GOOGL, TSLA, BIDU, PONY, WRD, UBER, AMZN, 002594.SZ · US · A-shares · Hong Kong
Robotaxi industry deep-dive | US–China commercialization (2026H2)
Waymo remains paid-scale anchor; China Apollo/Pony/WeRide grow fleets & overseas faster. Sleeve: GOOGL/BIDU/TSLA options + UBER/BYD shovels + PONY/WRD satellites. See freight.
Cite a section with a deep link, e.g. /en/r/robotaxi-passenger-industry-h2-2026#thesis
Snapshot (~2026 H1)
- Waymo
- ~11 cities / ~500k paid rides/week
- China fleets
- ~4–5k operational vehicles combined
- Overseas ASP
- Gulf ~$1+/km vs China ~¥2/km
- Listed map
- GOOGL·BIDU·PONY·WRD·TSLA·UBER
As-of 2026-09-25 (weekly refresh; equities aligned to §A. Missing series are N/A/null. Not investment advice.)
Industry deep-dive (data ~2026 H1–H2). Not equity §A; fleet/ride metrics vary by source. Not investment advice. Freight companion: autonomous freight industry.
Thesis
Robotaxi entered paid commercial expansion in 2026, but investability still splits into three tracks:
- Scale anchor in the U.S.: Waymo (Alphabet) leads paid rides and miles; capital-intensive vertical ops set the “mature-market” bar.
- Fleet slope in China: Apollo Go / Pony.ai / WeRide expand fleets and cities faster and push overseas (Gulf, Europe, Hong Kong); absolute revenue remains far below Waymo.
- Platforms & bodies take more measurable cash: Uber monetizes distribution; BYD supplies fleets—cleaner near-term KPIs than pre-profit L4 software names.
Allocation: core exposure via parent options (GOOGL / BIDU / TSLA); PONY / WRD as high-β satellites; platforms/OEMs as hedges. Do not linearize profits from fleet counts.
Market stage & demand
| Stage | Traits | 2026 read |
|---|---|---|
| Closed / testing | Safety driver; free or token fares | Still common |
| Paid ODD | Geofence + weather/time limits | Core of Waymo / Apollo / Pony / WeRide |
| National rules | Unified permits, insurance, liability | Incomplete in both CN & US |
| Positive unit contribution | After subsidies | Claimed by leaders; group NI still distant |
Demand: replace driver wage/benefits and peak shortages. Gulf fares ~$1.0–1.1/km vs ~RMB 2/km in China — same stack, different profit pools. Trust and accident risk still cap U.S. expansion slopes.
Value chain
Sensors/compute → AD stack → purpose-built or OEM vehicles → fleet ops → ride apps → riders.
| Layer | Who gets paid | Risk |
|---|---|---|
| Sensing/compute | Sensors, NVIDIA, domain controllers | Spec wars, export controls |
| L4 software | Waymo / Apollo / Pony / WeRide / Tesla | Burn, liability |
| Vehicle | BYD, OEM partners, Zoox purpose-built | Fleet bargaining, residual value |
| Ops | Operator or local mobility JV | Utilization, maintenance, tele-assist |
| Distribution | Uber / Lyft / Didi / own app | Take-rate vs brand |
Early customers concentrate in a few permit cities; overseas quota (e.g. Dubai AV caps) is both a constraint and a moat. See [[#risks]].
United States
Regulation by region
| Region | Status | Implication | Watch |
|---|---|---|---|
| 美国（州际为主） | 加州/亚利桑那/德州等商用牌照并行；联邦框架仍碎片化 | Waymo 扩城靠州许可堆叠；Tesla 受州数据库与安全要求约束 | 联邦 AV 规则、事故披露、保险责任 |
| 中国 | 城市试点牌照拼图；北京/广州/武汉等付费示范领先 | 车队扩张快于美国，但全国统一 L4 规则仍在整合预期中 | 跨城互通、无人化比例、地方补贴退坡 |
| 海湾 / 中东 | 迪拜/阿布扎比已有无人商用；配额与牌照成稀缺 | 单价（约 $1+/km）显著高于国内，改善单车经济 | 配额上限、本地合资要求 |
| 欧洲 / 东亚其他 | 伦敦测试、香港 RHD 测试、部分东欧商用试点 | 中国运营商出海差异化；右舵/左行是技术与合规门槛 | 公开道路商用时间表、与 Uber/Lyft 分发绑定 |
Waymo: ~10+ cities, ~3k vehicles, ~500k paid rides/week (tracker ranges). Vertical model; maps through GOOGL.
Tesla Robotaxi: vision-first, multi-metro launch; fleet still far below Waymo; state disclosure and safety rules set the slope. Maps to TSLA optionality.
Zoox / others: Amazon Zoox small commercial footprint; May Mobility–Lyft style partnerships show platform distribution path.
China & overseas
| Operator | Fleet/city snapshot | Commercial note |
|---|---|---|
| Pony.ai | >1,700 EoQ1; 2026 target ~3,500 | Robotaxi revenue up sharply YoY; unit-economics narrative |
| WeRide | China >1,000; multi-country | FY25 Robotaxi revenue still small; overseas ASP higher |
| Apollo Go | 10+ CN cities; ~28-city global footprint claim | Dubai driverless commercial; London/HK testing |
| Momenta et al. | More demo permits | Production-car data flywheel vs dedicated operators |
China operational fleets ~4–5k combined estimates mid-2026 — fleet ≠ revenue. Overseas matters because Gulf pricing can be ~3× domestic and RHD/left-hand traffic raises the bar.
Unit economics
Daily contribution ≈ paid km × net $/km − energy − maintenance − insurance − tele-assist − depreciation
Drivers: utilization, BOM/residuals, driverless ratio, and asset ownership (Waymo vertical vs China asset-light JV where tech takes service fees). Until group NI turns, falsify with paid rides, unit contribution, cash burn.
Listed map
Operators & listings
| Name | Role | Key metric | Note |
|---|---|---|---|
| Waymo (GOOGL) | 美国 Robotaxi 规模锚 | ~11 城 / ~3k 车 / ~50 万付费单/周 | 付费里程与收入领先；资本密集垂直模式 |
| 百度 Apollo Go (BIDU) | 中国多城 + 出海 | 10+ 国内城；全球足迹约 28 城口径 | 母公司稀释；单量与贡献利润并行跟踪 |
| Pony.ai (PONY) | 中国 L4 纯 play（客运为主） | Q1'26 车队 >1700；目标抬至 ~3500 | 单车经济叙事强；高波动 / 地缘敏感 |
| WeRide (WRD) | 多形态 L4 + 出海 | 中国车队 >1000；多国商用 | Robotaxi/bus/van；产品线分散双刃剑 |
| Tesla Robotaxi (TSLA) | 视觉路线 + 扩城叙事 | 多城已开；车队仍远小于 Waymo | 监管与安全员政策是斜率关键 |
| Uber / Lyft | 分发与接单平台 | 与 Waymo / Apollo / WeRide 等合作 | 赚平台费；非纯 L4 软件 beta |
| Name | Role | Note |
|---|---|---|
| GOOGL | Waymo | equity note |
| TSLA | Robotaxi + FSD | High narrative β |
| BIDU | Apollo Go | Ads/cloud dilute |
| PONY | CN L4 pure-play | Satellite |
| WRD | Multi-form + overseas | Small revenue base |
| UBER | Distribution | “Picks & shovels” |
| BYD | Vehicle supply | equity |
Freight: autonomous freight.
Scorecard
Scores 1–5, subjective.
| Name | Commercial | Unit econ | Regulation | Position | Valuation | Risk | Total | Role |
|---|---|---|---|---|---|---|---|---|
| GOOGL (Waymo) | 5 | 4 | 4 | 5 | 3 | 4 | 25 | Scale anchor |
| BIDU (Apollo) | 4 | 3 | 4 | 4 | 3 | 3 | 21 | CN + overseas |
| PONY | 4 | 4 | 3 | 4 | 2 | 2 | 19 | High-β satellite |
| WRD | 4 | 3 | 3 | 3 | 2 | 2 | 17 | Overseas satellite |
| TSLA | 3 | 2 | 3 | 3 | 2 | 2 | 15 | Narrative option |
| UBER | 4 | 4 | 4 | 4 | 3 | 4 | 23 | Distribution |
| BYD | 3 | 3 | 4 | 4 | 3 | 4 | 21 | Vehicle supply |
Portfolio framework
| Sleeve | Example weight | Tools |
|---|---|---|
| Core options | 40–60% | GOOGL / BIDU |
| Distribution & bodies | 20–35% | UBER, BYD |
| L4 satellites | 0–20% | PONY / WRD / TSLA slice |
| Cash/hedge | Remainder | Avoid levered options as “core” |
Outlook & tracking
Base case 2026H2–2027: Waymo keeps paid-scale lead; China fleets push toward multi-thousand units with more overseas cities; Tesla expands but lags share; unit contribution improves while most groups stay loss-making.
Falsifiable watches: (1) Waymo weekly paid rides/cities; (2) Pony/WeRide fleet vs cash burn vs unit-econ claims; (3) Apollo Go overseas revenue recognition; (4) Tesla state registrations & safety-driver policy; (5) major accidents / permit freezes.
Risks
| Risk | Level | Note |
|---|---|---|
| Accidents & trust | High | One event can freeze permits/multiples |
| Regulatory fragmentation | High | State/city shocks; delayed federal rules |
| Dilution | High | Pure-plays keep raising equity |
| Price wars / take-rates | Medium | Overlapping CN cities; platform fees |
| Geopolitics / export controls | Medium | Chips, data, ADR policy |
| Parent dilution | Medium | Core business noise |
| Stack divergence | Medium | Lidar-heavy vs vision; ODD failures |
References
- Axios robotaxi rankings — https://www.axios.com/2026/06/17/robotaxi-rankings-waymo-china
- Presenc AI Robotaxi Expansion Tracker 2026 — https://presenc.ai/research/robotaxi-expansion-tracker-2026
- Axis Intelligence Robotaxi Statistics 2026 — https://axis-intelligence.com/robotaxi-statistics/
- iEVChina China Robotaxi Fleet Tracker H1 2026 — https://ievchina.com/ai-mobility/china-robotaxi-fleet-tracker-h1-2026-cities-vehicles-revenue-weride-pony-baidu/
- Baidu Q2 2026 results — https://ir.baidu.com/news-releases/news-release-details/baidu-announces-second-quarter-2026-results
- On-site: freight · GOOGL · BYD · JD · Meituan
Not investment advice. Metrics may differ by definition; prefer latest IR.
Comments
Sign in to comment