XPeng Bets on Physical AI, Targets Robotaxi Break-Even Per Vehicle by 2027
Claire Weston
XPENG is pivoting from carmaker to Physical AI company, banking on robotaxi and humanoid robots as its next growth engines; Chairman He Xiaopeng has set a clear deadline — per-vehicle robotaxi breakeven by H2 2027 — while near-term fundamentals remain under pressure.
Why is XPENG shifting from "building cars" to "building robots"?
XPENG swung back to a loss in Q1 this year; first-half vehicle deliveries fell 16% year-on-year. Selling cars alone is not working short-term.
He Xiaopeng is positioning Physical AI — machines that use AI to act in the physical world, from self-driving cars to robots — as the company's next growth engine.
This means → XPENG is using a long-term strategic narrative to offset near-term sales pressure, betting that "over the next decade, cars will look more and more like robots on four wheels."
How does the robotaxi make money — and how does it differ from Waymo?
XPENG will not operate its own fleet. It acts as a technology supplier: providing AI models, full-stack software and hardware, and vehicles to local partners, who handle day-to-day operations.
In plain terms = Waymo runs a taxi company; XPENG sells "the complete toolkit for running one."
Macquarie Capital's head of China equity strategy, Eugene Hsiao, notes the approach is closer to Tesla — both use a vision-only stack, develop robotaxi software and hardware in-house, and skip lidar.
What does breakeven by 2027 actually depend on?
The key variable He Xiaopeng cites is not total fleet size but vehicle density within a specific operating zone — in smaller cities, a fleet of just a few hundred cars can hit the breakeven threshold.
Internal testing is underway in Guangzhou but has not opened to the public; a new vehicle designed specifically for robotaxi use will launch in 2026.
This means → the breakeven logic is "raise density," not "spread scale" — so XPENG does not need to burn heavy capital on fleet expansion early on.
How important is the overseas market in this strategy?
XPENG is in talks with dozens of potential partners, more than half of them outside China.
He Xiaopeng expects overseas markets to account for the majority of robotaxi revenue, with plans to expand the service internationally over time.
This reflects XPENG's read that competition inside China is fierce — and that robotaxi margins may be wider abroad.
Humanoid robots — how far along is the "hard path"?
XPENG has invested roughly eight years in R&D, choosing to build humanoid robots directly rather than easier-to-commercialize industrial automation equipment.
Target: monthly production capacity for its IRON robot exceeding 1,000 units by end-2026, with global deliveries starting the following year.
He Xiaopeng identifies Europe as a key market (high labor costs) and retailers as the earliest customer segment; he wants to create a "ChatGPT moment" — robots reliable enough to mass-produce, sell to ordinary consumers, and continuously improve via OTA updates.
What is the biggest risk in this transformation story?
Eugene Hsiao's assessment is blunt: "The market may underestimate the breadth of XPENG's AI strategy, but it may also overestimate how quickly that strategy converts to profit."
In plain terms = building technology wide does not mean making money fast — progress does not automatically become profit.
Two verification checkpoints matter above all: can the robotaxi hit per-vehicle breakeven on schedule in 2027, and can the humanoid robot complete its production ramp in 2026?
Content is for reference only, not financial advice.