist of :
1.) revenue share from the ride fare
2.) annual licensing fee
3.) sale of the vehicle
This global strategy crystallized in Abu Dhabi, which became the centerpiece of WeRide’s Q3 narrative. After securing the first-ever city-level fully driverless commercial permit outside the United States, the company announced just days after its earnings release that the service had officially launched in a fully driverless configuration on the Uber platform.
The initial deployment focuses on Yas Island, covering roughly 12 square miles, where riders requesting UberX or Uber Comfort may be matched with a WeRide GXR operating without a safety operator. The fleet currently includes around 100 vehicles in the Middle East, with Abu Dhabi as the anchor market, and WeRide plans to scale that number to 500 vehicles in 2026.
Importantly, Abu Dhabi is not an isolated success. Riyadh and Dubai are also proving to be strong markets for WeRide. Riyadh is now the first city in Saudi Arabia to host publicly available robotaxi rides, again through Uber.
Dubai granted a self-driving trial permit earlier this year, with supervised Uber operations starting in 2025 and full driverless services expected in 2026. Together, Abu Dhabi, Riyadh, and Dubai form a regional corridor where WeRide has both regulatory traction and deep platform partnerships. No other Chinese player has achieved comparable breadth or speed across the Gulf region
What makes Abu Dhabi particularly strategic is that WeRide is now explicitly treating it as a profitability template. On the earnings call, management repeatedly referred to this as the “Abu Dhabi Model”. The logic is straightforward: use China’s cost base, hardware experience, and engineering processes to achieve breakeven in a high-fare Middle Eastern city that has fast-moving regulators and strong platform partners.
To reach positive unit economics, WeRide outlined a clear operational roadmap. The company wants to increase the human-to-vehicle ratio from roughly 1:3 today to 1:10, expand its operational design domain from 50% of the city core to full-city coverage, extend service hours from 12 hours to 24/7, and raise trips per vehicle from 12–20 today to 25–30.
Source: WeRide Investor Presentation
The company now intends to copy this model into Dubai, Riyadh, Singapore (via Grab), and eventually into selected European markets such as Switzerland, Belgium, and France.
Despite the global tilt, WeRide continues to strengthen its home market presence. China remains the company’s highest-density operating environment and plays a critical role in long-term profitability. Han was candid about what determines Chinese unit economic
“Profitability in Tier-1 cities can be achieved with three elements: a city-level driverless permit, average daily orders in the high double digits, and a relatively healthy fare level.”
WeRide’s performance in Guangzhou and Beijing shows clear progress. Vehicles there complete 23–25 daily trips, and the new PU/DO-free feature has significantly improved utilization and user experience. Tier-1 Chinese cities will remain essential for large-scale demand density, operational learning, and data feedback loops.
Finally, WeRide is no longer only an L4 company. With the launch of WePilot 3.0, an end-to-end L2+ system developed with Bosch, WeRide is following the same dual-architecture logic we are seeing across the industry. The idea is to create a shared data and model ecosystem where:
mass-market L2+ vehicles produce broad-distribution driving data to train world models,
while L4 robotaxis provide high-fidelity edge-case data and redundancy-heavy scenarios.
Han described this as a “double flywheel”: L2+ improves L4, and L4 improves L2+. The system will roll out on Chery EXEED models, with OTA upgrades for existing owners, and has already been nominated by GAC for several passenger car programs. This positions WeRide to compete not only where AVs operate today, but where consumer vehicles will evolve over the next decade.
Source: WeRide Investor Presentation
Pony.ai Q3/2025 Earnings Analysis
Pony.ai’s Q3 results reflect a company entering its industrialization phase. Revenue reached US$25.4 million, up 72% year over year, driven by strong robotaxi growth, accelerated licensing revenues, and steady contributions from its robotruck business. Gross profit improved meaningfully to 18.4%, more than doubling from last year, though the company still reported a US$61.6 million net loss, largely due to heavy R&D spending tied to its Gen-7 platform and world-model development.
Liquidity stood at US$587.7 million at quarter-end, but this figure does not reflect the company’s Hong Kong IPO, adding more than US$800 million in fresh capital. Combined, these numbers show a company with enough financial runway to sustain aggressive expansion and absorb the high upfront cost structure of scaling L4 autonomy.
The operational story began at home. Pony.ai spent most of the call emphasizing its progress in China’s tier-one cities, where it achieved citywide unit-economic breakeven in Guangzhou shortly after deploying its Gen-7 robotaxis commercially.
Management gave a rare level of detail behind the achievement: roughly 23 trips per vehicle per day, supported by growing demand density, shorter wait times, and improved pricing power. The company credits this to a reinforcing cycle. More vehicles reduce wait times, shorter wait times increase user adoption, higher adoption raises utilization, and higher utilization improves pricing and route efficiency.
This domestic momentum has been supported by a steady increase in fleet size. Pony.ai now operates 960 robotaxis, with more than 600 Gen-7 vehicles already delivered this year. Management told investors it will exceed 1,000 vehicles by year-end and is targeting more than 3,000 vehicles in 2026.
What makes this significant is that Pony is not just deploying vehicles faster; it is manufacturing them faster. Gen-7 production is now running ahead of schedule, and the company has already achieved a 70 percent reduction in BOM cost versus the prior generation, with another 20 percent reduction planned for 2026. The partnership with Toyota’s JV Zhuifeng allows Pony to combine automotive-grade hardware with a mass-production mindset, creating a cost structure that competitors will struggle to match unless they invest similar capital into vertically integrated manufacturing.
This week, Pony also renewed its partnership with Sunlight Mobility in China. Sunlight is one of Pony’s largest fleet-funding and fleet-operations partners and plays a central role in Pony’s asset-light model. The renewed agreement covers a larger purchase pipeline of Gen-7 robotaxis, expanded operational support, and deeper integration of Pony’s virtual driver into Sunlight’s fleet-management systems. In practical terms, this means Sunlight continues to buy, own, and operate the vehicles while Pony supplies the autonomous driving software and remote-assist capabilities.
This renewed partnership matters because it reinforces Pony’s capital-efficient scaling strategy inside China. Sunlight takes on the CapEx burden of fleet expansion, which allows Pony to grow robotaxi deployments without overextending its balance sheet. It also helps solve a structural bottleneck: the need for large, professionally managed fleets to reach city-level density in Tier-1 markets.
By deepening this relationship, Pony secures a predictable buyer for its Gen-7 vehicles. Just as importantly, it stabilises production volumes for the Gen-7 program, helping Pony lock in manufacturing efficiencies and maintain predictable output as it ramps toward its 2026 fleet targets.
With domestic economics stabilizing, Pony.ai has accelerated its international expansion. The company now operates across eight countries, including Singapore, South Korea, the UAE, Qatar, Luxembourg, and the United States, building a network of partnerships that mirrors WeRide’s global-first approach but executed at a different pace.
A major development this week was Pony’s announcement that it will expand into Europe through a partnership with Bolt. Bolt wants to integrate AVs into its network as a competitive differentiator, and Pony provides the L4 stack that can scale across multiple European markets without requiring platform rewiring. This adds Europe to Pony’s international roadmap at the exact moment when European cities are opening to supervised and driverless pilots
Pony’s presence in the Middle East is also strengthening. The company is already testing in Qatar, expanding in the UAE, and preparing additional deployments that follow regulatory pathways similar to those pursued by WeRide. South Korea remains another market, where Pony secured a nationwide testing permit that allows it to operate across multiple autonomous zones. The company’s strategy in these markets is consistent: bring the Gen-7 cost structure, localize the ODD and compliance layers, and rely on local ride-hailing partners for consumer access and demand aggregation.
How Pony.ai and WeRide Reflect China’s Global AV Strategy
WeRide operates vertically integrated in China and uses an asset-light structure overseas, enabling it to adapt to regulatory expectations while maintaining control in its core market. Pony follows an asset-light playbook in China and abroad, building a network of fleet partners that fund vehicles while Pony scales its virtual driver. Both approaches have strengths. WeRide’s model gives it deeper operational control in strategically important cities, while Pony’s model allows for faster capital-efficient international expansion.
The strategic outlook is clear. China now leads the world in AV deployment velocity. WeRide is demonstrating how multi-product operations and international regulatory alignment can turn global footholds into commercial networks. Pony is proving that an asset-light model can accelerate entry into markets like Europe, where platform integrations such as Bolt provide immediate scale. The United States still leads in technical maturity through Waymo, but Chinese companies are scaling geographically at a speed unmatched by Western competitors. The next phase of the AV race will be defined not by who has the best autonomy stack, but by who can put the most autonomous vehicles on the road, in the most cities, with sustainable economics. The companies that convert capital into fleets fastest will shape the global market in 2026 and beyond.
🔗 WeRide / WeRide (2) / Bloomberg / Bloomberg (2) / Bloomberg (3) / WeRide Earnings Call Transcript / The Verge / TechCrunch / TechCrunch (2) / Asia Nikkei / Road to Autonomy / Pony.ai / CNEVPost / PR Newswire / Pony.ai Earnings Call Transcript
Enjoying the AV Market Strategist?
If you find value in these deep-dives and enjoy the updates on autonomy each week — you can support my work.
Your contribution helps me access better data sources, industry reports, and research tools to keep every issue sharper and more insightful.
The newsletter will always remain free to read, whether you choose to support or not — but every bit of backing helps me make it even better.
💡 Quick Takes
100 new self-driving cars to roll out for broader testing in South Korea
Korea will deploy 100+ autonomous cars in a single “AV city” and has already logged 13 million km across 132 vehicles and 47 test zones. The measure is part of an initiative announced by the Ministry of Land, Infrastructure and Transport and other related ministries to boost the country’s global competitiveness in the industry.
Zagreb set for robotaxi launch in 2026
Project 3 Mobility is running 60 verification prototypes in Zagreb, aiming for a commercial robotaxi launch in spring 2026, backed by €89.7m National Recovery and Resilience Plan funding. “Zagreb is the first step, and then from Zagreb to the whole world. I hope we can all be a little more optimistic and trust each other a little more,” Rimac said.
Tier IV and Horiba Mira collaborate on UK autonomous driving tech
Tier IV is bringing its Minibus 2.0, sensor stack and software into Horiba Mira’s UK facilities, starting with an autonomous bus on the proving ground and targeting future Level 4 operations. Horiba Mira will provide access to its facilities for localisation and safety validation of the software and hardware, as well as engineering support to ensure compliance with regulatory and operational requirements.
🔗 SMMT
Tesla aims to double Austin robotaxis
Musk says the robotaxi fleet in Austin will roughly double in December, with service also active in the Bay Area and a new ride-hailing permit in Arizona. All of this still uses safety monitors in the car.
🔗 Reuters
GreenMobility to deploy 2,000 autonomous car sharing vehicles in Denmark
Tensor and GreenMobility signed a non-binding LOI for up to 2,000 Level 4 Robocars, starting in Copenhagen. Vehicles would be owned by GreenMobility, with an option for private Tensor customers to make their cars available to the fleet when idle. Tensor is working with GreenMobility to engage Danish authorities including the Transport Ministry and Road Authorities. Discussions have focused on test frameworks, safety requirements and the path toward commercial approval.
📚 Worth Reading/Listening
Harry Campbell: Uber, Taxis, and the Shift to AVs (Hansu Kim - Flywheel CEO)
📊 Weekly Performance
Note: Stock performance data as of November 30, 2025. Past performance does not indicate future returns.
Thanks for reading!
If you found value in this newsletter, please consider sharing it with a friend or colleague who might benefit from these insights.
And if you haven’t already, subscribe to stay updated on the latest developments in the autonomous driving industry.
Disclaimer: This newsletter is for informational purposes only and should not be considered investment advice. All opinions are my own.
thanks for the analysis ;)
I really love these analyses. Daniel, do you see one approach ultimately outcompeting the other for global scale, or will the market sustain both models based on regional partnerships and capital constraints?