KEY TAKEAWAYS
- Q2 deliveries: 98,330 vehicles, down 11.5% YoY. Revenue: RMB 25.7B, down 15.1% YoY. Vehicle margin: 9.4%, down from 19.4% a year ago. Net loss: RMB 1.7B vs RMB 1.1B net income last year.
- Q3 delivery guidance: 95,000 to 100,000 vehicles, representing 1.9% to 7.3% YoY growth. Revenue guidance: RMB 26.6B to RMB 28.0B. July deliveries of 30,468 confirm the L-Series refresh is ramping. Cash: RMB 87.5 billion.
- In-house MACH M100 chip shipments exceeded 50,000 units. OTA 9.1 improved MACH VLA performance by 20% and nearly doubled urban ADAS mileage penetration. Li i9 flagship BEV and new Li MEGA both launch mid-September. Battery electric vehicles now 50% of total sales.
The vehicle margin story requires some context before the numbers make sense. Li Auto went from 19.4% vehicle margin in Q2 2025 to 9.4% in Q2 2026. On the surface that looks like a business deteriorating rapidly. It isn't. It's a business going through a simultaneous refresh of its entire product line while absorbing semiconductor and battery cost inflation and refusing to pass the cost increase to customers. The CEO was explicit about all three. Whether you find the explanation convincing depends on whether the margins recover in Q3 and Q4 as management expects.
What they reported
Total Q2 revenue of RMB 25.7 billion, down 15.1% year-over-year but up 11.7% sequentially. Vehicle sales of RMB 24.1 billion, down 16.7% year-over-year. Deliveries of 98,330 units, down 11.5% year-over-year and a steep drop from the kinds of volumes the company was generating in strong quarters. Vehicle margin of 9.4%, gross margin of 11.0%. Operating loss of RMB 2.3 billion, operating margin of negative 9.0%. Net loss of RMB 1.7 billion against RMB 1.1 billion net income last year. Free cash flow: negative RMB 1.3 billion, improving substantially from negative RMB 3.8 billion in Q2 2025 and negative RMB 7.4 billion in Q1 2026.
July deliveries: 30,468 vehicles. That's sequential recovery from the Q2 softness, and it's the data point management pointed to as evidence that the refresh transition is now behind them. Q3 guidance of 95,000 to 100,000 units implies July was below the quarterly run rate, meaning August and September need to run at roughly 32,000 to 35,000 per month to hit the midpoint. Achievable, given the new model launches. Q3 revenue guidance of RMB 26.6B to RMB 28.0B represents a return to near-flat year-over-year territory.
The model refresh — what actually happened
Li Auto refreshed its entire L-Series lineup this year, transitioning from the previous platform to a new architecture featuring the in-house MACH M100 chip, 800-volt active suspension, drive-by-wire chassis, and a third-generation range extender with 5C supercharging battery. The L9, L8, and L6 all moved to the new platform sequentially through 2026.
Model transitions create a specific set of financial problems. Old inventory has to clear before new inventory can ramp. New manufacturing lines take time to reach full efficiency. Customers wait for new models rather than buying the old ones. And the cost structure of ramping new products, tooling amortization, new supplier pricing, component changes, runs higher than steady-state manufacturing. All three of those things happened simultaneously at Li Auto in Q2. The CEO named them specifically: "clearing old inventory, ramping up new models, and sales policy transitions have all created short-term operational headwinds."
The L9 Livis variant, the premium drive-by-wire version, is accounting for over 85% of L9 sales since launch. That's a higher-ASP product mix than management expected. The L6 is at nearly 10,000 units per month demand, with the prior generation having delivered almost 400,000 units total, meaning strong brand loyalty carrying over to the new platform. The signals on the refresh are actually positive. The financial results of transitioning through it are not.
The cost inflation nobody is covering well
CEO Xiang Li spent considerable time on the call explaining the commodity and semiconductor cost environment. AI development is driving demand for chips and PCBs, which has pushed component prices up across the industry. Memory chip prices have risen. Lithium carbonate prices have fluctuated cyclically. Li Auto's products are more chip-intensive than average because they carry more intelligence, which means they're disproportionately impacted when chip prices move up.
Management decided not to pass the cost increase to customers. That's why vehicle margin compressed from 19% to 9% rather than from 19% to 15%. The strategic rationale: maintaining price parity while competitors absorb the same cost pressure means Li Auto's market position doesn't deteriorate. The cost will come back through scale on in-house components. The company is now shipping its proprietary MACH M100 chip across all L-Series models. Full in-house battery deployment across all models is expected within a few months. In-house silicon carbide modules for motor controllers are already deployed. Each of those reduces dependency on external suppliers and creates structural cost advantages as volumes scale. The Li Auto investor relations site has the detailed technology roadmap for each component program.
The chip roadmap — the part that's genuinely differentiated
This is the aspect of Li Auto's story that most financial coverage handles superficially. MACH M100 chip shipments exceeded 50,000 units as of the call. The OTA 9.1 update reduced end-to-end latency, improved MACH VLA performance by 20%, and nearly doubled urban ADAS mileage penetration compared to the previous generation computing platform. The Q4 OTA 9.2 update will move to a full 3D Vision Transformer architecture with 3x the parameter count and 4.6x the compute, improving perception range beyond 250 meters. The year-end OTA 9.3 will scale model parameters "exponentially" with enhanced text comprehension and reasoning.
CTO Yan Xie described a three-layer advantage from in-house chip development: hardware interface flexibility that allows chip, algorithm, and OS to be co-optimized; reinforcement learning built around the MACH platform enabling faster model improvement; and Halo OS enabling deep chip-application integration. These aren't marketing claims. They're specific engineering choices with measurable outputs, and the OTA 9.1 performance data supports them. Whether Li Auto can sustain this cadence against NVIDIA Orin-X and Thor-based systems from competitors is the question. For now, the data shows they're competitive and improving faster than alternatives on their own silicon.
The overseas expansion adds a longer-term growth dimension. Li Auto is launching at the Paris Motor Show in October, with the Li i6 entering European sales in Q4. Kazakhstan and Uzbekistan already have the all-new L9. Dubai launches in September. A strategic partnership with Allur for local assembly in Kazakhstan is underway. Right-hand drive Li i6 and Li MEGA are targeting Hong Kong, Singapore, and other RHD markets by year-end.
The one thing beginners should take away
Li Auto is in a transition quarter that produced genuinely bad numbers by any normal metric: a 19-to-9 vehicle margin drop, a net loss after a profitable year, deliveries declining year-over-year. The question isn't whether Q2 was bad. It was. The question is whether the causes are structural or temporary. Model refresh transitions, ramp costs, and commodity cost inflation are all quantifiable and temporary. In-house chips reducing structural cost dependence, the new platform's premium model mix exceeding expectations, and 30,468 July deliveries showing recovery are all specific, measurable evidence that the transition is progressing. RMB 87.5 billion in cash means the company can absorb another couple of quarters like this without existential risk. Whether Q3 and Q4 margin recovery arrives on schedule is what the next two earnings calls will answer.