XPEV · Forward model · He Xiaopeng case
The He Xiaopeng case, 20 quarters out
Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.
Built in US dollars at the single disclosed convenience rate of RMB6.79 to US$1.00, the 30 June 2026 H.10 rate XPeng uses in the results release. Every historical quarter here is the reported RMB figure divided by that one rate, so year-over-year growth in this model is RMB growth. That is deliberate and it does NOT match data/companies/xpev/series.json, which carries each quarter's own convenience translation at rates that moved from RMB7.2993 to RMB6.7852 over the tracked window - on that basis the basis quarter looks like 14% year-over-year growth when the reporting-currency growth was 8.0%. XPeng reports exactly two revenue lines, Vehicle sales and Services and others, and they sum to total revenues to the thousand in all eight quarters carried here, so no allocation is estimated and no quarter is marked estimated. It has one reportable segment under ASC 280 and the CODM reviews only consolidated revenue and gross profit, so there is no per-vertical operating profit: the margin field on each vertical is that line's DISCLOSED GROSS MARGIN, and all R&D and SG&A sit in corporate overhead at 27.4% of revenue - the basis quarter's 27.41% and the trailing four quarters' 27.57%. The consequence is honest and uncomfortable: the model's blended gross margin reaches about 24% in the final quarter against that fixed 27.4% overhead load, so the base case narrows the operating loss from 7.4% of revenue in the first projected quarter to 3.4% in the last and never crosses into profit. The valuation therefore rests entirely on the exit multiple. There is no robotics or robotaxi revenue line anywhere in this model. Dogotix has a filed US$6.3bn post-transaction valuation and the CEO has stated 2027 humanoid volumes, but no robot price, no robot revenue and no fleet size is disclosed, so that story appears only as an exit multiple in the He Xiaopeng case. Depreciation and amortisation are not disclosed quarterly, so the line labelled EBITDA here is gross profit less corporate overhead - an operating figure before any D&A add-back, which at the basis quarter is about RMB1.33bn negative against a reported loss from operations of RMB1.14bn; the difference is other operating income the model does not project. One limitation is worth naming plainly: the base case burns roughly US$0.35-0.44bn of free cash flow every quarter for twenty quarters - about the rate XPeng actually burned in the first half of 2026, RMB11.72bn - against net cash of US$3.05bn. A company on that path raises capital, and this model does not dilute for it, so the per-ADS fair value is generous to that extent in every case shown.
The CEO's physical-AI case, and the first time it carries an outside price. On 24 August 2026 Dogotix signed a US$900m Series A at an implied US$5bn pre-transaction and US$6.3bn post-transaction valuation, leaving XPeng about 68.41% on the fullest-dilution basis - a stake worth about US$4.31bn inside an enterprise value of US$7.6bn to US$8.2bn. On the same call He said IRON reaches mass production at scale by the end of 2026, ramping to several thousand units a month in 2027, and that driverless robotaxi passenger operations start in 2027. Note what this case does NOT do: it adds no robot revenue, no robot price and no fleet, because none is disclosed. It moves only the exit multiple, to 1.6x, which is what paying for the private mark on top of the car company looks like. Two cautions travel with it: US$200m of the US$900m was subscribed by a group entity and US$100m more by entities owned by the CEO and the co-president, so external money is US$600m; and the model still never earns an operating profit.
Latest: $4.70B (2031Q2E)
| Period | Value |
|---|---|
| 2024Q3 | $1.49B |
| 2024Q4 | $2.37B |
| 2025Q1 | $2.33B |
| 2025Q2 | $2.69B |
| 2025Q3 | $3.00B |
| 2025Q4 | $3.28B |
| 2026Q1 | $1.92B |
| 2026Q2 | $2.91B |
| 2026Q3E | $3.32B |
| 2026Q4E | $3.55B |
| 2027Q1E | $3.69B |
| 2027Q2E | $3.78B |
| 2027Q3E | $3.86B |
| 2027Q4E | $3.93B |
| 2028Q1E | $3.99B |
| 2028Q2E | $4.05B |
| 2028Q3E | $4.10B |
| 2028Q4E | $4.16B |
| 2029Q1E | $4.21B |
| 2029Q2E | $4.27B |
| 2029Q3E | $4.32B |
| 2029Q4E | $4.37B |
| 2030Q1E | $4.43B |
| 2030Q2E | $4.48B |
| 2030Q3E | $4.53B |
| 2030Q4E | $4.59B |
| 2031Q1E | $4.65B |
| 2031Q2E | $4.70B |
What drives each segment
Vehicle sales
Units × priceA clean volume-times-price line: XPeng publishes total deliveries every month and every quarter, so dividing vehicle revenue by deliveries gives an ASP with no allocation. The constraint is volume, not price - deliveries were 103,295 against 103,181 a year earlier - and the vehicle gross margin has fallen for three straight quarters to 12.1% on what management calls the product generation transition. Growth from here has to come from overseas volume and from the GX and MONA L03 mix, not from raising prices in China.
Latest: $3.78B (2031Q2E)
| Period | Value |
|---|---|
| 2024Q3 | $1.30B |
| 2024Q4 | $2.16B |
| 2025Q1 | $2.12B |
| 2025Q2 | $2.49B |
| 2025Q3 | $2.66B |
| 2025Q4 | $2.81B |
| 2026Q1 | $1.62B |
| 2026Q2 | $2.51B |
| 2026Q3E | $2.90B |
| 2026Q4E | $3.11B |
| 2027Q1E | $3.23B |
| 2027Q2E | $3.31B |
| 2027Q3E | $3.36B |
| 2027Q4E | $3.40B |
| 2028Q1E | $3.44B |
| 2028Q2E | $3.48B |
| 2028Q3E | $3.51B |
| 2028Q4E | $3.54B |
| 2029Q1E | $3.57B |
| 2029Q2E | $3.59B |
| 2029Q3E | $3.62B |
| 2029Q4E | $3.64B |
| 2030Q1E | $3.67B |
| 2030Q2E | $3.69B |
| 2030Q3E | $3.71B |
| 2030Q4E | $3.74B |
| 2031Q1E | $3.76B |
| 2031Q2E | $3.78B |
Assumptions & reasoning
- ASEASONAL, after testing for a season and rejecting it. Ratio-to-centred-four-quarter-moving-average on the fourteen quarters of consolidated revenue in data/companies/xpev/series.json (2023 Q1 to 2026 Q2; vehicle sales is 86-91% of that line) gives normalised factors of 0.82 for Q1, 0.89 for Q2, 1.00 for Q3 and 1.28 for Q4 - a peak-to-trough signal of 0.455. The window-to-window spread is 0.257 for Q1, 0.178 for Q2, 0.164 for Q3 and 0.373 for Q4, so on every single quarter the spread is larger than that quarter's own distance from 1.0. This line's own eight quarters of split history yield exactly one window per quarter index, so their spread is not merely wide, it is unmeasurable.
- The shape also fails out of sample on the most recent complete cycle, which is what settled it. Applied to deliveries, those factors predict 147,700 units in 2025 Q4 against 116,249 actually delivered, and 95,000 in 2026 Q1 against 62,682. On this line's own history the seasonal shape removes only about a tenth of the residual variance around a log-linear trend (RMSE 0.223 aseasonal against 0.201 with the best candidate set); on deliveries it removes about a fourteenth. Almost all of the variance is the 2024 MONA M03 launch and the 2026 Q1 collapse, and neither is a season.
- The cost of leaving it aseasonal is stated rather than hidden: the model cannot reproduce the March-quarter trough. 2026 Q1 deliveries of 62,682 were 39% below the quarter before, so the projected 2027 Q1 prints a year-over-year gain that is mostly a base effect, and projected FY2027 revenue grows faster than the business does for the same reason. Read the annual totals from 2028 onward, not the 2027 comparison.
- No per-vertical operating profit exists. XPeng has one reportable segment under ASC 280 and the CODM reviews consolidated revenue and gross profit only, so R&D and SG&A stay at corporate and the margin field on this line is the disclosed VEHICLE GROSS MARGIN of 12.08%, not an EBITDA margin.
- Capex intensity of 4.77% is derived, not disclosed: XPeng publishes no quarterly capex, only FY2025 capital expenditures of RMB3,347.1m. Holding the services line at an assumed 1.0% and solving the remainder against FY2025 vehicle revenue of RMB68,378.9m gives 4.77%, so total modelled capex reconciles to the disclosed annual figure rather than falling 8% short of it.
- Overseas revenue is disclosed only as a six-month geographic total covering both revenue lines - RMB8.23bn in H1 2026 against RMB5.07bn in H1 2025 - so the overseas ramp that management sized at above 20,000 deliveries in the basis quarter and above 40,000 by 2026 Q4 cannot be carved out as its own sub-line here. It is inside this line's volume growth, and the base case does not assume it all lands.
Services and others
Growth pathA 75%-gross-margin line that nearly doubled year over year and carried the entire consolidated gross-margin improvement while vehicle margin fell. XPeng has never sized any component of it - technical R&D services for Volkswagen, parts, supercharging, financing and insurance, carbon credits - so it gets a growth driver and nothing more. Its level is hostage to Volkswagen milestone recognition, which the FY2025 auditor flagged as a critical audit matter.
Latest: $926M (2031Q2E)
| Period | Value |
|---|---|
| 2024Q3 | $192M |
| 2024Q4 | $211M |
| 2025Q1 | $212M |
| 2025Q2 | $205M |
| 2025Q3 | $343M |
| 2025Q4 | $469M |
| 2026Q1 | $300M |
| 2026Q2 | $397M |
| 2026Q3E | $417M |
| 2026Q4E | $437M |
| 2027Q1E | $458M |
| 2027Q2E | $479M |
| 2027Q3E | $501M |
| 2027Q4E | $523M |
| 2028Q1E | $546M |
| 2028Q2E | $569M |
| 2028Q3E | $594M |
| 2028Q4E | $619M |
| 2029Q1E | $645M |
| 2029Q2E | $672M |
| 2029Q3E | $700M |
| 2029Q4E | $728M |
| 2030Q1E | $758M |
| 2030Q2E | $789M |
| 2030Q3E | $821M |
| 2030Q4E | $855M |
| 2031Q1E | $890M |
| 2031Q2E | $926M |
Assumptions & reasoning
- ASEASONAL. Ratio-to-centred-four-quarter-moving-average on this line's own eight quarters gives 0.93, 0.73, 1.05 and 1.30, but there is exactly one window per quarter index, so the window-to-window spread is not one point wide - it is unmeasurable. The shape is also indistinguishable from what actually happened: a single level shift when Volkswagen technical R&D milestones started landing in 2025 Q3, taking the line from RMB1.39bn to RMB2.33bn and then RMB3.18bn. A ramp seen once is not a season, so no factors are carried.
- The company has never sized any component of this line. The only sub-disclosure is an H1 2026 recognition-timing split - RMB1.77bn at a point in time against RMB2.65bn over time - which is a revenue-recognition cut, not a product cut, and must not be used as a proxy for a product split.
- The basis quarter grew 32.6% sequentially, and the driver deliberately does not start there. Volkswagen technical R&D revenue is recognised on progress toward completion using the input method, which the FY2025 auditor flagged as a critical audit matter: the quarterly level depends on management's estimate of total contract costs and is lumpy by construction, so a strong quarter can be followed by a weak one with nothing changing underneath.
- The margin field is the disclosed SERVICES AND OTHERS GROSS MARGIN of 75.07%, not an EBITDA margin, for the same ASC 280 reason as the vehicle line. It is the highest in the disclosed history and the terminal 65% assumes the near-100%-margin milestone revenue fades into parts, charging and financing.
- Capex intensity of 1.0% is assumed, not disclosed. XPeng publishes no capex split; this line is largely people and intellectual property, and the consolidated capital programme is attributed to the vehicle line instead.
Where each case comes from
Bear case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Bear column is what happens if they are taken at face value.
Interim results announcement, six months ended 30 June 2026
Bull case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Bull column is what happens if they are taken at face value.
He Xiaopeng case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the He Xiaopeng column is what happens if they are taken at face value.
Dogotix subscription and connected transaction announcement
Second quarter 2026 earnings call, 24 August 2026
From cash flow to fair value
| Present value of free cash flow, 20 quarters | -$5.98B |
| Terminal-year revenue | $18.48B |
| Terminal-year EBITDA | -$703M |
| Exit multiple, on revenue | 1.6x |
| Terminal value | $29.56B |
| Discounted at 13.0% a year, terminal value becomes | $16.04B |
| Enterprise value | $10.07B |
| Net cash | $3.05B |
| Equity value | $13.12B |
| Shares | 0.96B |
| Fair value per share | $13.71 |
| Against the current price of $11.33 | +21% |
0.9x terminal revenue, discounted at 13%. The reference is the tape. XPeng closed the results session of 24 August 2026 at $11.15, down 8.5% on the day, and $11.71 two sessions later; on 956.97m ADS that is a market capitalisation of $10.67bn to $11.21bn, and taking off net cash of RMB20.72bn ($3.05bn at the disclosed RMB6.79 rate) leaves an enterprise value of $7.62bn to $8.16bn against trailing revenue of RMB75.41bn ($11.11bn). So the market is paying 0.69x to 0.73x trailing revenue right now, and 0.9x assumes a modest re-rating as losses narrow, nothing more. No peer multiple is used: NIO and Li Auto are not tracked here and no verified quote for either was obtained, which the research brief records as an open question rather than filling with an unverified number. The 13% discount rate is the highest on this site and is meant to be - a loss-making China ADR that burned RMB11.72bn of operating cash in the first half, lifted borrowings to RMB19.76bn and moved its own gearing ratio from 41.8% to 73.2% in six months. This multiple decides more of the answer than every operating slider combined, because the model never turns free cash flow positive in the base case.
Read the other way round: at $11.33 the market is paying 1.4x terminal-year revenue, holding every other assumption on this page fixed. That is the number to argue about.
The projected path
| Quarter | Vehicle sales | Services and others | Revenue | YoY | EBITDA | Capex | FCF | R40 | PV of FCF |
|---|---|---|---|---|---|---|---|---|---|
| 2026 Q3E | $2.90B | $417M | $3.32B | +11% | -$246M | $142M | -$388M | -1 | -$376M |
| 2026 Q4E | $3.11B | $437M | $3.55B | +8% | -$267M | $152M | -$419M | -3 | -$394M |
| 2027 Q1E | $3.23B | $458M | $3.69B | +92% | -$275M | $157M | -$431M | +81 | -$394M |
| 2027 Q2E | $3.31B | $479M | $3.78B | +30% | -$276M | $160M | -$436M | +19 | -$385M |
| 2027 Q3E | $3.36B | $501M | $3.86B | +16% | -$273M | $162M | -$436M | +5 | -$374M |
| 2027 Q4E | $3.40B | $523M | $3.93B | +11% | -$270M | $164M | -$434M | +0 | -$361M |
| 2028 Q1E | $3.44B | $546M | $3.99B | +8% | -$265M | $166M | -$431M | -3 | -$348M |
| 2028 Q2E | $3.48B | $569M | $4.05B | +7% | -$260M | $167M | -$427M | -4 | -$334M |
| 2028 Q3E | $3.51B | $594M | $4.10B | +6% | -$254M | $168M | -$422M | -4 | -$321M |
| 2028 Q4E | $3.54B | $619M | $4.16B | +6% | -$248M | $170M | -$417M | -4 | -$307M |
| 2029 Q1E | $3.57B | $645M | $4.21B | +6% | -$241M | $171M | -$412M | -4 | -$294M |
| 2029 Q2E | $3.59B | $672M | $4.27B | +5% | -$234M | $172M | -$406M | -4 | -$281M |
| 2029 Q3E | $3.62B | $700M | $4.32B | +5% | -$226M | $173M | -$399M | -4 | -$268M |
| 2029 Q4E | $3.64B | $728M | $4.37B | +5% | -$218M | $174M | -$392M | -4 | -$256M |
| 2030 Q1E | $3.67B | $758M | $4.43B | +5% | -$210M | $175M | -$385M | -4 | -$243M |
| 2030 Q2E | $3.69B | $789M | $4.48B | +5% | -$201M | $177M | -$377M | -3 | -$231M |
| 2030 Q3E | $3.71B | $821M | $4.53B | +5% | -$191M | $178M | -$369M | -3 | -$219M |
| 2030 Q4E | $3.74B | $855M | $4.59B | +5% | -$181M | $179M | -$360M | -3 | -$208M |
| 2031 Q1E | $3.76B | $890M | $4.65B | +5% | -$171M | $180M | -$351M | -3 | -$196M |
| 2031 Q2E | $3.78B | $926M | $4.70B | +5% | -$160M | $181M | -$341M | -2 | -$185M |
Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.
Model revisions
Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.
| Date | Changed | Fair value then | Note |
|---|---|---|---|
| 2026-08-27 | all | $6.37 | First publication, built on the 2026 Q2 results filed 24 August 2026. Two disclosed revenue lines, a unit driver on vehicles calibrated to the disclosed 2026 Q3 guidance midpoint, a growth driver on services and others, and both lines carried aseasonal after a centred-moving-average test found the seasonal shape smaller than the spread between windows. |