EN / 中文

XPeng Inc. Q3 2026 Earnings: Record Deliveries, Widening Loss and Its High-Stakes Physical AI Bet

by guangzixingqiu·November 19, 2026

Written by | Xu Zhi, Edited by | Wang Pan

On November 16, XPeng Inc. released its financial results for the third quarter of fiscal year 2026, delivering another mixed report card.

Revenue for the quarter grew 19.1% year-over-year, and deliveries reached 118,400 units, both setting new quarterly records. Meanwhile, automotive gross margin and cash reserves improved on a quarter-over-quarter basis.

However, the net loss for the quarter widened to CNY 870 million, an increase of over 120% year-over-year. A key driver was the sustained growth in R&D spending, which rose17.5% year-over-year.

While the core automotive business showed steady improvement, the company faced a scenario of rising revenue without corresponding profit growth. The underlying cause is the comprehensive "cash drain" from the physical AI business.

The shift has been foreshadowed since late last year when XPeng upgraded its corporate positioning, culminating in the company's name change this April. XPeng’s "engine" has effectively switched from automobiles to physical AI. During the earnings call that evening, CEO He Xiaopeng reaffirmed the strategy: the short-to-medium-term goal is to achieve positive cash flow and profitability in the automotive business, thereby securing funding for long-term strategic investments in physical AI.

Just as it once aggressively pursued intelligent driving technology, XPeng remains the most radical among automakers venturing into physical AI. In the embodied intelligence sector, compared to NIO's conservatism and Li Auto's wavering stance, XPeng has gone all-in.

Yet, the new bet is far from paying off. Physical AI will inevitably fail to pull the heavy lever of XPeng's overall performance in the short term, while the automotive business also harbors certain hidden issues.

The biggest challenge for XPeng lies in balancing the time lag—ensuring that the self-sustaining capability of new businesses catches up quickly with the rate at which the automotive business must subsidize them.

The Cost of Extending the "Health Bar"

XPeng’s automotive core business "health bar" is significantly extending.

Although the growth rates in revenue and deliveries are no longer comparable to the previous high-speed expansion phase, given the current extremely weak and fiercely competitive market environment, XPeng’s performance can be considered passing.

The push into overseas markets, particularly establishing a foothold in Europe, has been the biggest contributor.

In the third quarter, total deliveries grew 2.1% year-over-year and 14.6% quarter-over-quarter. Overseas deliveries reached approximately 27,900 units, surging over 150% year-over-year, accounting for more than 20% of total deliveries. In October alone, XPeng’s overseas deliveries exceeded 13,000 units, growing 166% year-over-year.

Looking solely at the automotive business, overseas markets are showing a trend of taking over from domestic operations as the primary engine. He Xiaopeng has set a target of surpassing 100,000 overseas sales this year, doubling that figure next year, supported by expanding overseas stores to over 1,000 locations.

By comparison, as of September 30 this year, XPeng had 731 domestic stores. This means that as early as next year, XPeng’s number of overseas stores could exceed its domestic ones.

Overseas sales volume provides visible support, but the more important implicit value lies in boosting profits. Management revealed that the export gross margin is more than double that of the domestic business.

Previous comparisons have shown that prices for many of XPeng’s global models in European markets are typically around 2.5 times their domestic counterparts. Driven by higher vehicle prices, even after excluding taxes and localization costs, per-unit profits are far more attractive than in China.

Furthermore, XPeng’s guidance for the fourth quarter indicates continued growth, with expected deliveries ranging from 169,000 to 174,000 units, representing a substantial increase of about 45% both year-over-year and quarter-over-quarter. Revenue is projected to grow 36.61% to 42.9% year-over-year.

Of course, extending the "health bar" comes with irreversible negative costs in the short term.

First, the comprehensive rollout of low-price strategies has severely limited brand strength and the gross margin ceiling in the domestic market. The direct manifestation of declining brand power is the sluggish sales of mid-to-high-end models. Currently, the MONA series has become XPeng’s absolute sales pillar, with cumulative deliveries exceeding 350,000 units over two years.

Management noted during the earnings call that as the sales share of the MONA L05 increases, XPeng’s average selling price (ASP) in the fourth quarter will decline further, and gross margins may dip slightly.

Second, a key support for the sales curve comes from the frequent launch of new models. In other words, this growth is partly built upon consuming the lifecycle vitality of older models.

XPeng’s fundamental base remains automotive, but its positioning has shifted to fueling the physical AI business. Therefore, it cannot adopt the contractionary measures used by other automakers to weather the winter. Management explicitly stated that R&D investment related to physical AI will surge to CNY 7 billion this year and continue to grow substantially thereafter.

Switching Engines

When XPeng officially changed its name to XPeng Group this April, He Xiaopeng posted on WeChat Moments: "From the journey of smart electric vehicles to flying cars, Turing AI chips, VLA autonomous driving models, humanoid robot IRON, and Robotaxis, we are turning dreams into reality step by step on our global quest for physical AI."

It is fair to say that XPeng’s self-perception is no longer that of an automobile company.

Organizational restructuring is the most intuitive reflection of this shift. Early this year, XPeng merged its Autonomous Driving Center and Intelligent Cockpit Center into the General Intelligence Center, led by Liu Xianming. Positioned as an AI organization rather than just a driving assistance department, a separate Robotaxi business unit was also established.

In June, He Xiaopeng personally took on the role of CEO for the robotics business, handing over the automotive business to Wang Fengying. This top-down move signaled a determination to leverage the entire group’s resources to safeguard robot mass production.

The deep-seated reason for switching engines from automotive to physical AI stems from judgments about future competition. He Xiaopeng believes physical AI represents a business opportunity with higher gross margins and stronger barriers to entry. Robots are the typical representative; he optimistically stated, "In terms of business model, I believe the sales potential and gross margin scale of XPeng IRON could significantly surpass those of the most popular high-end vehicles currently on the market."

This is a proactive choice. XPeng has always been a somewhat radical "geek" among automakers. Beyond its heavy bets on intelligent driving, its comprehensive technical and business benchmarking against Tesla serves as further proof.

The technological accumulation in the automotive base has laid the groundwork for XPeng’s physical AI business. Direct support includes software and hardware capabilities such as the second-generation VLA and Turing AI chips, while indirect support involves the reuse of engineering and supply chain capabilities, along with data and scenario backing.

From this perspective, the value of the automotive business extends beyond financial subsidies; it acts as a pathfinder and testing ground for the physical AI business, enhancing capabilities in cost control and efficiency.

A few simple examples illustrate this: XPeng IRON and Robotaxi utilize the same second-generation VLA and foundation models as the cars. Over 85% of IRON’s supply chain partners overlap with the automotive system. Moreover, XPeng’s existing store network will serve as the first deployment site for IRON robots.

The combination of software/hardware technical foundations, engineering capabilities, and store application scenarios constitutes XPeng’s greatest advantage compared to other embodied intelligence players. This is why He Xiaopeng asserts, "The foundation and core of physical AI should be physics, not just AI."

This narrative has gained recognition from capital markets. In August, XPeng’s robotics business completed its first equity financing round, raising over USD 900 million. This set a new record for single-round private financing in China’s embodied intelligence industry, with a post-money valuation exceeding USD 6.3 billion.

For comparison, XPeng Group’s market capitalization recently stood at approximately USD 9.5 billion. The situation where a spun-off, non-listed business segment’s valuation approaches that of the listed parent company has previously occurred with Kuaishou.

Sitting at a High-Stakes Table

The path to the stars and seas is not smooth. He Xiaopeng previously described, "The difficulty of building robots is 20 times that of building cars."

The high valuation of XPeng’s robotics business is largely attributable to the capital boom in China’s embodied intelligence sector this year, rather than being a true reflection of its operational strength.

Supported by the automotive business, XPeng chose the most difficult starting point. It has essentially sat down at a high-stakes table requiring continuous cash flow supply, with the timeline for realizing returns being much longer than in the automotive sector.

First, regarding embodied intelligence, XPeng selected the longest monetization path: the general-purpose route. From the outset, IRON was defined as a high-level general-purpose robot. Rather than lowering its standards to create a vertical robot capable of quick tasks, it aimed directly at the endgame, similar to Tesla.

In terms of scenarios, IRON targets both commercial and home environments. While industrial scenarios are the preferred choice for the vast majority of players, with business models focused on B2B or customization for large clients, XPeng refuses to follow this path. It seeks to provide "plug-and-play" standardized products to small and medium-sized enterprises and households.

The rationale again stems from a "begin with the end in mind" judgment: Industrial robots have lower safety redundancy requirements but extremely stringent demands on performance metrics and cost control. To go from China to the world, commercial services and homes represent the optimal scenarios.

He Xiaopeng believes XPeng’s core advantage in robotics lies in high-level intelligence and scenario generalization capabilities. This is a path that is difficult at first but easier later. There are no guaranteed orders from large clients to fall back on initially, and the business model must be explored independently. However, once standardized products succeed, the ceiling will be far higher than that of most competitors.

Based on the current pace, XPeng IRON is on the eve of mass production. The first automated production line was launched in September, with hardware mass production preparation scheduled for completion by year-end. IRON will enter XPeng stores in Q1 next year and be released to external commercial customers in Q2.

At that point, XPeng’s robots will truly face market scrutiny and direct comparison with competitor products. If XPeng’s robot valuation is currently supported by vision, the real battle begins only after mass production and delivery.

Flying cars represent another marathon. Although XPeng’s "Land Aircraft Carrier" has secured over 7,000 orders, and the Guangzhou factory plans an annual capacity of 10,000 units with early deliveries starting by year-end, commercial operation remains unlikely to see the light of day within the next two to three years.

Uncertainty stems from regulatory policies, which are crucial in determining whether flying cars can transition from the "experience" stage to becoming genuine transportation tools. Zhao Deli, CEO of XPeng AeroHT, previously stated that the world will enter a period of rapid development for flying cars in the next decade. Interpreted optimistically, it will still take at least 5-8 years for ordinary consumers to own or use flying cars as easily as ride-hailing services.

In contrast, Robotaxi may be the fastest-moving of the three physical AI business lines. Recently, XPeng obtained L4 demonstration operation qualifications and has begun public operations across Guangzhou. It expects to refine a replicable product system by year-end and output solutions to global partners next year.

However, XPeng’s Robotaxi must still answer two questions: verifying the reliability of pure-vision solutions and validating the per-vehicle profitability model. Neither Tesla (for the former) nor Apollo Go, WeRide, or Pony.ai (for the latter) have fully cracked these challenges. For a newcomer like XPeng, this is likely not a game that can be "speedrun" in a short time.

Overall, new businesses cannot monetize quickly, yet high levels of investment cannot stop. This means XPeng will find it difficult to completely escape the gravitational pull of four wheels and fully enter the new phase of physical AI in the short term.

Therefore, differentiated dual-front warfare is destined to be XPeng’s mainstream narrative for the coming years. On one hand, it casts a long line in the physical AI track to catch big fish, betting on a leading position in a new battlefield. On the other hand, it pragmatically casts short lines in car manufacturing to catch small fish, using sufficient bread and milk to sustain itself until dawn.