Xpeng Loss Narrows to $53.2m in Q3, Ramps Up Push into Embodied Intelligence

For the fourth quarter, Xpeng is confident about turning profitable—but how can profitability become self-reinforcing?

Xpeng released its third-quarter financial report last night.

To put it simply, the company is still unprofitable, posting a net loss of RMB 380 million ($53.2 million).

Xpeng delivered 116,000 vehicles in the third quarter, up 149.3% year-on-year and 12.4% quarter-on-quarter. The results are clearer when viewed through the financial lens:

• Revenue: RMB 20.38 billion ($2.85 billion), up 101.8% year-on-year and 11.5% quarter-on-quarter;
• Automotive sales revenue: RMB 18.05 billion ($2.53 billion), up 105.3% year-on-year and 6.9% quarter-on-quarter;
• Overall gross margin: 20.1%, versus 14% in the same period of 2024 and 17.3% in Q2 2025;
• Automotive gross margin: 13.1%, compared with 6.4% in Q3 2024 and 14.3% in Q2 2025;
• Net loss: RMB 380 million ($53.2 million), versus RMB 1.28 billion ($179 million) a year earlier and RMB 480 million ($67.2 million) in Q2 2025;
• Cash and equivalents: RMB 17.53 billion ($2.45 billion), compared with RMB 18.59 billion ($2.60 billion) a year ago.

Ahead of the moment when clarity finally emerges, Xpeng did not deliver a dramatic last-minute profit surprise.

Instead, it doubled down on R&D investment in pursuit of a more durable long-term position. The company’s AI Day on November 5-6 and the unveiling of the X9 Extended-Range edition illuminate the trajectory ahead.

The M03 has maintained monthly sales of around 15,000 units and remains Xpeng’s most reliable model. But looking further forward, Xpeng needs many more M03-level successes.

For the fourth quarter, Xpeng is confident about turning profitable—but how can profitability become self-reinforcing? Last night’s earnings call offered insight.

Seven Models in 2026

The extended-range Xpeng X9 launch event on November 20 marks a significant strategic pivot for the company.

This is because it inaugurates Xpeng’s “dual-power” era.

After the extended-range X9, Xpeng will launch three additional super-extended-range models in the first quarter of 2026, and a total of four new “dual-power” models throughout 2026. In other words, Xpeng plans to introduce seven dual-power vehicles in 2026.

Featuring a 1,602-km CLTC combined range and a third-generation 1.5T range-extender paired with an 800-V hybrid SiC coaxial electric drive, the extended-range X9 has already exceeded pre-sales expectations and helped Xpeng reach inland and northern consumers who traditionally resist pure EVs.

This is only the starting point. Xpeng intends to unleash full offensive capability in the extended-range segment in 2026, targeting multiple market categories and price bands.

The first goal is rapid, precise territorial expansion.

Automakers entering the extended-range race see it as a path to scale—and ultimately, profits.

CEO He Xiaopeng noted during the call: “For large vehicles, extended-range penetration will be higher, while for small A-class cars, pure EV penetration will remain dominant.”

The extended-range strategy is another opportunity for Xpeng to move upward.

With a similar product matrix to Q2, Xpeng’s Q3 automotive sales revenue growth (+6.9% QoQ) lagged behind revenue growth (+11.5% QoQ) and deliveries (+12.4% QoQ), while automotive gross margin (13.1%) fell short of Q2 (14.3%).

Average selling price dropped again to RMB 156,000 ($21,840), down from RMB 164,000 ($22,960) in Q2 and RMB 189,000 ($26,460) in Q3 2024.

Portal data shows that both the G9 and X9 delivered fewer than 1,000 units in October, meaning large-vehicle penetration remains weak.

Xpeng’s “super-extended-range” roadmap may aim to create an expanded “M03 family.”

Beyond extended-range vehicles, Xpeng’s R&D spending surged 49% year-on-year in Q3 to RMB 2.43 billion ($340 million), supporting not only powertrain upgrades but also the three most frequently mentioned keywords in the call: VLA 2.0, Robotaxi and humanoid robots.

“A Global Embodied-Intelligence Company”

These three pillars are central to Xpeng’s future.

He Xiaopeng stated: “The continued improvement in operations reinforces our determination to invest in physical AI, accelerating mass production of VLA 2.0, Robotaxi and humanoid robots in 2026.”

Every decision today is underwriting the future.

In He’s vision, Xpeng ultimately becomes “a global embodied-intelligence company.” VLA 2.0, Robotaxi and humanoid robots together form the foundation of that identity.

He described VLA 2.0—which will open early access later this year and deploy widely in Q1 2026—as “a major leap forward,” with parameter counts 10 times those of the previous end-to-end model, pushing the intelligent-driving experience to a new level.

But its strategic meaning extends far beyond driver assistance.

VLA 2.0 is intended to serve as the unified “physical AI brain” across all future embodied-intelligence products.

This platform has received external validation: Volkswagen increased investment and became the first launch partner for VLA 2.0, and Xpeng’s Turing AI chip secured Volkswagen designations.

Xpeng aims to evolve from a provider of R&D services to a supplier of core AI models and chips for embodied-intelligence products—high-value components with strong profitability potential.

Commercialization will focus on Robotaxi fleets and humanoid robots.

Robotaxi-ready configurations will be available on existing models, with trial operations in China in 2026, and an SDK to expand development scenarios.

The spotlight during the earnings call was directed toward the IRON humanoid robot.

For the IRON robot, which dominated Xpeng’s AI Day discussions, He revealed an ambitious target: annual sales exceeding one million units by 2030.

This confidence stems from technological conviction.

He said IRON demonstrated only “a small portion of its capabilities” so far, adding: “Once humanoid robots cross the inflection point of intelligence and electrification like China’s EV industry, explosive growth will follow.”

IRON is currently at generation seven; the eighth generation will be true mass production. Based on his observations, parts of the market are still at generation three to five, still relying on joint-drive architectures that face commercialization challenges.

Internally, Xpeng is already thinking about how to improve public acceptance.

He shared a key insight: “When a robot has muscles and skin, people dare to hug it.” With no skin, no one was willing to touch it—what people want a humanoid robot to do is as important as what it is capable of doing.

To achieve scale, Xpeng is now moving at full speed.

He said the next-generation production model will enter the ET0 phase next month, with large-scale production of advanced humanoid robots planned by late 2026, initially targeting guidance, retail assistance and inspection scenarios.

Cost structure will differ dramatically from traditional hardware: software and AI could account for 50% from the outset, comparable to automotive total costs, reshaping pricing logic and explaining Xpeng’s confidence in this bet.

Finding the Balance Point

Stable sales of the M03 and P7+, incremental uplift from the extended-range X9, and RMB 2.33 billion ($326 million) in “services and other revenue” featuring a 74.6% gross margin—largely driven by the Volkswagen partnership—suggest further operational improvement in Q4.

With validation across both product and technology dimensions, Xpeng is actively pursuing its next revenue engine.

Challenges remain, including pricing-versus-volume trade-offs and the pace of transformation into a technology-driven company.

But as He Xiaopeng puts it, “The future of automotive must be defined through the lens of AI, combining full-stack self-development and cross-domain integration to create differentiated product capabilities.” From horsepower and range to ecosystems and AI, Xpeng is already moving.

The probability of success is uncertain—but everyone is watching closely.