Record October Deliveries Show Volume Momentum, But Margin Quality Remains Unproven

What happened

XPeng reported a record 50,237 deliveries in October 2026, up 22% month-over-month and 20% year-over-year. This print extends a multi-month acceleration and suggests the GX flagship and MONA line are generating sustained order conversion beyond the spring launch window. The previous DeepValue report had flagged weak GX initial deliveries (284 units in May) and a low-ASP MONA mix at ~44%, making volume quality the central risk. October's record volume, while impressive, does not disclose model mix or pricing, so it is premature to conclude margins have stabilized. Management's own commentary about price discounts and downward pressure on gross margin remains the key overhang until Q3 financials are released.

Implication

Investors should treat the record October print as evidence that XPeng's product cadence is working on the volume side, but not as confirmation of margin recovery. The company's ability to scale GX and reduce reliance on the sub-RMB160k MONA remains the pivotal test, and October's data does not provide that breakdown. Until Q3 earnings show vehicle margin holding at or above 12% with a healthier mix, the stock's re-rating potential is capped by the same discount-driven concerns outlined in the prior report. For those already holding, the print supports patience; for those underweight, it may be too early to chase without margin proof. The next catalyst is the Q3 earnings release, where any sequential decline in vehicle margin or continued high MONA mix would negate the positive volume signal.

Thesis delta

The delivery acceleration is now undeniable, shifting the risk from volume shortfall to margin quality. Previously the WAIT rating hinged on Q2 deliveries and GX ramp; October's record 50,237 units clears the volume hurdle with room to spare. However, the margin thesis is unchanged: XPeng must prove it can grow without buying share through discounts, and that evidence only comes with Q3 financials.