
In Q3, XPeng still lost 870 million yuan, with the loss year-on-year expanding by more than 100 percent. However, stay calm; XPeng has entered acceleration mode. In Q4, unexpectedly, under non-GAAP, it might even profit.
On the evening of November 16, XPeng Motors released its 2026 Q3 report and held an earnings conference call that same evening.In Q3, deliveries reached 118,390 units, a year-on-year increase of only 2.1%, yet total revenue increased 19.1% year-on-year to 24.28 billion yuan.
Revenue outperforming sales by 17 percentage points indicates a significant improvement in XPeng's operational quality.
Future may be better — He Xiaopeng predicted on the call: 60,000 units sold per month in November.
"Our automotive sector has already entered a high-growth period driven by multiple star products and global expansion. Over the next few quarters, XPeng's scale effect will continue to improve," said He Xiaopeng.
As for 2027, technical services, Robotaxi, and robotics businesses outside of XPeng's automotive business will be fully launched. Correspondingly, He Xiaopeng has also set tasks for himself for 2027.
01
Q3 Records High
In 2026, the Chinese auto market fell significantly. However, XPeng Motors is still growing and hitting new highs.
Q3 2026 was the quarter with the most car sales and highest revenue for XPeng.

But there is a contrast: Q3 deliveries were 118,390 units, up 2.1% year-on-year, up 15% quarter-on-quarter.
Automotive sales revenue was 21.63 billion yuan, up 19.8% year-on-year, up 26.9% quarter-on-quarter.

The reason is simple: improved vehicle product mix led to an increase in average selling price.
Rough calculation:Q3 revenue per car was about 183,000 yuan, about 156,000 yuan at the same time last year, an increase of about 17% in a year. The main force raising prices was product structure: Q3 deliveries for GX exceeded 20,000 units, accounting for nearly 20% of XPeng's total sales, ranking among the top in new energy models above 300,000 yuan; MONA L03 deliveries exceeded 20,000 units in October, and capacity is still being pushed up.

This improvement translated to real money on the income statement, with automotive gross margin at 13.1%, up 1 percentage point quarter-on-quarter, flat compared to the same period last year.
However, group gross margin was only 18.9%, down 1.8 percentage points quarter-on-quarter. The contrast lies in another business segment: Service and other revenue was 2.65 billion yuan, up 13.7% year-on-year, mainly from carbon credit transaction revenue, but down 1.9% quarter-on-quarter; Service and other profit margin was 67.1%, 74.6% in the same period last year, 75.1% last quarter. The reason is not complex — technical and R&D service revenue recognized after major milestones were achieved in the previous year and last quarter was lower this quarter, including the technology licensing revenue from the cooperation with Volkswagen in Q2.
In addition, XPeng has high expense outlays. R&D expenses were 2.85 billion yuan, up 17.5% year-on-year; Selling, general and administrative expenses were 2.78 billion yuan, up 11.4% year-on-year. This was due to increased marketing and advertising investment, as well as increased commissions paid to franchise stores.

The result was a net loss of 870 million yuan, more than double the 380 million yuan at the same time last year; but narrowed by 35% compared to the 1.34 billion yuan in Q2.
Cash flow is not bad, He Xiaopeng pointed out:Q3 operating cash flow exceeded 2 billion yuan; as of September 30, cash on hand was 45.41 billion yuan, nearly 5 billion more than the end of June.
What supported the entire income statement was going global.
Wu Jiaming, VP of Finance and Accounting of XPeng Group, said,Q3 exports accounted for more than 20% of total deliveries, export business gross margin is more than double that of domestic business, profit contribution and net profit level are several times that of domestic, profit margins are in the industry's first echelon.
02
Last Two Months, Average 60,000 Per Month
XPeng drew a quite steep line for Q4: Deliveries of 169,000 to 174,000 units, up about 45.4% to 49.7% year-on-year, up about 42.75% to 46.97% quarter-on-quarter; Total revenue 30.4 billion yuan to 31.8 billion yuan, up about 36.6% to 42.9% year-on-year.
October already delivered 50,237 units. Working backward based on guidance, the lower limit requires 118,763 units combined for November and December, averaging about 59,400 per month; the upper limit requires 123,763 units combined, averaging about 61,900 per month.
He Xiaopeng said that November is expected to be the first time XPeng achieves monthly sales exceeding 60,000 units.
What supports 60,000 units are hit new products.
MONA L05 launched at the end of October, quickly becoming a hit in China's large five-seat market. 98% of customers chose Ultra and Max versions equipped with advanced intelligent driving. He Xiaopeng hopes its production ramp-up speed will break company records, with first-month deliveries exceeding 10,000 units. The relay of GX, L03, and L05 is exactly the high-growth period driven by multiple star products he mentioned.
Will it be profitable in Q4?
Not very clear. Gross margin may decline. Wu Jiaming gave a heads-up in advance:As deliveries of models like L05 begin, average selling price is expected to drop slightly, gross margin will drop slightly quarter-on-quarter, the Volkswagen technology licensing revenue in Q2 will not be repeated, and the group gross margin for the year will remain at a relatively high 10-plus level. Based on Q3's 18.9%, revenue of 30.4 billion to 31.8 billion corresponds to gross profit of 5.75 billion to 6.01 billion yuan.

Expenses are the dividing line. Q3 R&D plus SG&A totaled 5.63 billion yuan. Q4 is the quarter for boosting deliveries, also the marketing peak season, SG&A is hard to compress, R&D is also hard to cut due to intelligent driving, AI, and new models. Neutral assumption is 5.6 billion to 5.8 billion yuan. Subtracting expenses, plus about 100 million yuan other net income,probability of turning positive on Non-GAAP basis is high; but for GAAP net income, it still needs to deduct hundreds of millions in share-based compensation and net interest per quarter, much more difficult.
So He Xiaopeng described short-to-medium-term operating goals quite modestly:Achieve positive cash flow and profitability for the automotive business, and ensure strategic investment in physical AI. Translated, it means proving in the remaining time of 2026 that this business can support itself.
03
He Xiaopeng's 2027 Tasks
Management's statement on 2027 is more confident than any year in XPeng's past.
He Xiaopeng said,XPeng's growth speed in 2027 will significantly outpace the industry. Overseas, at least a three-fold growth is expected — and 2026 full-year overseas sales will break 100,000 units for the first time. October overseas sales reached 13,000 units, up 166% year-on-year.
Looking at the structure, Co-President Gu Hongdi gave regional breakdown: Europe about 50%, Southeast Asia and Asia-Pacific about 30%, Middle East and Latin America about 20%. The increase mainly comes from G9L, GX, L05 and other newly released global models.

On November 12, XPeng Group held the UK Brand Day and the launch conference for XPeng MONA L03 and XPeng G9L, with Gu Hongdi giving a speech.
As for next year, overseas sales and service outlets target over 1,000 locations. Supporting the premium is intelligent driving: At the October Paris Auto Show, XPeng was the only Chinese automaker to officially cooperate with the organizer to conduct autonomous driving experiences. NGP passed the Arc de Triomphe roundabout, expecting to obtain the first batch of EU regulatory permits in 2027 and launch city NGP in Europe first.
Policy variables are also on the table:Export tax rebates in 2027 will drop from 13% to 9%. Wu Jiaming said the response strategy is localization — Europe has begun manufacturing through cooperation with Magna Austria, Southeast Asia is laid out in Malaysia, and 2027 new product pricing has also considered tax rebate changes.
Besides selling cars and technology export, XPeng also has two longer and harder to judge curves: Robotaxi and Robotics.
In October,XPeng officially launched the Technical Solutions business, also known as AI Engineering, outputting physical AI capabilities such as Turing AI chips, models, autonomous driving, robots, etc.Gu Hongdi personally led the formation of the BD team, defining this as productizing capabilities accumulated from cooperation with Volkswagen, and revealed that revenue generation is expected to start next year. Robotaxi obtained Guangzhou's intelligent connected vehicle application demonstration operation qualification on October 26, able to carry out L4-level public operations throughout Guangzhou. Next year, it will output overall solutions to global partners. He Xiaopeng judges that scaling will start from overseas in 2028.

The story of robotics is inseparable from competition. Goldman Sachs directly asked on the call:Now major large model giants are accelerating into physical AI. Where is the moat for XPeng robots? Is it necessary for future robot manufacturers to develop their own models? Is the core competitiveness in models, data, motion control, hardware-software combination, or scaled deployment capabilities?
He Xiaopeng didn't beat around the bush. He first pulled his vision back to the early years of autonomous driving: 2015 to 2020, many companies could "achieve" autonomous driving in computer simulation, but today it hasn't been implemented yet, due to accuracy and safety issues; around 2018, many L4 companies produced beautiful demos, but implementation speed was far from imagined, stuck on cost; China's autonomous driving once appeared to have edge-cloud combined architecture solutions, also didn't work through, most critically is latency. So according to his judgment,trying to achieve combined capabilities of System 01 and 02 through a general model is hard to implement.
His answer lies in "two directions, four capabilities".
The first is hardware native capability. He emphasized that the vast majority of people extremely ignored this point. The basis and core of physical AI should be "physical", not "AI";
The second is scaled commercial manufacturing and sales capability. Only scale has commercial capability, and only scale can have data capability;
The third is high-quality, diverse data capability — now many large model companies purchase and collect large amounts of visual data, but for robots, tactile, pressure, temperature, body sensation and other tactile data are more important. "Tactile data is equivalent to visual data in the autonomous driving field, extremely important". Without hardware, without real data feedback, true physical scenario generalization cannot be achieved;
The fourth is compute power on the edge side and cloud side. Edge inference, cloud inference, training and simulation all need huge compute support. He judges that Chinese in-house car manufacturers have good reserves and barriers in edge compute power.
This way of thinking falls on the product, which is the route he drew for XPeng robots. Today, robots in China and even globally mainly follow a to Large Business, customized business logic.
He Xiaopeng said he does not expect to make products that need customization, but expects it to be a highly standardized commodity, serving Small and Medium Business, or even individuals and family customers in the future. In terms of scenarios, he also made distinctions: From a Euro-American perspective, the best scenario is industry, because industry has slightly lower requirements for safety systems, higher requirements for hand capabilities, and stricter cost requirements; from China to the world, he thinks the best scenarios are commercial and home. These scenarios have higher requirements for safety and overall design aesthetics, and slightly lower requirements for hand capabilities — the difference is very large, which rather helps XPeng establish differentiation in different sub-segments.
On the production end, XPeng's first high-level general-purpose robot automated production line commissioned in September has a core process automation rate exceeding 80%, mass production according to automotive-grade quality standards, supply chain project agreements have also been signed, mass production started on January 1, 2027,in the first quarter of next year, the first batch of mass-produced units will enter some XPeng stores for trial, He Xiaopeng hopes it will become a "star employee" in the store, and in the second quarter of next year it will be officially released to external commercial customers.
His judgment on the business side is high: XPeng Robot's sales potential and gross profit scale are expected to far surpass the current hottest high-end models on the market. Wu Jiaming added from the income statement perspective,the profitability of the robotics business is expected to be significantly better than high-end electric vehicles on the market.
As for the commercialization path, He Xiaopeng's plan is,with mass delivery, massive real-scenario data will greatly accelerate AI model iteration speed, further expanding leading advantages.
Putting the automotive, technology export, Robotaxi, and robotics these four lines together, what really needs to be watched in 2027 is not how many cars XPeng can sell, but whether the latter two curves can really record revenue in the financial statements.
He Xiaopeng's own stance is,the development of physical AI is not a 100-meter sprint, but a marathon for the next ten years. This means XPeng must keep pace in automotive business and other businesses.
He Xiaopeng also stated his 2027 task:"I plan to devote more energy to the comprehensive improvement of operations, enhance the company's self-sustaining capability, while promoting the mass production and delivery of advanced robots, and significantly improve XPeng's global localization layout and brand momentum."