Xpeng Unveils Ambitious AI-Powered Future: Robotaxis, Humanoid Robots, and Flying Cars Set for 2026 Launch

Chinese EV maker Xpeng to launch robotaxis, humanoid robots with self-developed AI chips

Guangzhou, CHINA -- Chinese electric car company Xpeng plans to launch robotaxis next year after previously claiming it wouldn't be a real business in the near future and took the wraps off of its latest humanoid robot model. Xpeng's technology push mirrors one of its key rivals Tesla, as the Guangzhou, China-headquartered company looks to position itself as more than just an electric car firm. The automaker announced on Wednesday as part of its "AI Day" that it is launching three robotaxi models. The vehicles will use four of Xpeng's self-developed "Turing" AI chips. Xpeng claims the chips represent the combined highest in-car computing power in the world, at 3,000 TOPS, an industry measure. The semiconductors power Xpeng's "vision-language-action (VLA)" model, now in its second iteration. This type of AI models take into account inputs like visual cues that can help with applications like driverless cars or robotics. Alibaba announced Wednesday that it is partnering with Xpeng on robotaxis through the e-commerce company's digital mapping subsidiary AutoNavi and Amaps app, which also includes a ride-hailing portal. The Xpeng robotaxi includes an external display of speed and other information on the vehicle's sun visors. Xpeng said it plans to start testing robotaxis in Guangzhou and other Chinese cities next year. Co-president Brian Gu told CNBC last week that robotaxis will "ultimately be a global phenomenon" but that it would take time to get there, especially given regulation. Back in April 2024, he cautioned that self-driving taxis wouldn't become a significant business for at least five years. During a group interview with reporters on Wednesday, Gu addressed his change in tone from last year toward robotaxis. "The tech is happening faster than we anticipated," Gu said. He noted that the AI developments and the significant increase in computing power "give us the confidence we are near the inflection point" for robotaxis. Xpeng's strategy for robotaxis is to make two categories of cars: one for commercial self-driving shared vehicles, and another for fully autonomous personal cars that may be only shared among family members.

Chinese EV-maker Xpeng reveals plans to have its flying cars on the road - and in the skies - by the end of 2026

You may not have heard the name before, but if you're at all interested in 21st century transport then ambitious Chinese EV-maker Xpeng should definitely be on your radar - pun intended. Far from being content playing in just the automotive space, it's in the process of branching out into the more futuristic fields of electric aircraft and robotics. The company holds an annual product and technology showcase in Guangzhou, dubbed AI Day, which this year took place on 5 November under the general theme of 'Emergence'. A big focus was on its development of AI systems designed to interact directly with the physical world - in cars, robots and aircraft, for example. The core announcement was VLA 2.0 -- a vision-to-action AI model that turns camera input straight into control output, which Xpeng claims allows for faster responses, especially when compared to the language-in-the-loop reasoning seen in most competing AI. At the event, Xpeng also teased three new robotaxis, demoed a humanoid robot and previewed a tiltrotor hybrid aircraft. We also got updates on previous announcements, and confirmation that the Land Aircraft Carrier we loved at CES is nearing mass production. In Australia and the UK, Xpeng recently launched the G6 SUV EV, while the larger G9 SUV is coming soon. Xpeng hasn't yet entered the US EV market, so most people will probably know the brand for its flying car projects and CES demos. Xpeng expects small scale aircraft to become a major new industry in China, and its flying-vehicle arm, Aridge, is building two options specifically for this market. The Land Aircraft Carrier is a mobile, self contained system designed for short flights with two people, while the A868 full tiltrotor hybrid aircraft is aimed at efficient multi-passenger travel over longer ranges - so basically, a flying car. Together, they are designed to create a futuristic transport system that can handle everything from business use to tourism. Unveiled at AI Day 2025, the A868 uses a tiltrotor setup on Xpeng's "Kunpeng Super Extended-Range Architecture" with a self-developed "aviation-grade hybrid-electric core". To explain, the "Kunpeng Super Extended-Range Architecture" means the A868 operates like a hybrid car, with an internal combustion engine and generator that charges the battery and supplies power to the electric motors. Importantly, this setup helps reduce the range and turnaround limits of purely electric aircraft. The A868 is designed as a 6-seat aircraft with about 500 km (310 miles) of range and up to 360 km/h (224 mph) cruise speed, with the program now in what the company calls a critical stage of flight verification. Xpeng also highlighted the practicality of the design, and says the A868 is intended to lift off and land from compact sites instead of relying on runways, and that it can use takeoff and landing areas as small as half a basketball court. This means it can target markets currently served by helicopters, and use existing infrastructure such as helipads. The Land Aircraft Carrier, first announced in 2023, is not far off mass production and Xpeng says it's received over 7,000 global orders for the $280,000 / £214,000 / AU$430,000 vehicle. The Carrier pairs a six-wheel hybrid-electric carrier with a small two-person electric copter. The flying air module has six lift motors and docks on the carrier for terrestrial transport and fast charging, with up to six flights possible per outing. The copter has a range of 20 km and flight times up to 20 minutes - enough to suit short scenic hops or point-to-point business use. Xpeng says the cockpit controls are designed to be easy to learn, with a single stick that handles the main control. The vehicle has integrated backups across its power, batteries, flight control and communications systems, so one can cover another if something goes wrong. The flying module also has a safety mode that can still bring it down for a controlled landing even if two of its six motors (provided they're on opposite corners) stop working. To support production scale, Aridge's dedicated plant began trial production on the 3rd of November, with the first Land Aircraft Carrier officially rolling off the line. Xpeng describes it as the first flying-car factory to use a modern automotive-style assembly line, targeting 5,000 units per year at first, but then scaling up to 10,000 at full capacity -- that's one aircraft every 30 minutes. The announced commercialization plans start with a tourism route in China's northwest -- a low-altitude self-driving tourism route in 2026, marketed as a "self-driving flight" experience that adds scenic flying to established cultural itineraries. While flying cars often feel like a sci-fi dream - something that's perpetually off in the far future - Xpeng's approach with the A868 and the Land Aircraft Carrier is quite practical and realistic, even if the timelines still seem quite ambitious. Based on the number of global orders it's received though, a hybrid tiltrotor for longer business-class hops (where speed and range matter), and a simpler flight experience platform (tuned for tourism and shorter flights) seem to have broad appeal. The underlying thread in practically all of Xpeng's vehicles and robots is VLA 2.0, an AI model that uses 'vision-to-action'. Instead of converting camera input into written language and then using that to decide which actions to take, it goes straight from pixels to outputs to control driving or robotic manipulation. For cars, Xpeng says it trained the AI on about 100 million videos and now runs a distilled model with billions of parameters; comparatively, rivals generally use models with only tens of millions of parameters. Cars and mobile robots don't have endless power and cooling to supply data-center levels of compute, so the more efficient 'vision-to-action' is a solid choice for getting the system up and running now. If you're keen to know more about Xpeng's VLA 2.0 AI model, I've covered it in more depth on our sister site, TechRadar. While the VLA 2.0 AI system will be used in passenger cars for supervised self-driving, at AI Day 2025 Xpeng also teased us with an announcement about three new robotaxis it's building. They use the same VLA 2.0 AI model, but will operate in fully driverless autonomous mode in defined zones. This is equivalent to SAE (Society of Automotive Engineers) Level 4 automation, which has been achieved by companies such as Waymo, but only on comparatively sensor-packed robot taxis. Xpeng's Robotaxi platform keeps the company's pure vision approach and hardware capable of up to 3,000 TOPS (Tera Operations Per Second, a common way to measure performance of AI hardware) of processing power. Xpeng says it will use what it learns from passenger cars to fast-track the validation process of the Robotaxis, with a planned pilot program and then mass-production goal of 2026. While this is another ambitious timeline, using the same core hardware and software as retail cars would seem to be a sensible choice that should help lower costs, improve testing scale and get Robotaxis operating sooner. Iron is Xpeng's next-generation humanoid built with a humanlike frame: bionic spine, flexible skin and 82 separate ways (known as 'degrees of freedom') that its joints can move across the neck, arms, hands, torso and legs. The hands offer more joints than many rivals, and aim for fine manipulation so the robot can slot into existing jobs currently done by humans without new tooling. Under the shell, three AI models work together. VLA (vision-language-action), VLT (vision-language-task) and VLM (vision-language model) handle interaction, walking and task control. Processing stays on the robot using hardware rated at 2,250 TOPS, keeping latency low and private data local. Full specs are still to come, but Xpeng is aiming for large-scale mass production by late 2026, and suggested that jobs the bot could perform early on include tasks like guided tours, retail assistance and back-office logistics.

Your own private aircraft carrier? Chinese EV-maker Xpeng's annual 2025 showcase just gave us a tantalizing glimpse of the future of transportation

If the brand name Xpeng is unfamiliar, that may soon be about to change. The major Chinese EV manufacturer has its sights set on becoming a leader in all kinds of future tech, from robots to electric aircraft and beyond. In terms of cars, it sells the G6 and G9 SUVs, the P7 sedan and the X9 MPV, plus the MONA M03 in China. In the UK and Australia, Xpeng is well known for the recent launch of the G6, with the G9 coming soon. In the US, you can't actually buy the company's EVs yet, so it's perhaps best known for flying cars and CES demos. AI Day is Xpeng's annual showcase, and it recently took place in Guangzhou, with 2025 themed around the concept of Emergence, with a large focus on new capabilities for the companies 'physical AI' systems - ones designed to interact with the physical world - across cars, robots and aircraft. The main headliner was VLA 2.0, Xpeng's vision-to-action model that introduces new capabilities for efficiently turning camera vision to control output. During the event, the brand announced the new VLA 2.0 model, teased a mass production robotaxi, demoed a next-generation humanoid robot and showed off the upcoming tiltrotor hybrid flying car. We also got loads of updates on previously announced products, plus new insight into how Xpeng plans to build an AI ecosystem with partners, and even make its latest AI model open source. Xpeng announced a new tiltrotor aircraft called the A868 and gave us loads of new information on the Land Aircraft Carrier hybrid vehicle with launchable copter, including the news it's near to mass production. Watch the video below to see how this innovative system works and to read more, check out my Xpeng coverage on our sister publication, Tom's Guide. Xpeng's VLA 2.0 model takes a vision-to-action approach to processing visual data, so instead of translating what its cameras see into text and then turning that into control output, the model goes straight from pixels to driving or manipulation. That contrasts with the 'vision language action' path that is used, for example, in Google's recent robotics work and autonomous driving research. Vision language action has an AI model that first describes a scene in natural language, then decides what to do from that description. This approach is flexible and can fold in extra information, such as written instructions or web context, but it adds extra processing, power draw and latency that current hardware and batteries struggle to supply in mobile uses, like vehicles and robots. Vision to action is more akin to fast automatic reflexes, and Xpeng uses variants of the VLA 2.0 model across cars, aircraft and robots - each with its own specific training. For cars, Xpeng says it trained the AI on about 100 million videos, and runs a distilled model with a parameter count in the billions. In comparison, Xpeng has said rivals are using models with parameter counts in the tens of millions. Making this system work reliably on the road still needs serious computing power and the effective on-device compute is quoted at 2,250 TOPS (Tera Operations Per Second, a common way to measure performance of AI hardware) in consumer vehicles and up to 3,000 TOPS in the Robotaxi and and the Iron robot. In comparison, Tesla's current Hardware 3 platform has 144 TOPS, and unofficial estimates for the not-yet-released Hardware 4 range from roughly 430 to 1,150 TOPS. Given the limitations of current energy and thermal budgets inside cars and mobile robots, vision-to-action is a pragmatic choice. It sacrifices some flexibility compared with language-in-the-loop systems, but can deliver big payoffs in efficiency, responsiveness and deployability at scale. Plus, Xpeng's approach of throwing plenty of computing power at the problem shows it's serious about making it work now, rather than trying to optimize for more efficient models or hardware in the future. While we await the consumer rollout of VLA 2.0 and real-world feedback, Xpeng is confident enough in its design to plan on VLA 2.0 underpinning and helping create a wider physical ecosystem that other manufacturers can build on, and plans to make VLA 2.0 open-source to business partners. VLA 2.0 will run on Ultra vehicles - Xpeng's 2025 top-spec trims with the full sensor suite and compute platform already built in. Using VLA 2.0, Ultra vehicles get what Xpeng calls XNGP - short for Navigation Guided Pilot. For now, in passenger cars, this is equivalent to SAE (Society of Automotive Engineers) Level 2 driver assistance, where the driver needs to stay attentive and supervise the AI control. The system is designed to handle most driving situations, including tight streets, parked-car pinch points, complex merges, roundabouts and construction zones. Lane choice and gap selection are part of the training, so Xpeng says it can ease into traffic, negotiate with other drivers and recover if something unexpected happens. It also supports parking and low-speed maneuvers that use curb detection and path planning to handle tight spots, and avoid any bumps or scrapes. Rollout of VLA 2.0 will reach new models and existing Ultra cars through over-the-air updates, beginning with a small early-access test in China, then wider release once safety goals are met. Xpeng highlights that their approach is all about low latency, predictable behavior and road skills rather than piling on more features. For example, VLA 2.0 is built to work with basic mapping (or none at all), to better handle road changes between map updates. In the driver assist landscape, Xpeng and Tesla are the main companies pushing to use camera vision by itself (rather than adding sensors like LIDAR) for consumer cars. Older Xpeng vehicles included additional sensors, so this is a bold approach with a lot of advantages, but also risk, and we've seen Tesla struggle to take camera-only vision to higher levels of autonomy. Based on demo videos, Xpeng's high compute strategy looks very promising though, and having the same system underpin multiple platforms should help it scale faster. Xpeng also announced a partnership with Volkswagen, which has confirmed it plans to use Xpeng's technology in its China-market vehicles. There's also an existing Xpeng and VW partnership for two jointly developed vehicles that will be built and sold in China. At AI Day 2025, Xpeng also announced three upcoming Robotaxi models using the same VLA 2.0 model, but here they were operating in fully driverless autonomous mode in defined zones. This is equivalent to SAE Level 4, like what has been achieved by companies such as Waymo with their sensor-packed vehicles. Xpeng says it'll use the passenger-car learnings to fast track the validation process of the Robotaxis, with a planned pilot program and then mass production goal of 2026. Passenger cars are also expected to get a fully autonomous mode trim that uses the same Robotaxi hardware on select models. Iron, Xpeng's next-gen humanoid robot, strongly leans into anthropomorphism and has features like a bionic spine, flexible skin and 82 degrees of freedom (82 independently controllable movement axes). There's also a big emphasis on hands: Xpeng highlights higher-degree-of-freedom hands than many rivals, aiming for fine manipulation to make them more useful in everyday tasks and a better fit for tasks currently done by humans without retooling. Iron uses three AI models in concert: VLA (vision-language-action), VLT (vision-language-task) and VLM (vision-language model) for conversation, walking and interaction. All computation runs on the robot itself thanks to powerful hardware delivering 3,000 TOPS of processing. This is the same core hardware setup as Xpeng's vehicles, and again shows the company is willing to use a lot of computing resources to make it work. Compared to competitor robots like Tesla Optimus and the Figure 01, Iron's bet is on dexterity and on-device processing. This matters because object manipulation is a key bottleneck for robots doing real work today, and having everything run on the robot should both reduce latency and improve privacy. Detailed specs of Iron haven't been released yet, but Xpeng is targeting large-scale mass production by the end of 2026. By using the same underlying systems across robotics and vehicles, Xpeng has a big advantage in scaling mass production of core hardware and software. The company suggests initial use cases could include jobs like giving tours, performing retail tasks and back-office logistics.

Xpeng AI Day: new AI model powering robots, robotaxis, and flying cars

Today, Xpeng held its AI Day 2025 at its new headquarters in Guangzhou. The automaker is positioning itself as an AI company with what it calls new "AI-defined applications" powered by the new Xpeng VLA 2.0 vision-centered model. This model is going to power the autonomous features in Xpeng's current EV lineup, the self-driving system in its newly announced robotaxis, humanoid robots, and "flying cars." At the event, CEO Xiaopeng He focused a lot on what he calls "Physical AI", which he describes as the merging of AI into physical products, aka robots. This is similar to how Tesla CEO Elon Musk has been describing Tesla's cars as "robots on wheels," and it is now also working on humanoid robots. Xiaopeng has never hidden that he has drawn a lot of inspiration from Tesla. He even admitted having started Xpeng after Tesla "opened up" its patents. Much like Tesla's transition to an "AI and robotics company", Xpeng has undertaken a similar transition in recent years, and it is now culminating in these new products announced today. Xpeng's new VLA operating system, Xpeng VLA 2.0, is the basis of all other products announced today. Unlike traditional "Vision-Language-Action" architectures that translate visual inputs into language before generating actions, the new VLA uses a direct "Vision-Implicit Token-Action" path -- removing the language bottleneck entirely. The result is an AI that can interpret and respond to physical reality faster and more intuitively, almost like human reflexes. To run such a massive model on production cars, Xpeng reengineered its stack for its Turing AI chips, achieving real-time performance with 2,250 TOPS of compute on its vehicles with its "Ultra" level of driver assist -- an order of magnitude more complex than typical automotive AI models. Ultra is Xpeng's top level of ADAS in consumer vehicles, but the company also announced a new level today: Robo: Robo is going to be for robotaxis. Xpeng announced that it will launch 3 new purpose-built electric vehicles for robotaxi services next year. There will be a 5-seater, a 6-seater, and a 7-seater. I've seen the prototypes driving around Xpeng's headquarters, but for now, these are the only images that Xpeng is sharing: These vehicles will feature four Turing AI chips delivering 3,000 TOPS and dual hardware for redundancy. Like its autonomous driving systems for consumer vehicles, Xpeng is betting on a vision-only approach powered by its new VLA model. Trial operation is expected to start in 2026. Xpeng is expecting to work with partners to deploy its robotaxi, with the first one already announced: map and navigation company Amap. When it comes to ADAS systems in its consumer vehicles, Xpeng sees its new VLA model expand capabilities quite a bit. They demoed their latest version of NGP (Navigation Guided Pilot), which is basically its Tesla FSD competitor. I did a test ride on the latest software and was impressed. It appears to be the equivalent of Tesla's FSD v14 with parking to parking capacity. For both its robotaxis and NGP software, Xpeng has talked about "open sourcing" the systems. Although "open" and "open source" were used alternatively, it sounded more like they plan to license their technology to other companies. Volkswagen has been announced as the first automaker to use Xpeng's new VLA-backed autonomous driving system. This is another area where you might think Xpeng is following Tesla, but it's the other way around in this case. Xpeng has been working on robots for 7 years. First, it started with quadruped robots and gradually evolved into several generations of humanoid robots. Brian Gu, Xpeng's vice chairman, noted that he believes the company began its humanoid effort ahead of Tesla. The previous generation of its IRON robot has been around for a while, and it even gave us a tour of its HQ in Guangzhou today: Today, Xpeng unveiled the latest generation of its IRON humanoid robot. A more advanced version with new humanoid bones, bionic muscles, and "full-coverage soft skin." The previous version was already impressive with relatively smooth walking that we were told was not pre-programmed nor tele-operated. But the new generation already features much smoother walking and a more production-ready look. Powered by three Turing AI chips (2,250 TOPS) and combining VLT + VLA + VLM cognitive models, IRON can see, move, and interact in real time -- effectively performing tasks that require physical dexterity and conversation simultaneously. It also features more dexterous hands with 22 additional degrees of freedom. The new version of the IRON uses an all-solid-state battery, though they didn't release any specs on capacity or autonomy. When it comes to commercialization, Xpeng is being way more conservative than competitors like Tesla and 1X, who are both predicting mass production and commercialization in 2026. Xpeng also aims to start mass production next year, but it only plans to use the robots in its own commercial operations, primarily as guides, receptionists, and even salespeople at its locations. Xiaopeng said the company tried to operate IRON in its own manufacturing operations for a year, doing what it believed was the easiest human task to replace: tightening screws with a drill on the assembly line. However, Xpeng concluded that it wouldn't be efficient at scale due to the robot not performing as well as employees, especially on a cost basis, due to high repair and replacement costs. Humanoid robots are way more complex and currently not as sturdy and efficient as industrial robots, which can lead to high repair costs. Xiaopeng believes that humanoid robots will one day be widely used in manufacturing operations and in homes for household chores, but he thinks this is further down the road than the competition predicts. He was reluctant to give a clear timeline, but when pressed, he said likely 3-5 years for industrial applications and 5-10 years to be safe and useful in homes at scale. Xpeng says it will ensure its robots follow the three laws of robotics and add a fourth law to ensure they also respect human privacy. Next year, you should expect to see IRON in Xpeng stores across the world. Xpeng is launching an IRON SDK for developers to help develop more applications. Baosteel, a giant steel production company based in China, will be an early partner. Finally, the latest "intelligent AI-powered" product that Xpeng highlighted at its AI Day is its latest lineup of "flying cars" under its ARIDGE brand. To be fair, "flying car" is a term used loosely. As the CEO pointed out himself, they are more accurately described as low-altitude eVTOL (electric vertical takeoff and landing) aircraft. The main new aircraft unveiled is the A868, an electric-hybrid, full-tilt rotorcraft. I saw the first prototype at Xpeng's HQ today: The A868 uses an aerospace-grade hybrid system based on Xpeng's "Kunpeng Super Range" platform, promising 500 km of range, 360 km/h top speed, and a six-seat layout targeting business travel. It is currently entering its flight test phase, according to Xiaopeng. Closer to production is Xpeng's 'Land Aircraft Carrier' - a modular electric-hybrid 6×6 vehicle that carries a small eVTOL. The aircraft is much smaller than the A868 as it literally fits inside a van, and it also has a much shorter range of 20 to 30 km. It is reportedly equipped with full safety redundancy and six dual duct propellers. The Xpeng Land Aircraft Carrier is entering pre-production, with over 7,000 global preorders and the first mass-production unit already off the line as of November 3. The company's ARIDGE execs are required to accumulate a minimum of 5,000 km of flight in the aircraft - a policy seemingly meant to create confidence in the product. Xiaopeng also says that he will personally fly in the first production unit. While individuals can buy the Land Aircraft Carrier with the manned drone, the company is also currently deploying "flying camps" across China to offer "tourism tours" with the electric aircraft. The first of such camp is in Dunjuang. Xpeng is aiming for 200 flying camps by the end of 2026. There are a lot of products that Xpeng unveiled that made me raise an eyebrow, especially with the "flying car" stuff, but I have to give credit where credit is due. There's some real and impressive technology at this company. The comparisons with Tesla are obvious, but at least, Xpeng is being way more realistic about timelines and capabilities and not selling you on products that don't exist yet. For example, Xpeng is investing heavily in humanoid robots, but it doesn't see near-term commercialization scenarios for home or factory robots. Nonetheless, it is designing the robot for mass production, and it aims to reach that goal by the end of 2026. The goal is that by then, AI capabilities will improve enough to make IRON useful in applications beyond the tour-guide/salesman applications it is first aiming for. The SDK should help with that. But Xiaopeng is not out there claiming that it is a "multi-trillion-dollar opportunity" and that it will start producing millions of robots within the next 2 years. He did say that he believes humanoid robots are a big opportunity that AI is enabling through new "real-world physical models", but there's still a lot of work to be done. During a press conference that followed the AI Day presentation, Xiaopeng and Gu were repeatedly ask to differentiate Xpeng and Tesla. They admitted there are many similarities between the two companies, but noted that Xpeng is more open than Tesla, partly out of necessity given its smaller scale. When asked specifically about why Xpeng's valuation was so much lower than Tesla even when adjusted for scale, Dr Gu was not shy about pointing that Tesla's valuation was based on potentially premature commercialization announcements while Xpeng prefers to wait until it has the technology before announcing commercialization. The "flying cars" might be the exception, but I think that's fair for the rest. Xpeng is not exactly announcing the commercialization of its humanoid robots beyond use in its own stores, and based on the demonstration we saw from a previous-generation IRON, it looks very plausible. Unlike demonstrations from competitors such as Tesla's Optimus and 1X's Neo, Xpeng claimed that its IRON tour guide demo was not pre-programmed (beyond the product speeches) and was not tele-operated. The IRON was navigating the showroom in real time using the same neural networks as its vehicles. Speaking of its vehicles, its NGP system, a Tesla FSD competitor, is offered standard and is clearly marketed as a level 2 driver assistance system that requires driver attention at all times. Xpeng also has an advanced driver monitoring system to prevent driver abuse, which has been a problem with Tesla's FSD and Autopilot - leading to the automaker currently being flooded with wrongful death lawsuits regarding crashes involving its systems. In contrast with Tesla, Xpeng is more conservative, not promising or prematurely selling additional capabilities to owners before they are ready. That said, Xpeng clearly has ambitions to reach level 4 autonomous driving in consumer vehicles, and it looks like it is neck-in-neck with Tesla FSD in performance right now. At least in China. And unlike Tesla, which has been trying to license its FSD technology to other automakers for years, Xpeng already has a taker in VW.