4Q25 revenue rose by 9.2% QoQ to RMB22bn, exceeding forecasts on non-automotive growth from carbon-credit trading and VW technical services. The company achieved its first-ever quarterly profit of RMB383m bolstered by government subsidy, but record-high OPEX and shortened payment cycles triggered a return to cash burn of RMB1.56bn. 1Q26 guidance points to March deliveries of 26k-31k units, underperforming NEV rivals. While management expects sequential sales recovery driven by market seasonal upturn and new launches, we anticipate substantive YoY growth will likely be deferred until 2H26, contingent on MONA new model rollout. We revise down our 2026- 2027E revenue forecasts by 15-16% to reflect lowered delivery estimates of 480k and 550k units, respectively. To counteract softening auto business, XPeng is aggressively pivoting towards a "Physical AI" ecosystem encompassing Robotaxis and humanoids. While these new ventures lack near-term contribution to revenue/profits, they offer valuation optionality, with key catalysts including humanoid sector's IPO momentum and XPeng Iron's mass production target of 1,000- unit/month by year-end. Hence, we maintain BUY but lower TP to US$21/HK$83 (1.5x 2026E P/S).
Key Factors for Rating
4Q25 revenue and gross margin modestly beat on non-vehicle gains though vehicle margin slightly missed. In 4Q25, total revenue rose by 9.2% QoQ to RMB22.3bn, above our prior forecasts on greater QoQ non-auto business rise by RMB854m (incl. over RMB300m contributed by after-sales finance, while the rest from carbon credit trading and VW tech services). Elevated high-margin non-auto mix boosted blended gross margin 1.2ppts QoQ to 21.3%, whereas vehicle margins stagnated at 13% QoQ in 4Q25, missing estimates even with higher-margin X9 penetration and export volume upticks.
First-ever quarterly profit bolstered by subsidy but surging OPEX and shortened payment cycles triggered a return to cash burn. In 4Q25, OPEX hit a record RMB5.67bn, bumping the OPEX ratio up 1.4ppts QoQ to 25.5%, topping our estimates. Elevated other operating gains, mainly led by government subsidies, yielded the first quarterly GAAP profit at RMB383m. However, with surging OPEX and shortened payment cycles, the company returned to cash burn in 4Q25 with net cash decrease by RMB 1.56bn, snapping four straight quarters of positive runs.
1Q26 delivery guidance sharply trails NEV rivals, yet non-auto proportion uplift set to anchor overall margin resilience. XPeng’s 1Q26 delivery outlook sharply lags NEV rivals on 30%-35% YoY plunge, guiding March sales to 26,000-31,000 units, shy of competitors. Though the mgmt. anticipated QoQ volume gains from sector upturn and intensive launches from 2Q25, we reckon the substantive YoY lift is likely deferred until 2H26 driven by two fresh MONA launches. On margins, vehicle margin should stay in the low-teens against sales headwinds, supported by greater X9 penetration and export growth, while non-auto contributions are set to anchor overall blended margins at 20% in 1Q26, as guided by the company.
Bold Robotaxi strategy though still lacking visible model for scaled profitability. XPeng's Robotaxi, fitted with second-generation VLA, has secured Guangzhou testing permits and is running routine L4 public road trials at present. The company targets safety-driver pilot demonstrations in 2H26 for testing business validation and economic feasibility, aiming to implement fully driverless service at scale in early 2027 and align with small number of global leaders in Robotaxi services. Although the company outlines an ambitious trajectory for Robotaxi development, we consider it highly preliminary, with no established open-tech infrastructure or a visible blueprint for scaled profitability.
Humanoid robots eyed for 1,000 per month run-rate by year-end, starting with XPeng store integrations. During the conference call, the mgmt. confirmed that XPeng’s homegrown humanoid robot Iron is on track for forthcoming mass production. Construction of a dedicated Guangzhou facility for mass production—at a monthly run-rate of 1,000 units by end-2026 (annualised capacity of 10,000+ units)—has kicked off, supporting a rollout timeline that evolves from commercial to industrial and home scenarios, starting in 2026 at XPeng stores and campuses for sales support and tour guidance. Scaling operations are expected to compress hardware costs akin to vehicle manufacturing, thereby realising high-margin potential through superior AI integration and premium market positioning.
Valuation
To reflect softened delivery outlook and tepid new-car demand, we trim our 2026-2027E sales volume forecasts to 480,000 and 550,000 units. Accordingly, we slashed 2026-2027E revenue estimates by 15%-16% to RMB92.3bn and RMB104bn. Combined with lower vehicle margins and ramped R&D, especially aggressive bets on physical AI, we project wider non-GAAP net losses of RMB3.25bn and RMB1.57bn for 2026-2027E.
Since the second half of last year, the core automotive business has been faltered, with sales volume and margins consistently falling short of expectations primarily due to a stalled push into the premium segment. Efforts in the mid-tohigh- end market, such as the G7 and the new P7 models, have flopped commercially, disrupting the company’s aspiration toward brand upscale.
To offset the softening auto business, XPeng is pivoting towards a "Physical AI" ecosystem including Robotaxis and humanoids. While these initiatives leverage its core automotive tech, we view them as highly nascent, lacking both mature business models and immediate monetisation paths. Consequently, they are unlikely to contribute meaningfully to revenue/profits in coming years. Meanwhile, the core NEV business, currently the most viable application of physical AI, is facing heightened competition and deteriorating cash flow, limiting its ability to fund these ambitious new ventures. Together with ballooning R&D outlays, we deem XPeng is likely to return to loss-making trajectory after brief 4Q25 profitability.
In the near term, the new launches over the coming three quarters, especially two MONA SUV models in 2H26, may trigger a recovery in vehicle sales, but subdued vehicle margins are likely to cap the extent of profitability gains. Over the longer term, progress in humanoid robotics has the potential to act as a catalyst, supported by forthcoming public listings of Unitree/Agibot/Galbot and XPeng Iron’s ramp-up to year-end production targets, which may justify a separate valuation for that segment. The upside from these new business ventures may help weather vehicle headwinds and bolster the valuation to some degree. Hence, we maintain BUY but revise down our TP to US$21/HK$83, based on 1.5x 2026E P/S.