XPeng Motors: A car company that doesn’t really want to be one in the first place

[HONG KONG] Right before He Xiaopeng went on stage to launch the 2025 XPeng X9 car, he battled what he called a “miserable” cold.

“I slept for 30 out of the past 48 hours. I wanted to be fully energetic to meet all of you,” the co-founder of China’s XPeng Motors told a large crowd of reporters, investors and supporters at an event in Hong Kong on Apr 15.

That anecdote was a sign of just how important the X9 is to the ambitions of 11-year-old XPeng Motors and He himself. But the futuristic, seven-seat multi-purpose vehicle (MPV) might merely be a Trojan horse for XPeng’s killer weapon in the electric vehicle (EV) market’s next battlefront.

Founded in 2014, XPeng has sold more than 600,000 vehicles since inception, a milestone that serves to underline how the company is still very much a startup. In comparison, market leader BYD takes only two months to sell that many cars.

Still, with 60,158 vehicles delivered in the first quarter of 2025, XPeng is showing signs of building the kind of volume it needs for sustainable growth. It operates in over 30 countries and plans to double its overseas presence to 60 markets, with a goal in place for half its volume to eventually come from outside China.

At the unveiling – the brand’s first global car event – He positioned the X9 as the opening act in what he called a “new decade” for XPeng, one that will put artificial intelligence (AI) at the centre of the New York and Hong Kong-listed company’s strategy.

For two hours, the 47-year-old laid out a case for how AI will underpin everything XPeng does, using the new X9 as a prime example. Engineers updated 35 per cent of its components, but one change mattered more than the rest: a switch from Nvidia’s chips to XPeng’s self-developed Turing AI chip.

The 40-core processor can run large models with 30 billion parameters locally, meaning without help from the cloud. Mass production is scheduled to begin in the second quarter of 2025 in China, potentially giving XPeng a head start over rivals still reliant on external suppliers for high-performance automotive AI.

XPeng says the chip could give its cars enough self-driving ability to need just one intervention from a human driver every hundred kilometres (once regulations allow that level of autonomy).

AI will also make its cars more energy efficient by learning how drivers behave in different traffic conditions, then adapting the motor responses to suit, or make them more comfortable by using cameras to read the road ahead and tune the suspension for bumps before they hit them.

The same cameras will also make the AI multi-modal – if you are curious about the new Mercedes driving ahead of you, for example, you could simply ask your XPeng to identify it and give you the run-down on its performance figures.

In fact, by forming the technological backbone for a sprawling product ecosystem, the Turing chip could turn out to be XPeng’s real product offering.

“XPeng was never built to be a traditional automaker. We never want to become one,” He said. “What really drives us is creating boundary-pushing innovations, whether it’s intelligent driving, AI-powered cars or beyond.”

What’s meant by “beyond” has so far taken shape as flying cars and humanoid robots, both of them with Turing AI chips for brains.

The humanoid machines, which XPeng named Iron, are 178 cm tall and weigh 70 kg. They have 60 joints and camera vision and are already on trial in the company’s factories, where they work side-by-side with humans.

He expects the Turing chip to let an Iron robot figure out problems by itself, instead of mindlessly repeating tasks that programmers set for it.

“If a robot does not have the intelligence that it’s required to have, it is going to be useless,” he said bluntly.

Perhaps sensibly, AI isn’t intended to make XPeng’s flying car think for itself, but to make it easier to control. At the X9’s launch party, AeroHT – a startup that became an XPeng subsidiary in 2020 – showed off the Land Aircraft Carrier, a six-wheeler that comes with a two-seater drone.

The flying machine fits neatly into the vehicle, which also charges its batteries. The design solves two major headaches that come with owning a traditional helicopter, namely storage and refuelling, according to Tan Wang, who co-founded AeroHT.

AI removes another pain point by making it a breeze to control. Tan told reporters that helicopters are fiendishly hard to fly, but claims that the Turing chip does enough of the heavy lifting that anyone can learn to operate the XPeng in just three minutes. “In three hours, you would be a master,” he said.

The Land Aircraft Carrier might seem like a pie-in-the-sky idea, but He said XPeng is on track to start delivering them to customers in 2026.

The company says it has more than 4,000 pre-orders, at just under two million yuan (S$359,000) each. The target is to sell 10,000 a year, which seems almost outlandishly ambitious, given that no aviation company has ever managed annual sales of more than a thousand.

The opportunity is significant. The Chinese government elevated the development of the “low-altitude economy” to national priority status for the first time in its 2024 Government Work Report, giving the sector a significant boost. The Civil Aviation Administration of China forecasts that the market could reach 3.5 trillion yuan by 2035.

Despite that, He is measured in his expectations about how quickly his clutch of AI-powered machines could change the way we live, putting the timeline at “10 or 20 years” when asked how long it would take for robots to become ubiquitous.

“We will have a robot at home and outside we will have robot cars. Flying cars would also be a type of robot. But I think it may take decades of R&D to get us there,” He said. 

Yet, it’s far from certain if XPeng will survive long enough to participate in that future, let alone create it. Like many startups, it struggles with profitability. It has never run in the black, and in the fourth quarter of 2024 its net loss was 1.33 billion yuan, slightly better than the 1.35 billion yuan loss in the same period of 2023.

With margins improving and vehicle sales rising, management expects the company to start breaking even by the end of this year. But the founder sounds a cautious note about XPeng’s chances.

“I believe that the next decade will witness very brutal competition in China, and I believe that ultimately, we will have only five to seven carmakers left,” He said. “Right now, we are ranked seventh, so there’s a long way to go.”