A Shareholder Critique of XPeng, and What the Numbers Say

A summary of a Chinese shareholder-group discussion has been circulating on Weibo, and it is worth reading not because it is right — much of it cannot be checked — but because it is the kind of document a company's investor relations team is ultimately judged on.

The post, from 姬永锋, quotes the account's own earlier summary of "the issues discussed in the group these past few days". Three complaints:

1. Investor relations. The company did not communicate well with shareholders through the share-price decline, and He Xiaopeng did not make even a symbolic share purchase.
2. Domestic sales are poor, down sharply year on year, with the problem concentrated in older models — while conceding that the X9, P7+ and G6 are competitive in their respective segments.
3. A forecasting miss on the GX. The summary says the top trim's demand was mis-estimated, which has affected deliveries.

The post's own comment adds the framing that has travelled with it: capital consensus and market sentiment are fickle; product structure and aesthetic design are what must change; the "information cocoon" and IR's attitude are the hard parts; and the AI and overseas strategy is what does not change.

That is sentiment, not reporting — the post has no engagement, the sources are unnamed group members, and the numbers behind the claims are not given. So here is what the public data actually shows.

What the numbers say

Deliveries were not collapsing, sequentially. XPeng's own release on 1 October reported 41,256 deliveries for September, up 5% month on month, taking the third quarter to 118,390, up 15% quarter on quarter. That does not refute the complaint — a month-on-month rise says nothing about the year-on-year comparison, and the domestic/export split is what the group is really arguing about — but "sales are very poor" needs that context.

The whole market had a bad September. Citi's read was that Chinese automakers' EV orders fell 22% week on week, with full-month retail heading for a 7–12% year-on-year decline. A weak September in China is a sector event, not an XPeng one. Both things can be true: the market is down, and XPeng is underperforming it domestically.

The financial picture is the real source of the share-price argument. In the first half, XPeng's revenue fell 3.8% to ¥32.78 billion, gross margin improved to 20.6%, and the attributable net loss widened 173% to ¥3.12 billion, against a debt-to-asset ratio of 73.4%. The second-quarter breakdown is starker: a 12.1% gross margin on vehicles, against 75.1% on the services segment that carries carbon-credit trading and technology services. Exports are the growth story — 8,125 units in August, up 169% year on year, with 49,424 in eight months, already past all of 2025 — but the domestic car business is where the loss is made.

The "product structure" complaint has already been answered

This is the most useful thing to note about the thread's timing. On 28 September, XPeng consolidated its four product lines into two, merging the F (P-series) and I (overseas products) lines into G, leaving G for everything above the MONA band and D for MONA. The stated rationale was exactly the criticism in the post: concentrate R&D, cut cost, and stop models competing with each other on positioning and price.

A company that reorganises its product lines four days before a shareholder group complains about product structure is not ignoring the complaint. What the thread is really about is pace and communication — whether shareholders can see the plan working before the next reporting date.

The two claims that cannot be settled from outside

• "He Xiaopeng did not buy shares." We found no insider-purchase disclosure either way, so the honest version is "no purchase has been disclosed" rather than "he did not buy". A founder buying into a decline is a signal; its absence is not proof of anything on its own.
• "IR does not communicate." This is a judgement about responsiveness, and it is unfalsifiable from outside the room. It is also the complaint that management can fix fastest, which is presumably why it appears first.

What would settle it

• October and November delivery numbers, split between domestic and export.
• Third-quarter results, and whether the product-line consolidation shows up as cost per model or margin.
• Any founder or insider purchase — a disclosure question, not an opinion.
• The overseas ramp of the L03 and G9L, which is where the growth and the awards have been coming from (the X9 just took the Luxury class in German Car of the Year 2027, a useful counterweight to the domestic gloom).

Caveats

• This is a shareholder-forum summary, not reporting. The three complaints come from unnamed group members; the numbers behind them are not provided.
• The GX forecasting claim is an inference by the group. We could not verify a specific top-trim demand miss, and a long wait for a high-spec trim is not on its own evidence of a forecasting error.
• No share price is quoted here deliberately: we could not date a live quote reliably, and a stale price would be worse than none.
• Nothing in the thread names a source for the domestic sales decline. The market data above is from Citi via AAStocks; the delivery figure is XPeng's own September number.
• Correction, 1 October 2026: an earlier version of this article cited a figure of 41,581 deliveries for September (+10.3% m/m). That was XPeng's September 2025 number. XPeng's own release for September 2026 gives 41,256 (+5% m/m), and the quarter reconciles with it exactly (41,256 + ~39,291 + ~37,843 = 118,390). The company's figure is used above — which makes the year-on-year comparison marginally negative, not positive.

The bottom line

Strip out the sentiment and the thread is a fair list of the three things management actually controls — how it communicates, how focused its product plan is, and whether the founder puts his own money behind it — next to the two it does not (capital mood and the market), and the one strategic bet it has already made and cannot hedge: AI and going overseas. The first three are judged quarterly. The last one is judged in years, which is precisely the mismatch that produces posts like this.

Source: 姬永锋 (Weibo), with market data from Citi via AAStocks and financials from our earlier coverage — translated & adapted for XPENGTech.