Sudu Technology
A Shanghai-based general-purpose embodied intelligence developer co-founded by ImageNet contributor Prof. Hao Su and serial entrepreneur Zheng Han — a simulation-first robotics bet that reached a $2B+ valuation within eleven months of incorporation
Shanghai Sudu Technology Co., Ltd. (“Sudu”) is a privately held embodied-AI company incorporated in Shanghai in May 2025. It pursues full-stack in-house development across models, data, simulation, robot hardware and control systems. The founding structure pairs one of the field’s most cited academics (Hao Su) with a two-time exit founder in hardware and platform businesses (Zheng Han). We view this as a combination of academic infrastructure assets and commercial execution capacity, and we regard the team’s scientific credibility as high relative to the cohort of Chinese embodied-AI companies that crossed the RMB 10B valuation line in 2026.
Undergraduate and doctoral degrees from Beihang University (BUAA) and a Stanford Ph.D. in computer science. Joined the ImageNet project in 2008 under Fei-Fei Li; led ShapeNet and PointNet during his Stanford doctorate. Joined UCSD in 2018, became tenured associate professor and embodied-AI lab director, and built the SAPIEN and ManiSkill platforms. Returned to China in April 2026 as Distinguished Professor at Fudan University and dean of its General Physical Intelligence institute. Citations of roughly 150,000 per press reports.
BUAA graduate with prior research experience at Microsoft Research Asia. Founded smart-wearables company ZEPP (later acquired by Huami) and co-founded smart-office platform Rocket Technology (later acquired by UCommune). Sudu is his third venture. At the August 2026 World Robot Conference he argued that scaled deployment depends on high-reliability generalization within a limited set of valuable skills.
Former investor at Source Code Capital, reportedly involved in that firm’s investment in Unitree. We read the appointment as importing an investor’s view of robot supply chains and bill-of-materials discipline into the hardware roadmap.
Prior roles at ABB, Huawei and BlueRun Ventures; invested in several embodied-AI and advanced-manufacturing companies before returning to operating roles. We would expect him to anchor industrial-customer access and capital-markets communication.
(1) Title discrepancy. Hao Su is described as co-founder, chairman and CTO in a Sep 30, 2026 ZhiDongXi report, but as chief technical advisor in Leiphone (Jun 2026) and QbitAI (Apr 2026) coverage. Registered corporate titles and shareholding are not disclosed.
(2) Key-person and conflict-of-interest exposure. Prof. Su holds a concurrent Fudan University deanship, and the allocation of intellectual property between the university and the company has not been disclosed. A Fudan-affiliated fund (Fudan Sci-Tech) participated in the Pre-A. Until the structure is clarified, we assume a conservative scope for academic assets attributable to the company.
Sudu positions itself as a developer of a general physical-intelligence base that transfers and is reused across tasks and scenes, rather than a vendor of single-task solutions. Its in-house hardware-plus-software system, #Sudo R1, unveiled in April 2026, integrates a 3D world model with reinforcement learning and is trained on simulated data only, with no real-robot demonstrations. Real-world data collection cost is the industry’s most commonly cited bottleneck; if the approach holds, we believe it implies a structural advantage in unit economics and development cycle time.
| Date | Milestone | Stage | Our Comment |
|---|---|---|---|
| May 2025 | Incorporated in Shanghai | Founding | Launched as an academic-plus-serial-founder structure |
| Apr 2026 | Sudo R1 unveiled; $500M Pre-A announced | Tech Reveal | First external validation attempt via technical blog and 60-minute video |
| Jun 2026 | First on-site hardware showing at ICRA 2026 (Vienna) | Public Demo | Grasping of varied materials and shapes at a leading academic venue |
| Jul–Aug 2026 | WAIC and WRC presentations: bimanual, deformable-object and precision-assembly skills | Capability Expansion | Task difficulty moving from grasping toward assembly, a more industrially relevant frontier |
| Sep 30, 2026 | New round of more than RMB 1B announced | Capital Raise | Next test is scaled verification across industries, customers and production stations |
Commercialization positioning: The company is reported to be collaborating with CATL on battery production and logistics scenarios and to be building a multi-station robot system. Management says proceeds will fund full-stack technical work and scaled validation in real industrial settings, specifically whether one general intelligence base can transfer, run stably over the long term and be reused across lines. We read this as the company entering the transition from lab prototype to industrial productivity.
(1) Performance figures are self-reported. The 98% and near-100% success rates rest on the company’s technical blog and company-produced video. We have found no independent verification or public benchmark data. Test environment, object set and robot configuration would need to be disclosed for comparability.
(2) No revenue, order or customer disclosure. Revenue, backlog, units delivered and pricing are not disclosed, and CATL is the only named counterparty. The contractual nature of that collaboration (paid versus proof-of-concept) is unclear. We believe the current valuation reflects a premium on technology, team and capital structure rather than commercial traction.
Two large rounds are publicly confirmed: the April 2026 $500M Pre-A and a round of more than RMB 1B announced on September 30, 2026. The Pre-A post-money valuation exceeded $2B within eleven months of incorporation, making Sudu one of the headline names among sixteen embodied-AI companies that joined China’s RMB 10B valuation club in the first half of 2026. The investor base combines platform capital (Alibaba, Tencent, Ant), industrial capital (CATL), established venture firms (Hillhouse, IDG, BlueRun) and Shanghai state-linked capital. We read this mix as a three-part positioning across technical validation, industrial demand and policy support.
Leiphone reports a 2024 seed round, followed by multiple rounds in 2025 with roughly twenty funds competing for allocation, but round names, amounts and investors are not disclosed. Because the date precedes the May 2025 incorporation, a predecessor entity or a reporting error is possible.
Multiple rounds after incorporation are reported, but no official announcement has been located. Xiaomi, named as an existing shareholder in the September 2026 round, does not appear in the April Pre-A investor list, so we infer it invested in this earlier period (our inference, unconfirmed).
Post-money valuation above $2B (approx. RMB 13.4–13.6B), announced alongside the Sudo R1 reveal. Platform, industrial, venture and state-linked capital converged in a single round.
Co-led by Shanghai Guotou Xiandao, Guotai Haitong and Futeng Capital, with participation from Daohe Long-Term, Dongfeng Asset, China Development Bank Capital, Langmafeng and Liangtu Zhice, and follow-on investment from existing shareholders including Xiaomi. Proceeds are earmarked for full-stack technical work and scaled validation in real industrial settings. Futeng, an investor in the prior round, returned as a co-lead.
(1) Valuation cannot be cross-verified. The Pre-A post-money is converted to RMB 13.408B in ZhiDongXi (Sep 30) and RMB 13.64B in an April report from Robot Qianzhan, likely reflecting different FX assumptions. One outlet cited a July 2026 valuation of RMB 20B, but its sourcing is unclear (cited via Sohu, unconfirmed). No valuation has been disclosed for the September round.
(2) Round label and size are imprecise. The September round is described only as a “new round,” with a disclosed floor of RMB 1B and no stated size or round name. The ordering of the three co-leads also differs by outlet. QbitAI additionally names Digital Future and a CATL-affiliated vehicle among Pre-A investors, which the official list does not. We have used the official list as our baseline.
(3) Ownership structure undisclosed. Founder stakes, the employee pool and any rights attached to state-linked investors are not available. As is common for private Chinese issuers, disclosure is limited to press releases.
China’s embodied-AI market is in an overheated phase: sixteen companies joined the RMB 10B unicorn tier in 1H26 alone. Earlier movers such as Galbot, AgiBot, Galaxea (Xinghaitu), D-Robotics-adjacent peers and Zhipingfang already trade at RMB 15–20B or more reported valuations. We see Sudu’s relative differentiation at three levels: technical approach (simulation-first), founder assets (academic infrastructure) and capital structure (industrial and platform investors side by side).
Policies are trained on simulated data only, with no real demonstrations. If it holds, this structurally lowers real-robot data collection and tuning costs, shortens development cycles and eases portability across robot bodies and stations. Performance proof remains at the self-reported stage.
The designer of simulation and evaluation platforms widely used by embodied-AI labs worldwide leads the technical line. We see this as an asset in evaluation standards and talent access that later entrants cannot replicate quickly.
Owning both the software brain and the hardware shortens the learn-and-deploy loop. A dedicated hardware lead and an in-house robot system point to an integrated supplier rather than a model-only vendor. Body manufacturing scale and cost competitiveness remain unproven.
CATL, a Pre-A investor, is named as a collaboration partner in battery production and logistics. A strategic investor doubling as a validation partner can ease access to line data and early references, though the commercial commitment has not been disclosed.
Alibaba, Tencent, Ant and Xiaomi sit alongside state-linked investors such as Shanghai Guotou Xiandao and CDB Capital and a brokerage-affiliated investor. We believe this supports follow-on financing and a future listing path. Futeng says it has facilitated engagement with Hong Kong’s securities regulator.
Management frames generality as a means rather than an end, prioritizing high-reliability generalization on a limited set of valuable skills. Anchoring on uptime and stability in production, rather than demo breadth, is consistent with what industrial buyers require.
| Company | Reported Valuation | Positioning |
|---|---|---|
| Sudu (苏度) | $2B+ (Pre-A, Apr 2026) | Simulation-first, full stack, industrial manufacturing focus |
| TARS (它石智航) | ~RMB 13B (Pre-A, Apr 2026) | General embodied model plus industrial and general-purpose robot lines; Hillhouse and Sequoia led |
| Galbot (银河通用) | RMB 20B+ | National AI fund and SOE backing; record RMB 2.5B single round |
| Galaxea (星海图) | ~RMB 20B (Apr 2026) | VLA-model approach; single strategic platform investor (Ant) |
| Zhipingfang (智平方) | RMB 20B+ (Jun 2026) | Twelve rounds in a year; batch delivery on semi-automated lines |
Our overall assessment: Sudu is being priced on conviction in its technical route rather than on commercial metrics. With category leaders reportedly valued at RMB 20B or more, the gap to Sudu’s Pre-A valuation of RMB 13.4–13.6B suggests room for re-rating in follow-on rounds. At the same time, absent external validation of the simulation-first approach, evidence of stable operation on customer lines and clarity on body cost structure, we see material risk of valuation compression at this stage. We would monitor three items through 2H26: (1) whether independently verifiable performance data is released, (2) paid customers and line deployments beyond CATL, and (3) disclosure of the IP allocation tied to Prof. Su’s Fudan appointment.
Editorial note: the peer table mixes valuation dates and sources; figures are as reported and not normalized to a common date.