The monthly delivery reports landed across Saturday morning in China, and they told one story four different ways. Leapmotor delivered 101,267 vehicles in July — up 102% year-on-year, up 8.45% from the 93,376 it managed in June, its fourth consecutive record month and the first time it has cleared 100,000 in a single month. Through seven months it has delivered 457,754 vehicles, 68.42% ahead of last year.
Now the three names an American brokerage account can actually hold. Nio delivered 35,934, up 71% from a year ago but down from 40,597 in June. XPeng delivered 38,027, up 3.57% year-on-year and down 5.23% from June. Li Auto delivered 30,468, against 30,731 in July 2025. Add all three together and you get 104,429 — which means one company most Western investors cannot buy did 97% of the combined volume of the three they can.
The mix inside Nio’s number is the part worth sitting with. Of those 35,934 vehicles, 20,008 wore the premium Nio badge, up 57.9%. The family-oriented Onvo brand did 10,155, up 69.9%. Firefly, the small cheap one, did 5,771 — up 143.9%. The fastest-growing line in China’s best-known premium EV company is its least expensive car, and the two budget sub-brands now account for 44% of the parent’s volume.
Our take: This is not a China story, it is a price story, and the U.S. market is running the same experiment one segment behind. Every one of these companies was built on the thesis that electric vehicles would be sold up-market — software, screens, autonomy, margin. July says the volume is at the bottom and the growth is at the bottom, and the companies winning are the ones that got there first. Leapmotor’s cars undercut the premium field by roughly half on sticker. Nio’s own growth engine is a small hatchback. XPeng’s answer to a 12.8% year-to-date decline is a 123,800-yuan SUV. When four competitors independently conclude that the fix is a cheaper product, that is not a strategy — that is the market telling them the price ceiling is real.
The premium tier stopped compounding two years ago
Li Auto is the clearest case. Its 30,468 is essentially flat against last July’s 30,731 — and last July was itself down 39.74% from the year before. Two years of no growth off a base already 40% below the peak. Cumulative deliveries have reached 1,764,155, and the company spent July launching a new L6 and pushing the L9 past 300,000 units, which is a fine quarter of product work that has not yet moved the monthly line.
XPeng’s July was a miss against expectations rather than against last year. Dealers surveyed by Deutsche Bank in late July looked for as many as 45,000 units; the constraint was supply of the new Mona L03, the low-priced SUV launched globally in Munich on July 16 at 123,800 yuan (about $18,240) with first customer deliveries on July 22. XPeng says nationwide deliveries accelerate in August, and Deutsche Bank puts non-cancellable L03 orders above 50,000 with 13-to-17-week waits on the battery-electric version. Its seven-month total of 204,004 is still down 12.78% year-on-year.
What to watch
- XPeng’s August number. The L03 backlog is the whole thesis. If a 50,000-order book and a supply ramp cannot lift a 38,027 month meaningfully, the problem was never production.
- Leapmotor’s math against its own target. 457,754 through July is 46% of the one million it has publicly stuck to. Hitting it requires roughly 108,400 a month for the final five — above July’s record, every month, with no misses.
- Nio’s brand mix. Premium Nio is 55.7% of volume and falling. Onvo and Firefly carry lower prices and, presumably, lower margins; the September quarter will show what 71% growth costs.
- Where the cheap cars go next. A 100,000-unit-a-month low-cost platform does not stay domestic. Europe’s tariff structure is the only thing standing between that volume and Western dealer lots.
The uncomfortable read for anyone holding these tickers: the three companies listed in New York represent the shrinking half of the market they operate in. The one printing records is listed in Hong Kong and part-owned by a European automaker. Access and exposure are not the same thing.