The robot is hard to ignore. Its unmistakably female form stands out with its nearly uncanny proportions and realistic, model-eqsue gait.
The 5’4”-ish IRON robot’s appearance and walk were so convincing to the public that journalists and influencers inside and out of China were convinced that rising automaker Xpeng hired a woman to wear a cosplay-esque morph suit and robotically do a supermodel walk across the stage of its new auditorium, like a more convincing version of Tesla’s first appearance with its Optimus robot.
Xpeng Tour 2025 Photo by: Kevin Williams/InsideEVs
But whether or not you believe Xpeng’s robot is fake—and from seeing the unit in person, it appears to be the real deal, at least when it comes to its ability to walk—the headline-making robot is important in a lot of ways.
Sure, Xpeng’s own development team can’t really give a good answer yet as to why they designed the robot to look the way it did. Additionally, Xpeng remains unclear about the commercial viability or path forward for the robot itself, aside from a vague promise to mass-produce the androids and sell them in 2026.
Xpeng Tour 2025
Photo by: Kevin Williams/InsideEVs
But the whole project speaks volumes about Xpeng, which used the recent AI Day event in Guangzhou to break into the global consciousness for the first time.
Xpeng’s robot project may be unclear, but like Tesla, it clearly wants to be more than just a car company. Yet, unlike other car companies turned tech giants, Xpeng’s path forward seems far more viable than what I’ve seen from other brands—especially Chinese ones that are flailing for profits, relevance and attention in a market that is turning into survival of the fittest.
After spending the better part of a week at its new Guangzhou world HQ, I think Xpeng could be the one to watch, both in and outside of China. It’s clearly using Tesla’s playbook—except Xpeng might be a faster reader, with better comprehension.
What Is Xpeng?
There are a lot of Chinese EV companies, so I’m not going to hold it against you if you’ve never heard of Xpeng. The brand calls itself a “tech startup,” although at 11 years old and more than 1 million vehicles produced around the globe, I think we’re a little well beyond the title of “startup.”
Xpeng Tour 2025
Photo by: Kevin Williams/InsideEVs
The brand sells a full line of tech-forward EVs and now plug-in hybrids and extended-range EV models as of this year. All of them fall into a sort of near-premium-ish space in China. It’s directly analogous to Tesla’s market position, with refinement and a focus on technical gadgetry rather than outright luxury.
Famously, a former Tesla engineer was accused of stealing the Autopilot source code and selling it to Xpeng, but Xpeng says it has respected the intellectual property of all parties. (Tesla and the engineer eventually settled the lawsuit in 2021.)
Recently, Xpeng has taken steps toward expanding outside of China with its line of moderately priced midsize and compact SUVs. Success has been modest in Europe, with Xpeng’s deliveries hitting around 25,000 units there this year.
Xpeng Mona M03. Photo by: Kevin Williams/InsideEVs
Then Xpeng’s roughly $18,000 Mona M03 sedan showed up. In a little over a year, Xpeng has sold 200,000 examples of that car, and has plans for the model to enter other markets outside of China.
Recently, it unveiled a “Max” version of the Mona M03, which adds Xpeng’s Level 2 assisted driving features, at the time not super common in such an affordable car.
Despite those wins, it’s hard to stand out in the very crowded Chinese automotive space. That’s why Xpeng is clearly looking beyond cars.
Xpeng’s Approach To Autonomy: AI, LIDAR And More
In China, semi-autonomous and self-driving technology is practically a prerequisite for any carmaker. No, really. It’s hard to think of a car brand in China that isn’t making its own foray into above-Level 3 autonomy—meaning hands-off and eyes-off driving. The only thing holding it back is government regulations.
Xpeng XGNP Photo by: Xpeng
BYD’s excellent “God’s Eye” system, a camera-based assisted driving system, is an option on every single China-market model, including the dirt-cheap Seagull. I’ve also experienced the systems in Nio and Onvo vehicles, which use either LIDAR (Nio) or camera-only (Onvo) autonomy systems. Or, Zeekr, which uses LIDAR and some Nvidia Drive Orin Chips to do semi-autonomous driving. Or Li Auto, which uses some NVIDIA Drive Orin Chips and some LIDAR hocus pocus to move down the road.
Point being, go to damn near any brand in the People’s Republic, and you’ll find the same few steps applied verbatim to every car brand. It’s kind of like that for the whole Chinese car industry, which is why some brands are really struggling to stand out in a crowded market of otherwise very competent entries in China’s new-energy vehicle (NEV) market.
Xpeng XGNP
Photo by: Xpeng
But now that I’ve been home for a little bit of time and had the space to ruminate on what I saw (and recover from a very nasty cold), I understand that Xpeng’s focus on AI feels like they’re taking a page right out of Tesla’s book. And if they play it right, perhaps it could lead to a similar level of success, and not just in China.
What Is Physical AI?
Xpeng’s World AI Day started off with an announcement, a “new invention,” it purported, called “Physical AI.”
Xpeng Tour 2025
Photo by: Kevin Williams/InsideEVs
On its face, this is more of a marketing or tone-setting term for the whole conference, rather than a true definition or new invention. Slogging through the marketing and hype jargon, Xpeng’s focus on“physical AI”—a term we’re hearing more and more these days—refers to how artificial intelligence can and will interact and move throughout the real world and physical space. Xpeng was clear that this is a self-made model, not weird, meant for funny or uncanny generative videos via an LLM, or a Tesla Grok-style chatbot that’s installed into the car.
Xpeng sees physical AI as the manifestation of its large data models, making sense of the world. The founder and CEO of Xpeng, He Xiaopeng, wants the self-driving tech to have similar reasoning and response to that of a human, and less like a machine.
Xpeng Humanoid Robot Concept Photo by: Patrick George
To them, this means that the AI model should try and reason like a human, or at least simulate that. To Xpeng, human logic is effectively translating “visual information” to action, whereas a typical AI model will take visual objects alongside language to make sense of the world. In layman’s terms, AI models are constantly taking in data, whether that’s actual text, or information from via camera or LIDAR. It’s then organized and interpreted by the model itself. Keep in mind that this all happens while the car is driving, so new information is constantly being absorbed and processed.
Now, add in the fact that the car (or robot) has to navigate a world that is changing in real time; it’s a unique challenge for an AI model, especially compared to other AI models that don’t have to deal with changing information in real time.
Your brain does this all of the time, so it deserves more credit than you think. Getting a machine to do the same is very hard. To get something that’s usable, Xpeng says they essentially had to pare back how much data and what kind of data is interpreted by the AI model itself.
To create a more human-like autonomous experience, Xpeng says that they’ll need to streamline the amount of data coming and being processed. Sure, there’s a lot of data that Xpeng gets from what its cars see, but the ability to process it all isn’t realistic from a technical standpoint, or helpful to the end result.
Xpeng Tour 2025 Photo by: Kevin Williams/InsideEVs
Xpeng calls it VLA 2.0, short for Vision, Language, and Action—although with the 2.0 version, the Vision and Language aspects have been combined.
Essentially, in earlier VLA models, Xpeng would process both vision (objects), along with everything else in the background—most notably, language and text. This overabundance of data is why Xpeng thinks other AI models move too slowly; according to the folks I spoke with, the idea is that, most humans don’t need to read every single piece of text to interpret what action to take. Thus, AI should be the same.
Xpeng says it has collected some 100 million hours’ worth of driving data from driving videos. From there, it has been able to simulate “long tail” theoretical circumstance simulations, which help train the model how to think and react in the real world.
Xpeng Tour 2025
Photo by: Kevin Williams/InsideEVs
Does It Actually Work?
Xpeng offered several ridealongs in its entry-level Tesla Model 3 fighter, the Mona M03 Max, as well as its latest P7 sport sedan. We got some point-to-point semi-autonomous Level 2-ish urban navigation, meaning our eyes were on the road and our hands were still on the wheel.
Sure, the Mona and P7 were able to navigate traffic on public roads around Xpeng’s Guangzhou campus. They could also move around Xpeng’s maze of an underground parking structure, negotiating obstacles and the number of Xpeng employees headed to the office and contractors here to put the finishing touches on Xpeng’s new headquarters.
The car was even able to maneuver in and out of a parking space all on its own, even choosing a new parking space when a contractor in a Maxus-branded van unwittingly blocked the car’s original plan of action.
Xpeng European EV production at Magna Steyr’s facility in Graz, Austria
Photo by: Xpeng
However, this wasn’t VLA 2.0, nor was it particularly unique when it comes to Chinese market autonomous driving technology. Earlier this year, mere days after the Shanghai Auto Show, I watched a Zeekr 007 do something similar at Zeekr’s Hangzhou-area headquarters, except the car did it entirely driverless—no humans inside.
Xpeng P7 (2026)
Photo by: Xpeng
Then, just to showboat on over every westerner holding out their phones or DJI action cameras to report back home, it finished off the impressive demo by plugging itself into to a DC fast charger, which rocketed to nearly one megawatt worth of charging speed.
Hell, Nio and Onvo’s self-driving abilities were equally impressive as Xpeng’s, able to navigate nearly to the door, and go straight to one of its battery swapping stations. While Xpeng’s tech seems good, it’s not unique—nor was the software we tested part of its AI-forward announcement of VLA 2.0 we were told about the day before.
Yet, arguably, Xpeng’s software execution or proof-of-concept of what it currently can do isn’t meaningless when taken in the full context of its plans: world domination.
Xpeng Wants To Go Global
The VLA 2.0 model is kind of the linchpin for all of Xpeng’s future expansion plans. For starters, the company says that the VLA model will be ready to work outside of China with minimal work or training, including Europe, which is a big market push for the brand.
New Xpeng P7 2026
Photo by: Xpeng
This is a huge level up from other Chinese brands, which don’t always get the full suite of assisted driving features for several reasons, often relating to pricing as well as data privacy laws. BYD’s “God’s Eye” software has not debuted outside of China yet, while cars like the Zeekr 7x and Nio ES6 (called EL6 in Europe) lose their LIDAR-based autonomous driving features when sold in Europe.
Yet, Xpeng says its future self-driving AI model will be processed on the vehicle itself—and compliant with any data-sharing rules. Xpeng’s in-house processors, named the Turing Chip, are the backbone of Xpeng’s future push. Used in triplicate, the chips are meant to control all of Xpeng’s autonomous-driving future; forming the basis of its future Robotaxi business (set to launch in 2026), and the forthcoming “Robo” trim of its current crop of cars.
Xpeng Mona M03
Photo by: InsideEVs
Those are said to have robotaxi-like functionality, but via private car ownership. The brand says that it wants to have true Level 4 autonomous driving by next year. That’s quite a reach, considering Tesla’s camera-only Robotaxi is only just now getting off the ground in very specific geographic areas.
The same chip and AI model logic is said to underpin the IRON robot, although it’s not clear if any of that stuff was up and running for our lone proof-of-concept walking demo Xpeng showed during its conference.
That’s A Lot Of Promises. Can Xpeng Deliver?
Impressive as these ideas are, I’m still not entirely convinced that I’ve seen enough from Xpeng itself to confidently say that it has the sauce or not to pull it off. I don’t know if Xpeng’s new AI model works like executives and engineers say the way it does. A 2026 timeline for true Level 4 full autonomy and a robotaxi service based on cameras only feels ambitious. The concept of selling a car that can drive or run errands fully autonomously still feels like a pipe dream. Tesla promised that years ago, and it has yet to deliver.
Hell, Xpeng itself has said that it still doesn’t know what the path to commercial viability is for its IRON robot. Despite its “soft skin, bionic muscles and dexterous hands”, it straight up said it does not see the product as viable in factory settings, unlike Tesla’s plans for its Optimus robot.
Xpeng AeroHT CES Photos
Photo by: Xpeng
But there are things we do know, though. Xpeng’s current suite of assisted driving software works generally well. The cars that they’re in are stylish, well-finished, and well-priced. There’s a reason why it’s one of the fastest-growing brands in Europe.
Every Xpeng vehicle I got seat time with felt tight and high quality, unlike Tesla, a brand with sloppy initial quality and an old product line neglected by a founder resting on his laurels. Any sort of tech addition to Xpeng’s existing lineup only makes a strong lineup even stronger. Also, VLA 2.0 is good enough for Volkswagen China to be its first client. That has to mean something.
Still, it’s easy to go to any given Zeekr, BYD, Geely, Li Auto, Leapmotor, Deepal or whatever brand’s showroom and find something that feels technically impressive to use or drive. Like Xpeng, when it comes to a company’s bottom line, it’s clear that it’ll need to move past just building or selling cars if it ever expects to see big profits and cash coming in the door. This is something Tesla started planning for in the last few years, but it still hasn’t moved away from being a car business yet..
And, if Xpeng wants to survive what’s shaping up to be the great culling of Chinese brands on the horizon, then it must know that promises will only take it so far.
Contact the author: kevin.williams@insideevs.com
More Xpeng News We want your opinion!
What would you like to see on Insideevs.com?
– The InsideEVs team