WeRide Doubles Revenue as Overseas Robotaxi Fleet Surges and L2+/L3 ADAS Revenue Explodes 26x, But Path to Cash Flow Breakeven Still Three Years Out
Q2 2026 earnings call, August 12, 2026
WeRide's second-quarter results mark what CEO Tony Han called "an important inflection point," with total revenue nearly doubling year-over-year and jumping 103% sequentially to RMB 232 million. The headline numbers underscore a business transitioning from pilot-stage robotaxi deployments to actual commercial scale, particularly overseas, even as the company remains several years from profitability. CFO Jennifer Li reiterated a target of positive cash flow in a single quarter by 2028 and full-year breakeven in 2029, timelines that were not pulled forward despite the quarter's strength.
L2+/L3 ADAS Business Inflects From Design Wins to Mass Production
The most significant new data point this quarter is the acceleration of WeRide's L2+/L3 ADAS business, which the company also refers to as WRD 3.0. Revenue from this segment surged nearly 2,600% year-over-year and 219% quarter-over-quarter, with cumulative vehicle deliveries surpassing 30,000 units as of June 30. Management guided to more than 100,000 units by year-end and over 500,000 cumulative units by the end of 2027. This is a meaningful re-rating of the ADAS business from a nascent side project to what Li called an "increasingly important driver of our overall gross revenue growth."
Han was direct about how the company is achieving this with a comparatively small headcount, attributing it to WeRide's foundation models, Genesis and WIT. "With our very good AI infrastructure and foundation model, we can actually build the best ADAS solution in China and in the world with a team of 200 or 300 people," he said. He also disclosed a notable hardware efficiency claim: WeRide's onboard domain controller runs at roughly 200 TOPS versus competitors' 2,000 TOPS, yet the company still won six consecutive championships at China's Intelligent Driving Competition, which Han attributes to model distillation rather than raw compute.
Investors should treat the FSD comparison Han volunteered with some skepticism given the promotional framing, but it is nonetheless a new claim worth tracking: Han said WeRide's ADAS system is "on par with FSD" in Chinese urban scenarios and personally livestreamed a two-hour unassisted drive through Guangzhou's Zhujiang New Town CBD, calling it the first such demonstration by any autonomous driving company founder in a first-tier city.
Mercedes-Benz L3 Partnership Raises Questions on OEM Sourcing Strategy
WeRide disclosed a new L3 autonomous driving proof-of-concept program with Mercedes-Benz during the quarter, which management framed as validation from a top-tier global OEM. This drew the sharpest question of the call from LightShed's analyst, who noted that Mercedes is also an NVIDIA flagship partner running a full-stack autonomous program, and pressed management on what WeRide offers that NVIDIA's stack does not already cover. Han's response was notably evasive, citing contractual restrictions on specifics and instead pivoting to an analogy about DeepSeek outperforming better-funded competitors. This exchange is worth flagging for investors because it leaves unresolved whether the Mercedes relationship reflects genuine multi-sourcing economics, regulatory necessity in China, or a narrower scope than the "validation" framing suggests.
Overseas Robotaxi Economics Come Into Sharper Focus
Overseas revenue grew 164% year-over-year and roughly 170% quarter-over-quarter, now representing nearly 40% of group revenue. Management gave the clearest unit economics disclosure to date on this call, an important data point given prior ambiguity, with steady-state annualized technology service revenue per robotaxi estimated at $50,000, with Li providing a range of $40,000 to $60,000 with upside as fleet density builds in each city. Han emphasized that this figure is not comparable to ride-hailing gross fares, since WeRide operates a licensed "virtual driver" model, collecting recurring technology and knowledge-based fees from local partners rather than owning vehicles or fare revenue directly.
The overseas robotaxi fleet reportedly doubled to around 400 vehicles during the quarter, according to Li, a figure that conflicts with Han's earlier statement in prepared remarks that the total L4 fleet including robotaxi reached 3,400 units with robotaxi at "more than 1,800 vehicles," up 40% sequentially. This internal inconsistency in the transcript, whether a transcription error or a genuine discrepancy between domestic and overseas fleet figures, is worth clarifying with management, as the numbers cited for robotaxi fleet size vary meaningfully across the call.
New robotaxi commercial partnerships launched this quarter in Madrid, Zurich and Copenhagen, alongside continued expansion in Abu Dhabi, Dubai and Riyadh. WeRide now holds official autonomous driving licenses in eight countries and operates in twelve, a scale Li asserts is unmatched by any competitor. On the durability of platform partnerships with the likes of Uber and Grab, given Uber's history of rotating AV partners, including its recent wind-down of the Serve Robotics relationship, Li argued that WeRide's moat is not the platform relationship itself but the regulatory permit. "Everyone can be on the platform, but not everyone can get the regulatory permit," she said, adding that competitors with weaker safety records have faced regulator-imposed suspensions after accidents.
Margin Expansion and Operating Leverage Emerging, Though Losses Persist
Gross margin reached a record 37.5%, up 940 basis points year-over-year, driven by the mix shift toward higher-margin overseas L4 and L2+/L3 revenue. Total operating expenses grew just 9.2% year-over-year to RMB 533 million, well below revenue growth, which Li pointed to as early evidence of operating leverage. R&D spending rose 36% to RMB 434 million, tied to continued investment in the Genesis and WIT foundation models, while G&A actually declined on lower share-based compensation.
Despite these improvements, net loss narrowed only 1% year-over-year to RMB 401 million, and EBITDA loss narrowed 8.1% to RMB 335 million. The company holds RMB 5.4 billion in cash and liquid financial resources as of June 30, providing runway, but the modest pace of loss narrowing relative to the dramatic top-line growth suggests the path to the stated 2028-2029 cash flow targets will require sustained execution rather than a single quarter of momentum.
WIT Foundation Model Explained as Core Technology Moat
Management spent considerable time this call explaining the WIT model, launched during the quarter and named after philosopher Ludwig Wittgenstein. Han described it as a tool that decomposes video data into what the company calls "minimal fact units," atomic causal facts extracted from road footage, which then feed into the Genesis world model to generate synthetic long-tail driving scenarios. "Hallucination in digital AI may be entertaining, but it is fatal in physical AI," Han said, drawing a distinction between generative AI's tolerance for error and the zero-tolerance safety bar required for autonomous vehicles. Han also proposed what he called an industry qualification threshold, arguing that a company should not claim to be a "robotaxi company" unless it has operated a fleet of at least 100 driverless vehicles for six months without severe accidents, a pointed jab at OEMs and ADAS players publicly signaling robotaxi ambitions without comparable safety track records.