Simple AI深朴智能
LatestPre-A (hundreds of millions RMB, Jun 2026, led by DiDi)
Undisclosed (to be confirmed)
Beijing Shenpu Intelligent Technology Co., Ltd. (Simple AI)
Founded by a Tsinghua PhD; '1+2+N' flywheel iteration; a commercial-to-home feedback strategy; the Simple-World world model; the Xiaopu Simbot deployed in hotels; four funding rounds within a year
数据截至2026-08-19
About
Simple AI was founded in Beijing by Li Xiaofei in October 2024. Li is a Tsinghua PhD who studied under academician Li Keqiang, a doyen of autonomous driving, and was co-founder and general manager of the robotics division at Idriverplus, where he led the architecture of L4 autonomous driving technology. The company pursues a '1+2+N' flywheel iteration (one embodied model architecture, two real data pipelines, N real application scenarios) and is among China's earliest embodied AI companies focused on lightweight world models, proposing the unified foundation model Simple-World and a hierarchical memory-augmented agent architecture. Commercialization takes home as the destination and home-like scenarios as the starting point: the Xiaopu Simbot landed the industry's first embodied robot hotel POC at the Beijing Lido Grand Metropark Hotel in partnership with CTG Hotels; it joined Huawei Cloud's Industry AI Dream Factory to co-build the CloudRobo platform, and took global top three in the B-1K challenge (a domestic first). Four rounds of funding within a year, with the June 2026 Pre-A led by DiDi (DiDi's first lead investment in the embodied AI track); it plans to complete batch delivery in 2026 and launch an overseas version plus home/elder-care trial versions in 2027.
Technical Approach
'1+2+N' flywheel iteration: 1 embodied model architecture (Simple-World unified foundation + end-to-end VLA), 2 real data pipelines, N real application scenarios, continuously optimizing the model with scenario data to form a dual technology-and-business flywheel of 'smarter with use, more efficient with delivery'. Taking home as the destination and home-like scenarios as the starting point, accumulating real work data in hotels and elder-care communities to feed home scenario model training.
Commercialization（5）
- With CTG Hotels, landed the industry's first embodied robot hotel scenario POC at the Beijing Lido Grand Metropark Hotel; the Xiaopu Simbot handles item delivery, public area patrol, room service, and basic cleaning
- Working with the Hong Kong Polytechnic University and the Hotel ICON teaching hotel to set embodied AI hotel service standards, while building data compliance frameworks for Hong Kong and overseas markets
- Huawei Cloud INSPIRE 2026: among the first companies to join the Industry AI Dream Factory embodied AI zone, co-launching the ecosystem cooperation program and co-building the CloudRobo closed-loop development platform
- Top three globally in the BEHAVIOR 1K challenge (a domestic first); partnership with Shengshu Technology forming the industry's first embodied AI foundation-model-plus-application ecosystem loop; co-initiating the ECCV 2026 world model workshop
- Overseas expansion: plans to complete the leap from flagship customer trials to batch delivery in 2026; overseas and elder-care/home seed-customer trial versions in 2027
Position & Strengths（5）
- Named an embodied AI dark horse by robot industry media; four rounds (hundreds of millions RMB) within a year is rare in the industry
- Tsinghua plus Idriverplus technical pedigree (student of academician Li Keqiang, architect of L4 autonomous driving systems)
- Lightweight world model technical route endorsed by top capital (DiDi, Linear, Puhua, Baidu Ventures, Shunwei)
- The '1+2+N' flywheel iteration path is seen as a pragmatic commercialization route: commercial scenarios first, then data feedback, then the home endgame
- A global top-three B-1K challenge result validating model generalization