The competitive landscape in the intelligent vehicle sector is expanding beyond car sales into AI and robotics.

On August 24, XPENG Inc. released its second-quarter 2026 financial results and simultaneously announced that its humanoid robotics arm had signed initial equity financing agreements with several investors.

These two developments landing on the same day highlight the two distinct business lines currently shaping XPENG. On one side, vehicle deliveries have rebounded notably from the first-quarter trough, with new models, overseas markets, and technology service revenue jointly supporting the top line. On the other, the robotics business is now bringing in external capital, securing an independent market valuation, and beginning to raise funds for mass production and commercialization.

The earnings report shows XPENG generated RMB 19.74 billion in second-quarter revenue, up 8.0% year-on-year and 51.5% quarter-on-quarter. Gross margin came in at 20.7%, holding above the 20% threshold for a second consecutive quarter.

The improvement in blended gross margin was largely driven by high-margin businesses such as technology R&D services. Automotive gross margin stood at 12.1%, down 2.2 percentage points from the same period last year.

The robotics financing opens an independent funding channel for another long-term investment. The unit currently still operates jointly with the automotive business, with its income, expenses, and assets and liabilities continuing to be consolidated into the group's financial statements. The IRON robot is planned to enter volume production by the end of 2026.

The car business is still repairing its profit profile, while robotics has entered a pre-production investment phase. With both lines expanding simultaneously, the profitability and cash flow of the automotive operation will continue to determine how much XPENG can allocate to robotics.

Blended Gross Margin Holds Above 20%

In the second quarter, XPENG delivered 103,295 new vehicles, up 0.1% year-on-year and 64.8% quarter-on-quarter. The sequential rebound in deliveries was swift, though volumes were roughly flat versus the prior-year period.

Automotive sales revenue reached RMB 17.05 billion, up 1.0% year-on-year and 55.0% quarter-on-quarter, with the growth trajectory broadly consistent with delivery trends.

The more pronounced shift came in services and other income. According to the second-quarter report, this segment generated RMB 2.70 billion, up 93.9% year-on-year, driven primarily by reaching a milestone in technology R&D services provided to Volkswagen Group, alongside growth in parts and accessory sales.

The changing revenue mix directly influenced margin performance. XPENG's blended gross margin for the quarter was 20.7%, up 3.4 percentage points year-on-year and 0.1 percentage point from the first quarter. Automotive gross margin was 12.1%, flat sequentially but below the 14.3% recorded a year earlier. The company attributes the year-on-year decline in automotive margin primarily to product generational transitions.

Services and other businesses posted a gross margin of 75.1%. Based on the company's disclosed figures, automotive sales contributed roughly RMB 2.06 billion in gross profit, while services and other operations added approximately RMB 2.03 billion—the two are now nearly comparable.

This means the 20.7% blended gross margin cannot be directly equated with the profitability of the car-selling business. Technology service income enhances the resilience of the earnings structure, but whether automotive gross margin can recover as new models ramp up remains a key indicator of operational quality.

Expenses also reflect XPENG's current investment intensity. R&D spending reached RMB 2.91 billion in the quarter, up 32.1% year-on-year and roughly flat sequentially, with incremental spending directed toward new models and AI-related technologies. Selling, general, and administrative expenses totaled RMB 2.50 billion, up 15.2% year-on-year and 32.5% quarter-on-quarter, with the sequential increase linked to marketing investments and higher dealership commission costs.

XPENG posted a net loss of RMB 1.34 billion in the second quarter, narrower than the RMB 1.78 billion loss in the first quarter but wider than the RMB 480 million loss a year earlier.

As of the end of June, the company held RMB 40.48 billion in cash and cash equivalents, restricted cash, short-term investments, and time deposits, down RMB 1.61 billion from the end of March. The cash buffer still supports R&D and new business expansion, though increased spending has yet to translate fully into current-period profits.

The immediate priority for the second half is new-model production capacity. During the earnings call, He Xiaopeng noted that GX domestic deliveries surpassed 7,000 units in July. After the launch of the MONA L03, orders exceeded XPENG's previous model record, and new locked orders in the third quarter as of the call were up approximately 50% quarter-on-quarter. However, extreme weather and supply chain issues have affected the L03's production ramp, prompting the company to initiate double-shift manufacturing.

The product cadence also includes the G9L, launching and delivering in September, and the MONA L05, slated for domestic release in the fourth quarter. XPENG expects third-quarter deliveries of 115,000 to 121,000 units, up 11.3% to 17.1% quarter-on-quarter, with revenue guidance of RMB 21.7 billion to RMB 23.4 billion.

He Xiaopeng has also set a new target: monthly deliveries exceeding 60,000 units in the fourth quarter. Compared with an average of roughly 34,400 units per month in the second quarter, achieving this goal requires synchronized improvements in new-model orders, supply chain, and delivery capacity.

Overseas operations provide another pillar. XPENG's overseas deliveries surpassed 20,000 units for the first time in the second quarter, up 81% year-on-year. Overseas markets contributed over 25% of revenue in the first half, with management noting that the average selling price of export models exceeds EUR 40,000.

AI investment continues in parallel with the automotive product cycle. XPENG plans to push the first major upgrade of VLA 2.0 starting at the end of August.

Discussing overseas testing of VLA 2.0, He Xiaopeng displayed an unusual level of emotion during the call. He said he and his team recently completed localization acceptance testing in Germany. "To be honest, I was very excited at the time," he said, because the model, trained primarily on Chinese data, performed on German city roads in a way that was already relatively close to the experience in the Chinese market, with "almost no additional localized training data."

XPENG plans to pursue regulatory approval for VLA 2.0 in Europe as early as the first half of 2027, while exploring overseas software subscriptions and external technology partnerships. Whether road performance overseas can replicate domestic test results depends on factors including road environments, regulatory approvals, data compliance, and localization adaptation in different countries.

For XPENG, whether AI research can further convert into vehicle sales, product premiums, software revenue, or external technology services will determine whether heavy R&D spending yields more stable commercial returns.

IRON Sets Volume Production Milestone

On the day of the earnings release, XPENG announced that its robotics subsidiary, Dogotix, signed initial equity financing agreements with investors including IDG Capital, Gaorong Ventures, Tencent, and Alibaba. The round raised over USD 900 million at a post-money valuation exceeding USD 6.3 billion, setting a record for the largest single-round private financing in China's embodied intelligence sector.

The robotics business has thus secured an independent valuation and funding channel, though it will not become fully financially separate from XPENG Group in the near term.

According to the disclosed structure, of the approximately USD 900 million in subscription proceeds, external institutions contributed USD 600 million, XPENG Group internally invested USD 200 million, and entities controlled by XPENG Group Chairman and CEO He Xiaopeng, along with Honorary Vice Chairman and Co-President Gu Hongdi, jointly contributed USD 100 million.

Gu Hongdi stated that the automotive and robotics businesses currently continue to operate jointly and have not yet been spun off. The agreement includes a transition period of approximately 18 months. As XPENG retains controlling interest, the robotics business's income, expenses, and assets and liabilities will remain consolidated into the group's statements.

Compared with the financing scale, He Xiaopeng's responses on the earnings call regarding production ramp were more specific.

He noted that the mass-production logic for robotics differs from that of automobiles. "The production capacity challenge in the early stage is definitely about quality; in the later stage, it's definitely about sales."

Under the current plan, IRON will enter volume production by the end of 2026, initially deployed in XPENG stores and parks. In the first half of 2027, it will focus on proprietary scenarios before gradually expanding to external customers. By mid-year and into the second half, monthly production capacity will be raised to several thousand units based on market demand.

This response provided neither full-year 2027 sales figures nor a clear timeline for profitability.

Gu Hongdi said it remains too early to judge when the robotics business will turn profitable. He Xiaopeng added that robotics deliveries will be disclosed as "real" delivery data, similar to the automotive business. Rather than offering distant sales targets, this framing places the next verification focus on product quality, customer demand, and actual deliveries.

He Xiaopeng also emphasized that the technical challenges and innovation difficulty of advanced general-purpose humanoid robots far exceed those of intelligent vehicles, "at least 20 times greater." He further noted that the scarcity of high-quality supply would push the product's average price and gross margin well above those of automobiles.

The initial scenarios XPENG has chosen for IRON are also relatively restrained.

The company plans to first enter retail and service industry scenarios such as guided tours and shopping assistance, serving individual customers and small-to-medium enterprises. Industrial and home applications will come later, with an open SDK supporting partner secondary development. Stores and parks serve as both initial application scenarios and entry points for early testing, real-world data collection, and product improvement.

On the cost side, He Xiaopeng stated that over 85% of IRON's supply chain partners overlap with the automotive business. Pricing in the robotics industry typically ranges from 2.5 to 3 times material costs. He expects IRON's hardware gross margin to exceed XPENG's automotive business, with potential future additions from AI model upgrades, software, and subscription revenue.

The synergy between automotive and robotics primarily lies in chips, AI models, data infrastructure, supply chain, and manufacturing systems. IRON is equipped with three Turing AI chips with a stated effective computing power of 2,250 TOPS. The VLA and VLM models used in vehicles, along with portions of the simulation and training systems, can also be reused in the robotics business.

From this perspective, XPENG is not making a single choice between selling cars and robotics. Instead, it is attempting to leverage the automotive business for manufacturing, supply chain, and data infrastructure, while using independent financing to support robotics expansion.

Once IRON begins external customer deliveries in 2027, real orders and cash collections will provide the operational data to support the USD 6.3 billion valuation. This is an unavoidable step for any automaker entering embodied intelligence: technologies and supply chains can be shared, but the new business must still build its own order book and cash flow.