Liu\u2019s comments clarify that the company spends $500 million a year on AI model training alone. This marks a significant expense for the company, which reported cash and cash equivalents, restricted cash, short-term investments, and time deposits of $6.26 billion as of March 31. \u00a0<\/p>\n
XPEV Vs. TSLA<\/p>\n
The remark came the day after Liu shared the stage at the CVPR 2026 conference in Denver with Tesla\u2019s Ashok Elluswamy and executives from Nvidia and Waymo<\/p>\n
Liu said the company\u2019s latest vision-language-action (VLA) models have already reached parity with\u2014or even surpassed\u2014Tesla\u2019s FSD V14 in early testing this year. FSD, or full self-driving, is the name of Tesla\u2019s self-driving technology.<\/p>\n
XPeng follows almost the same philosophy as Tesla in enabling vehicle self-driving. The system relies on vehicle camera data, which is used to train the system further and make better decisions for enabling vehicle autonomy. However, while Tesla has stayed clear of ultrasonic sensors and radars for its system, XPeng continues to use them sparingly.<\/p>\n
\u201cCamera is the best solution,\u201d Liu said. More expensive radar and lidar are reserved strictly for redundant active-safety systems such as automatic emergency braking\u2014creating an \u201corthogonal system\u201d that operates independently of the main AI driver.<\/p>\n
Unlike Tesla\u2019s Elon Musk, Liu is not hopeful that cameras alone would ever be sufficient.<\/p>\n
\u201cWe hope so, but to be honest, it\u2019s not possible. We all make mistakes. System also makes mistakes. Even you can reach 99.9999%, you still have the chance to make mistakes. Adding another layer of redundancy definitely will help.\u201d Liu said. \u201cYou\u2019re not talking about chatting with ChatGPT and making a mistake \u2014 you just say, \u2018hey, this is so silly, redo.\u2019 You\u2019re talking about lives.\u201d<\/p>\n
Xpeng\u2019s Edge<\/p>\n
China\u2019s chaotic roads, Liu argued, actually give XPeng an edge in data diversity and corner-case collection as compared to Tesla, which is primarily training its FSD system in the U.S. \u201cIn China, you have a larger chance to get a corner cases and data,\u201d he noted, adding that this gives the company more chances to expand internationally owing to the diverse data. \u00a0<\/p>\n
XPeng isn\u2019t just trying to ship better driver-assist software; it sees itself as a \u201cphysical AI\u201d company first and an automaker second. \u201cWe need physical devices in the real world to make sure we get feedback, we get data. Just like Google is producing Pixel devices just trying to demonstrate, \u2018OK, this is what Android can do.\u2019 But on the other hand, we want to make sure we are an AI company,\u201d Liu said.<\/p>\n
Liu wants the same foundation models to generalize far beyond cars\u2014to robots and flying vehicles, areas where the company is also investing.<\/p>\n
The Volkswagen Partnership<\/p>\n
German automaker Volkswagen announced a strategic partnership with XPeng in 2023 to license the Chinese EV maker\u2019s self-driving tech for its own China-market models. Tesla, meanwhile, is yet to find takers for its FSD system, a fact that Musk has often expressed frustration about.<\/p>\n
How Did Retail Traders React?<\/p>\n
On Stocktwits, retail sentiment around XPEV stock fell from \u2018extremely bullish\u2019 to \u2018bearish\u2019 over the past week, while message volume fell from \u2018extremely high\u2019 to \u2018normal\u2019 levels. \u00a0<\/p>\n
A Stocktwits user applauded Xpeng as a \u201csolid company.\u201d<\/p>\n
XPEV stock<\/a> has fallen 22% this year owing to weak vehicle deliveries and slowing demand.\u00a0<\/p>\n Read More: WBD, PSKY Stocks On Track For Worst Day Since December As States Reportedly Plan Lawsuit Against $110B Merger<\/a><\/p>\n For updates and corrections, email newsroom[at]stocktwits[dot]com.<\/p>\n
![\"Follow](\"https:\/\/chunks-prd.stocktwits-cdn.com\/_next\/static\/media\/google-news-icon.d9c2bfd1.svg\"\/){kind=icon}
<\/a><\/p>\n