Westlake Robotics Deep Dive: Westlake University's First Robotics Spin-out Builds a Dual-Pretrain Brain-and-Cerebellum Architecture with WR1 Whole-Body Foundation Model and GAE Behaviour Expert
Westlake Robotics, the first Westlake University robotics spinout, pairs the WR1 general embodied foundation model (brain) with the GAE behavior-expert model (cerebellum) in a dual-pretrain architecture and the Westlake o1 body (29 joints, 165 N·m); five 2026 rounds totalling about RMB 500 million and booked orders near RMB 100 million.
Executive Summary
Westlake Robotics (Westlake Robotics Technology (Hangzhou) Co., Ltd.) is the first self-initiated commercialization venture of Westlake University in AI and robotics, founded in June 2021 in Hangzhou. Its three-in-one route spans a general brain, a humanoid whole-body cerebellum and self-built bodies: the brain side is WR1, a whole-body motion-operation integrated general embodied foundation model (a dual-pretrain architecture fusing a vision-language model with a world model, on a third-generation VLA-plus-RL route); the cerebellum side is GAE, a general behavior-expert model (programming-free motion generalization and whole-body control, supporting millisecond-level motion mirroring and one-operator-many-robots through its avatar system); the flagship body is the Westlake o1 (Titan o1) unveiled in 2026 at 134 cm, about 35 kg, with 29 self-developed integrated joints, 165 N·m knee peak torque and 110-plus TOPS of compute. Funding spans seven rounds from the angel round in November 2023; within 2026 it completed five rounds from February to September (Pre-A, Pre-A+, Pre-A++, an A round of nearly RMB 100 million led by Haiyuan Capital, and A1), totalling about RMB 500 million, at an estimated valuation of roughly RMB 1.8 billion (single source). Booked orders approach RMB 100 million across scientific teaching, humanoid data collection, power inspection, emergency response and home-textile and footwear scenarios.
Caveats: the company is private with no audited financials; the claims of the world's only dual-pretrain architecture and the world's first general motion model are company statements; the brain-versus-cerebellum assignment differs across reports (adjudicated in Discrepancies); orders are not revenue.
The verdict: the most complete scientist configuration and the most self-consistent architecture narrative in this cohort — Wang Donglin (nearly twenty years in reinforcement learning, a 2026 AAAI best-paper award) leads technology, Zhang Yue (ACL Fellow, eighteen years in NLP) covers the language side, and Du Haitao (a founding member of Tmall Genie) runs commercialization, mapping exactly onto brain, language and product. Layering the brain and cerebellum reflects the real frequency constraints of robot control better than the one-model-fits-all narrative common among peers.
Company Overview
The company was founded in June 2021 as the first robotics venture in Westlake University's commercialization system, with operations starting January 2024. The three founders divide cleanly. Founder Wang Donglin is a tenured professor at Westlake University, deputy head of the AI department and director of the Machine Intelligence Lab, chief scientist of a National Science and Technology Innovation 2030 major project; he earned his BS and MS at Xi'an Jiaotong University and a PhD in electrical and computer engineering from the University of Calgary in 2010, has published over 150 top-venue papers across nearly twenty years in robot and reinforcement learning, and won a 2026 AAAI best-paper award. Co-founder and chief scientist Zhang Yue is a tenured professor and vice dean of the School of Engineering, with a Tsinghua BS, Oxford MS and PhD, Cambridge postdoctoral work, eighteen years in NLP and election as an ACL Fellow in 2025. Co-founder and CEO Du Haitao was product and operations general manager of Alibaba's AI Lab and a founding member of Tmall Genie, having previously incubated two unicorn-scale AI-agent companies. The R&D team draws from Alibaba, ByteDance, Tencent and Huawei and from Oxford, Cambridge and Carnegie Mellon, covering mechanics, embedded hardware, embedded software, perception, motion control and AI, with over 300 team papers at top venues.
This configuration addresses the classic weakness of university spinouts — professor teams that know technology but not product. Investors repeatedly highlight Du Haitao's role in turning frontier technology into products users can actually touch and afford.
Funding is the other throughline: the angel round in November 2023; an angel-plus round of nearly RMB 100 million in January 2025; a RMB 100-million-scale Pre-A in February 2026; Pre-A+ in May 2026 (Xiaomiao Langcheng leading, with Xingzheng Capital and Beiyu Capital); a solo Pre-A++ in July 2026 (the Huirong Fund of Henan Investment Group); the A round in August 2026 (nearly RMB 100 million led by Haiyuan Capital); and A1 on September 11, 2026 (Haiyuan, Cuiwei Fund, Yinxinggu Capital, Westlake Sci-Tech Investment and Zhongshan-affiliated funds). The cumulative total is about RMB 500 million, with a valuation near RMB 1.8 billion at A1 (single source). Official use of proceeds names three targets: the WR1 model push, multi-scene commercialization and talent.
Technology
The Brain Side: WR1 and Dual Pretraining
WR1 is a whole-body motion-operation integrated general embodied foundation model that jointly pretrains a vision-language model with a world model, handling environment understanding, long-horizon task decomposition and autonomous decision-making on what the company calls a third-generation VLA-plus-RL route. Note that WR1 is a model, not a robot — early coverage of the whole-body general embodied foundation phrasing was widely misread as hardware, while the A-round use of proceeds lists conquering WR1, the foundation model, as its first item.
Dual pretraining means the brain (VLM plus world model) and the cerebellum (the GAE motion model) each undergo large-scale pretraining before joint fine-tuning. The company claims this architecture is the world's only one; the value proposition is balancing generalization against inference efficiency — when both layers are well pretrained, joint adaptation needs far fewer samples than training from scratch.
The Cerebellum Side: GAE, the General Behavior-Expert Model
The company defines GAE as the world's first general motion model, positioned as the robot's general cerebellum, with the analogy spelled out: ChatGPT for language, Sora for vision, GAE for motion. It generalizes to over a thousand actions without programming; the avatar-system demonstration has a motion-capture performer's movements mirrored by the robot at millisecond latency, with one operator driving multiple robots. The positioning is technically sound: general-purpose large models lack physical grounding and proprioception, while a behavior-specialist model trained on teleoperation, capture and simulation data is stronger on motion continuity and physical plausibility — at the cost of weaker cross-task and cross-body generalization. This is the same problem the action-optical-flow route of Zhongke Diwiji attacks, via a behavior-expert path rather than a unified action language.
The Bodies: Westlake o1 and the Product Matrix
The Westlake o1 (also branded Titan o1), released in early 2026, stands 134 cm, weighs about 35 kg, carries 29 self-developed integrated joints, an arm payload of about 3 kg, 165 N·m knee peak torque and 110-plus TOPS of onboard compute. The lightweight quadruped Xiaoxi and the industrial Aotian U1 (inspection, livestock, construction) complete the hardware matrix. A separate module line deserves note: the company sells motion-control, multimodal-perception and remote-teleoperation modules to third-party body makers, explicitly stating it does not compete with its customers and serves as an industry-common technology base — meaning Westlake runs both the own-robot and the brain-supplier routes, validating models on its own bodies while scaling the ecosystem through modules.
Technical Shortfalls
First, both exclusivity claims lack third-party benchmarks; GAE's ceiling for motion generalization (task complexity, success rate) has no public data. Second, the interface design of dual pretraining — how the brain's subgoals get executed by the cerebellum, how force feedback corrects the brain — is the make-or-break of such architectures, and no interface specification is disclosed. Third, the o1's 3 kg payload is low for industrial work; the humanoid initially serves validation and data collection. Fourth, training-data scale (hours, tasks, bodies) is undisclosed, and it is the decisive variable for a behavior-expert model.
WR1 Release and Home Scenario Hands-on Test (October 2026)
From October 21, 2026, media reported that Westlake Robotics had formally released WR1, referred to as WR1.0 in investor material, and completed a real-home hands-on test: removing trousers from a swaying clothes rail, cradling them on an arm and adapting to disturbance, with two robots perceiving and dividing labor to fetch, spread and fold laundry and the humanoid dexterous hand folding autonomously. An earlier arm-platform demo claimed a best of 36 seconds at 100 percent success. The architecture includes short and long-term motion-state recall, real-time action chunking and GAE look-ahead latency compensation. The release form and test results come from media and investor material and are not independently verified; the performance figures should be treated as media claims.
Product and Commercial Model
Orders and Scenarios
Booked orders approach RMB 100 million across five scenario classes: scientific teaching, humanoid data collection, power inspection, emergency response, and home textiles and footwear. The first two classes likely dominate — consistent with the validate-models-on-bodies positioning, and implying that true industrial and commercial volume has not started. Named partners include Alibaba, Intel, JD.com and the Hong Kong University of Science and Technology; a county-wide humanoid training base co-built with Longyou county merges order and data-engine design — the base is both a contract and a source of training data for GAE.
Business Model
Three revenue lines run in parallel: body sales (Westlake o1, Xiaoxi, Aotian U1), module licensing (motion control, perception, teleoperation, sold to body makers), and industry solutions. The modules-do-not-compete-with-customers stance is rare among domestic embodied companies — most peers bind the brain to their own bodies. It gives the company revenue elasticity independent of its own hardware shipments, but demands that module interfaces reach industry-base openness and stability.
Commercial Risks
Order nature (framework, firm contract, prepayment share) is undisclosed; research and data-collection orders carry weaker margins and durability than industrial ones; the 3 kg payload limits direct industrial penetration, so industrial volume depends on the module route or a next-generation body.
Competitive Landscape
Rivals in the embodied-brain segment fall into three classes: body makers with in-house brains (Zhiyuan, Unitree), independent brain companies (Galbot, Xingdong Jiyuan, Zhongke Diwiji and others), and academic spinout peers. Westlake's differentiation: first, the brain-cerebellum layering with dual pretraining is the most self-consistent architecture narrative, matching the real frequency stratification of robot control; second, the founding team pairs reinforcement learning (Wang), NLP (Zhang) and product (Du) in complete complement; third, the module-licensing line creates an ecosystem position most peers lack. The weaknesses: against Zhongke Diwiji it lacks a third-party benchmark signal of the ARNOLD-title kind; against Zhiyuan and Unitree it trails by an order of magnitude in capital and production; and the contest between GAE's behavior-expert route and unified-action-language routes has not been settled on public data.
Financing and Capital Structure
Seven capital events: the angel round in November 2023; angel-plus of nearly RMB 100 million in January 2025; a RMB 100-million-scale Pre-A in February 2026; Pre-A+ in May 2026 (Xiaomiao Langcheng leading, Xingzheng Capital and Beiyu Capital participating); a solo Pre-A++ in July 2026 (the Huirong Fund of Henan Investment Group); the A round in August 2026 (nearly RMB 100 million led by Haiyuan Capital); and A1 on September 11, 2026 (Haiyuan Capital, Cuiwei Fund, Yinxinggu Capital, Westlake Sci-Tech Investment, Zhongshan-affiliated funds, with Yiwei Capital as financial advisor). Cumulative funding is about RMB 500 million; the A1-time valuation is about RMB 1.8 billion (SoHoBlink, single source, unconfirmed by the company). Official use of proceeds: the WR1 model, multi-scene commercialization and talent.
DISCREPANCIES
1. The referent of WR1 is this report's most important correction. Early coverage widely misread the whole-body general embodied foundation phrasing as a robot body; cross-checked against the A-round use of proceeds (conquering the whole-body motion-operation integrated general embodied foundation model WR1) and multiple reports, WR1 is a foundation model on the brain side, while the body is the Westlake o1 (Titan o1). The body text follows this correction. Source grade: OFFICIAL (use of proceeds) versus MEDIA (misreading) conflict.
2. Brain-versus-cerebellum assignment has variants. Source A (NetEase report): GAE takes the cerebellum role (programming-free motion generalization and whole-body control) while the dual-pretrain architecture takes the brain. Source B (company releases): WR1 fuses the VLM and world model while GAE is described as an avatar system. The substance agrees (GAE is the cerebellum, WR1 the brain), but the avatar phrasing casts GAE as a teleoperation product rather than a model. Adjudication: read the architecture per source A and treat the avatar system as GAE's demonstration form. Source grade: MEDIA and OFFICIAL.
3. The four-rounds figure. Media uniformly say four rounds in half a year totalling about RMB 500 million; the full record runs seven tranches from November 2023 (angel, angel-plus, Pre-A, Pre-A+, Pre-A++, A, A1). The four rounds refer to the dense 2026 sequence. Source grade: MEDIA (incomplete count).
4. The RMB 1.8 billion valuation. A single SoHoBlink source at A1, no company disclosure. Source grade: DB (UNVERIFIED).
5. Two exclusivity claims. The world's only dual-pretrain architecture and the world's first general behavior-expert model are company claims without a defined comparison set or third-party benchmark. Source grade: OFFICIAL (self-reported).
6. Order composition. The near-RMB-100-million order book spans five scenario classes, but the share of scientific-teaching and data-collection work is undisclosed, as is order nature (framework, contract, prepayment); orders are not revenue. Source grade: OFFICIAL (self-reported).
7. Personal honors versus company capability. Wang's 2026 AAAI best-paper award and Zhang's 2025 ACL Fellowship are personal academic honors with no direct causal link to industrial product performance. Source grade: MEDIA (conflation risk).
8. Paper counts. Wang's 150-plus individual papers and the team's 300-plus cumulative papers are mixed in coverage; this report labels them separately. Source grade: MEDIA.
9. A third funding-round caliber. A Hangzhou Network report on October 21, 2026, states the company completed six rounds totalling several hundred million renminbi in the past nine months, inconsistent with both this report's seven tranches (about RMB 500 million) and the five-2026-rounds framing; the report disclosed no per-round amount or valuation. This report retains the seven-tranche caliber and flags the third as a media statement, not independently verified. Source grade: MEDIA (UNVERIFIED).
Risks and Diligence Gaps
First, the gap from architecture narrative to engineering proof. No public benchmark exists for dual pretraining or GAE; diligence should test motion-generalization success and cross-task transfer on held-out tasks against end-to-end single-model baselines.
Second, order-to-revenue conversion. Delivery cadence, acceptance terms and payment status of the near-RMB-100-million book are unknown; research and data-collection orders are less durable than industrial ones.
Third, body competitiveness. The o1's 3 kg payload and 35 kg weight sit at the low end for industrial humanoids; industrial volume requires the module route to earn first or a next-generation body.
Fourth, a crowded field with richer rivals. Zhiyuan and Unitree lead by a body length in capital, capacity and deliveries; among independent brain companies, Zhongke Diwiji carries a third-party benchmark title. Westlake must convert academic prestige into reproducible customer-side metrics.
Fifth, spinout governance. IP ownership, team equity and university revenue-sharing complexity will shape later rounds and the long-term capital path.
Sixth, production and delivery. The path from prototype to volume production requires lines, supply chain and quality control, none disclosed; running bodies and modules in parallel raises the delivery-organization bar.
Seventh, data compliance. The humanoid data-collection business involves operator imagery and behavioral data, and a county-wide training base involves public-space data — personal-information-protection and data-security regimes apply.
Eighth, the diligence checklist. Independent benchmark results, order contracts and margins, module customers and shipment volumes, o1 production-ramp data, and the valuation basis all need priority verification.
Composite Assessment
Technology: 7. The brain-cerebellum layering with dual pretraining is the right direction and a self-consistent narrative; GAE's motion-generalization analogy to ChatGPT and Sora is crisp; the modular line (motion control, perception, teleoperation) shows engineering depth. Deductions: no third-party benchmark, undisclosed interface design, unknown data scale.
Business: 6.5. Near-RMB-100-million orders are a scarce signal in a demo-rich, order-poor industry, and the Alibaba, Intel, JD and HKUST partner list plus the Longyou base give breadth. Deductions: order mix skewed to research and data collection, unclear order nature, industrial volume not started.
Finance: 7. Seven rounds totalling about RMB 500 million with dual state and market backing and repeated insider re-ups put capital adequacy in the upper-middle tier. Deductions: single-source valuation and a high burn rate across bodies, models and the training base.
Product: 7. The matrix is complete (humanoid, quadruped, industrial wheeled, plus three module classes); the o1's 29 self-developed integrated joints and 165 N·m knee torque show real actuator depth; the modules-do-not-compete stance shows ecosystem sense. Deductions: low payload and a short production record.
Market: 7. Embodied AI enjoys policy and capital tailwinds, the data-collection business rides the data-as-moat consensus, and module licensing meets a real brain shortage at body makers. Deductions: fierce competition and resource dispersion across three business lines.
Composite: 6.9. Westlake Robotics is this batch's best-matched team-to-thesis company: the three founders cover brain technology, language intelligence and commercialization exactly, the layered architecture is closer to the real constraints of robot control than one-model narratives, and the validate-models-on-bodies, scale-ecosystem-through-modules dual track is unique among peers. The physical-AI verdict: GAE's motion generalization must ultimately pass physical-constraint tests — the claim of programming-free generalization over a thousand actions is industrially meaningful only if it survives contact, friction and load variation; the depth of simulation training and physical verification is undisclosed, and that is the missing link between the motion-generalization narrative and industrial reliability.
References
- Tencent News, Ruicaijing (2026-09-11): the A1 round with Haiyuan Capital, Cuiwei Fund, Yinxinggu Capital, Westlake Sci-Tech Investment and Zhongshan-affiliated funds; proceeds to the WR1 model, multi-scene commercialization and talent; founding in June 2021 in Hangzhou; Wang Donglin (tenured professor, deputy head of AI, director of the Machine Intelligence Lab, chief scientist of a Science and Technology Innovation 2030 project); Du Haitao (ex-GM of Alibaba AI Lab, founding member of Tmall Genie)
- NetEase News (2026-09): the fourth round in half a year, about RMB 500 million cumulative; Pre-A February 2026, Pre-A+ May 2026 (Xiaomiao Langcheng leading, Xingzheng and Beiyu participating), A round August 2026 (nearly RMB 100 million led by Haiyuan); co-founder Zhang Yue (tenured professor, vice dean of engineering, NLP); GAE as the cerebellum with the dual-pretrain architecture as the brain; Westlake o1 with 29 self-developed integrated joints, about 35 kg, about 3 kg arm payload; partners Alibaba, Intel, JD, HKUST; the Longyou county-wide humanoid training base; near-RMB-100-million orders across scientific teaching, humanoid data collection, power inspection, emergency response, home textiles and footwear
- SoHoBlink (2026-09): the A1-time valuation of RMB 1.8 billion; WR1 as the whole-body unified general foundation model fusing VLA and world model; GAE as the avatar system — the world's first general motion model, millisecond mirroring of motion-capture performers and one-operator-many-robots; Westlake o1 (Titan o1) at 134 cm, about 35 kg, 165 N·m knee peak torque, 110-plus TOPS; over 300 team papers
- Leaderobot (2026): operations starting January 2024; Wang's 2026 AAAI best-paper award, Xi'an Jiaotong BS and MS, 2010 Calgary PhD, 150-plus papers; Zhang's 2025 ACL Fellowship, Tsinghua BS, Oxford MS and PhD, Cambridge postdoc, eighteen years in NLP; the solo Huirong Fund Pre-A++ in July 2026; investors including Saimeng (SAIF), Xiaomiao Langcheng and Haiyuan; the brain-plus-cerebellum-plus-body three-in-one route; third-generation VLA-plus-RL
- Chushen (2026-05): the angel round from November 2023, angel-plus of nearly RMB 100 million in January 2025, the RMB 100-million-scale Pre-A in February 2026, Pre-A+ in May 2026; Xiaomiao Langcheng's repeated re-ups; one-brain-many-bodies addressing stiff motion, programming cost, weak generalization and teleoperation latency; GAE generalizing to over a thousand actions without programming; the matrix of quadruped Xiaoxi, industrial Aotian U1 (inspection, livestock, construction) and humanoid Titan o1 (precision industrial operation and remote control); core modules (motion control, multimodal perception, remote teleoperation) sold to body makers without competing with them
- Technical background (public): the frequency stratification of cerebrum and cerebellum — hundreds-of-hertz real-time control versus seconds-scale reasoning; pretraining both layers before joint fine-tuning raises sample efficiency, with the interface (joint commands and torques versus skill selection and subgoals) as the make-or-break
- Technical background (public): behavior-expert models trained on teleoperation, capture and simulation data gain physical grounding and continuity over general large models; cross-task and cross-body generalization remains the shared weakness, forming a path contrast with unified-action-language routes such as action optical flow
- Technical background (public): the degrees-of-freedom, payload and cost triangle of humanoid bodies; 29 self-developed integrated joints and 165 N·m knee torque mark a self-developed-actuator route, with the 3 kg payload steering early use toward validation and data collection
- Commercial background (public): the industry's demo-rich, order-poor state; orders must be classified into framework agreements, firm contracts and prepayments; research and data-collection orders carry weaker margins and durability than industrial ones
- Industry background (public): data collection has become a standalone business line for embodied companies; the county-wide training-base model merges contract and data-engine design
- Competitive background (public): rivals in the embodied-brain segment include body makers with in-house brains (Zhiyuan, Unitree) and independent brain companies (Galbot, Xingdong Jiyuan, Zhongke Diwiji et al.); Zhongke Diwiji holds the CVPR 2026 ARNOLD challenge title as a third-party signal
- Policy and governance background (public): humanoids in national industrial planning, embodied AI in the government work report, and the RMB 1.2 trillion sci-tech refinancing facility covering high-end equipment; university commercialization involving IP ownership, team equity and school revenue-sharing
- Hangzhou Network and an investor official account, October 21, 2026: Westlake Robotics formally released WR1 (WR1.0) and completed a real-home hands-on test (laundry fetching, folding, two-robot cooperation), with an arm-platform demo claimed at a best of 36 seconds and 100 percent success; the same-day Hangzhou Network report said nearly nine months brought six funding rounds of several hundred million renminbi (a third funding caliber, with no per-round amount or valuation disclosed). Both are media or investor material, not independently verified.
This report is compiled from public sources and does not constitute investment advice. The company is private; revenue, margins, order nature and product test data are undisclosed. Funding and order figures are company statements, and the valuation rests on a single source; uncertainties are flagged throughout.
Related papers
Related tools
Related cases
Related reports66
- 51World Deep Dive: A Listed Digital Twin Company Where 51Sim Becomes the First Growth Engine While 51Earth Nearly Vanishes, and Two Irreconcilable Financing Naming Conventions Must Both Be Kept
- AMI Labs: A USD 1.03B Seed Bet on World Models, LeCun's Non-Generative Path and Its Blank Scorecard
- Anshi Pacific (PERA) Deep Dive: A Beijing Domestic-CAE Leader Founded in 2003 Whose PERA SIM General Simulation, AIROM Reduced-Order Models, AgentX Multi-Agent and Jingzhi iGPT/Whale Span from Industrial Simulation to Physics AI
- AspenTech Deep Dive: The Process-Simulation Incumbent Acquired by Emerson and Its Hybrid Industrial-AI Route
- Autodesk Deep Dive: Neural CAD Foundation Models and a $200M World Labs Bet on Physical-World AI, with NAVASTO Adding CFD Surrogate Capability
- Bentley Systems Deep Dive: iTwin Digital Twin and Generative AI Empower Full-Lifecycle Infrastructure Engineering
- BeyondMath Deep Dive: A Foundation Physics AI on the Neural-Operator Route, a $19M Consortium Budget, and Three Misread Facts
- Cadence Deep Dive: System-Level Multiphysics EDA and the Millennium M2000 Supercomputer Anchor Physical AI, with AuraStack Agentic AI and Fidelity GPU CFD as Key Levers
- COMSOL Deep Dive: Four Decades of Multiphysics Solvers, Entering Physical AI via DNN Surrogate-Model Training and All-Physics GPU Acceleration
- Dassault Systèmes Deep Dive: 3DEXPERIENCE and SIMULIA Virtual Twin Physics Behavior Anchor Industrial AI, with LEO Virtual Companion and GPU Explicit Dynamics as Physical-AI Levers
- DP Technology: From the Gordon Bell Prize to National-Level Unicorn Status and the Door of a Hong Kong Listing, the AI-for-Science Leader's Technical Delivery and Commercial Blank
- Emmi AI: From Linz Spin-out to a EUR 330M Mistral Acquisition in 17 Months
- Fainite Deep Dive: A Zurich Physics-AI Simulation Startup Whose Caltech, ETH Zurich, Cambridge and Google Team Uses Physics-Aware Deep Learning to Cut Simulation from Tens of Hours to Seconds, With Only Venture Kick Grants and No Priced Round
- Fysics AI: The PhysX Founder's Second Act, a Domestic Differentiable Physics Engine and Its Ecosystem Play
- Frontop (301313) Deep Dive: From Digital Exhibition Halls to a Physical-AI Data Foundation - FTE Engine, FunSim Simulation and the Gap Between Positioning and Revenue
- Godela Deep Dive: A YC X25 Physics-AI Simulation Startup Whose Ex-Apple, Google and Intel Team Uses Physics-Constrained Surrogate Models to Cut Simulation from Days to Seconds, at Seed Stage with Pilot-Only Access
- HollyWave (688682) Deep Dive: RDSim AI Electromagnetic Simulation and Engineering Agents Close the Simulation-Measurement-Product Loop
- IAT (Alt) Deep Dive: China's First Listed Independent Automotive Design Company Pivots to AI-native R&D with the Yuanfang, Yufeng and Taiyi 3.0 Product System and an NVIDIA Omniverse Partnership, Reporting RMB 975M Revenue and a Widened RMB 250M Net Loss in 2025
- Inductiva Deep Dive: A Closed Case in Open-Solver Orchestration, Eight Premises That Must Be Corrected, and Where Co-Founder and CTO Hugo Penedones Went Next
- Intesim (INTESIM) Deep Dive: A Home-grown Chinese CAE Leader Founded in 2009 with Self-developed Solvers for Structure, Fluid, Electromagnetics, Thermal and Acoustics and Multiphysics Coupling, a ~RMB 100M B Round from CNII and Shenzhen Capital, and 2025 Revenue and Bookings Both above RMB 100M
- iSoftStone (301236) Deep Dive: A ChiNext-Listed IT Services Giant Whose TianShu Platform Builds a Three-Layer Physical-AI Stack with Self-Developed TwinHub Digital Twin, FSim Industrial Simulation, CAE Multiphysics and Phyxis Embodied Control
- Jua Deep Dive: The Earth Physics Transformer as a Large Physics Model for Energy Trading
- Keysight (with ESI Group) Deep Dive: Virtual Prototyping and Hybrid AI Bridge Measurement and Simulation, with Natural-Language Assistants in EDA
- Lightwheel Deep Dive: A Physical AI Data and Evaluation Infrastructure Startup Founded by a Former NVIDIA Simulation Lead, Seven Rounds Totalling Roughly CNY 2.65 Billion at a Valuation Above USD 2 Billion, Seats on Both the Newton and EgoVerse International Standards Committees, With Two Incompatible Accounts of Q1 Orders
- Luminary Cloud: The Physics AI Model Factory Backed by NVentures
- MathWorks Deep Dive: MATLAB/Simulink and Simscape Multiphysics Power Physical AI, with R2026a Copilots and an Agentic Toolkit for Edge Intelligence
- Maniformer (Mifeng) Deep Dive: An Agibot-Spun Physical-AI Data Infrastructure Firm Whose Bodyless MEgo Capture Hardware Is in Mass Production and Which Closed Three Rounds in Half a Year Led by China Telecom
- Mistral AI: How Physics AI Moved From Language Models to EUV Lithography and Nuclear Power
- Monolith AI: The British Engineering-AI Firm That Cut Physical Test Counts Instead of Accelerating Simulation, Now Inside CoreWeave
- Monumo Deep Dive: A Cambridge Deeptech Company Redesigning Electric Motors with the Anser AI Engine, Tens of Millions of Multiphysics Simulations per Day, GBP 10.5M Raised in Total, and a Series A Being Raised
- NAVASTO: Moving Solver Insight Onto the Designer's Desk, and How Autodesk Bought That Entry Point
- Navier: Handing CFD Workflows to Agents, a Seed-Stage Bet on AI for Hardware Engineering
- Neural Concept: An EPFL Spin-out Betting on Neural Surrogates, with $138M Raised
- NP Co. (NP Company) Deep Dive: A 2025-founded Paris Inria-born AI-native simulation startup that raised a EUR 6M pre-seed led by Partech with Mistral AI co-founders as angels, claiming a pre-trained physics foundation model at 1,000x speedup — most claims unverified
- NVIDIA Deep Dive: Building the Full-Stack Physical-AI Infrastructure with PhysicsNeMo, Warp, Cosmos World Foundation Models and Omniverse
- Physical Intelligence Deep Dive: A $2.1B Bet on the GPT Moment for Robotics, With pi0.7 Self-Reporting 60-80 Percent Success on Unseen Tasks and Neither Pricing nor Revenue Disclosed for the Pi API
- Physical Superintelligence Deep Dive: A $58M Seed Bet on Virtual Physicists, With Emmy Targeting AI Data Center Multiphysics Optimization as Its First Commercial Proof Point
- PhysicsX: Pretrained Geometry and Physics for Industrial Engineering at a $2.4B Mark
- Prometheus: A $41B Mark, $18.2B Raised, and Not Yet a Product
- PTC Deep Dive: Creo/Onshape and the ThingWorx Digital Thread Power Physical AI, with NVIDIA Isaac Sim Bridging Robotics Design to Simulation
- Qianjue Technology Deep Dive: How a Body-Agnostic Embodied Brain Pairing the Predictive World Model Polibrain with Brain-Inspired Partitioning and Polynomial Representation Reached a Hundred Thousand Cleaning and Service Robots
- Quanscient Deep Dive: Cloud-Native and Quantum-Powered Multiphysics Simulation as the Data Engine for AI-Native Hardware Engineering
- Rescale Deep Dive: An HPC Operating System and Desktop Engineering Workbench in the Cloud, a $115M Series D, and Five Misread Facts
- Schrödinger Deep Dive: Physics-Based FEP+ Simulation and Machine Learning Power Molecular Discovery, with the 2026 Bunsen Agentic Co-Scientist
- Secondmind Deep Dive: A Cambridge Engineering-AI Firm Whose Bayesian Active Learning Cuts CAE Simulation and Test-Bench Runs by up to 90%, with Roughly USD 73M Raised and Mazda Both Customer and Investor
- Shichuang Technology Deep Dive: SupreCAST Cloud-Native Material-Processing Simulation and Industrial Agents Advance the CAE+AI Strategy
- ShonCloud (Shunyun) Deep Dive: Meshless Particle-Method CFD for EV and Nuclear Fluid Multiphysics
- Shuzhi Maijin Deep Dive: A Shenzhen Fully Self-Developed SNN-plus-LBM CFD Startup Claiming Second-Level Results on Hundreds-of-Millions of Grids on Consumer Hardware, With a Tens-of-Millions-RMB Angel Round From a Ministry-of-Finance-Affiliated Fund and an FPGA Migration Plan
- Siemens (incl. Altair) Deep Dive: The 2025 Altair Acquisition Unites Simcenter and Altair, with PhysicsAI Geometric Deep Learning and PhysicsAI Generate as the Physical-AI Core
- Silurian Deep Dive: A Generative Forecasting Transformer Earth-System Foundation Model for Weather Risk on Energy and Infrastructure Assets
- SimScale Deep Dive: A Cloud CAE Platform Built on Core-Hour Subscriptions, EUR 52M of Equity Kept Separate from EUR 25M of Debt, and Five Misread Facts
- Skild AI Deep Dive: A General Robot Foundation Model Incubated at CMU's Robotics Institute, With a USD 1.4B Series C at a USD 14B Valuation, a Conflicting Series B Size, and an ARR Figure That Is a Run-Rate Rather Than Revenue
- Songying Deep Dive: A Huawei-Cloud-Alumni Startup Building a Domestic Physical AI Operating System, With ORCA OS Positioning Against NVIDIA Omniverse, Three Incompatible Financing Narratives, and a Verifiable Third-Party Anchor in a CNY 20 Million Minority Investment
- Sudo Tech Deep Dive: Zero-Real-Data Simulation Training, the Fully In-House Sudo R1 Stack, and an Atomic-Skill Commercialisation Strategy from a Startup Valued Above CNY 20 Billion in 17 Months
- Suochen Technology: The Pain of Becoming a Physics-AI Platform from a CAE Vendor
- Synopsys (Ansys) Deep Dive: The 2025 Ansys Acquisition Turns the EDA Leader into a Silicon-to-Systems Engineering Powerhouse, with Ansys SimAI and the NVIDIA Physical-AI Partnership as the Core Physics-AI Levers
- TenFong Deep Dive: An Academician-Led Chinese Industrial Simulation Software Vendor Whose Official Timeline and Database Records Use Incompatible Round Names, With an Official Financing PDF and an Investment Database Both on File
- Tianfu (Nanjing Tianfu Software) Deep Dive: A 2011-founded Chinese Intelligent-CAE Veteran That Closed a Several-Hundred-Million-RMB Series C, Pivoted to an Industrial World Model in 2024, With Unreconcilable Financing and Valuation Records and No Disclosed Revenue
- Tongyuan Soft Control Deep Dive: MWORKS (Syslab/Sysplorer) as a Domestic MATLAB/Simulink Alternative, with MWORKS AI and the MOGICK Engineering Agent Advancing Home-Grown Physical AI
- Vinci AI: Turning Physics Simulation from a Checkpoint into a Continuous Constraint
- World Labs: Fei-Fei Li's Spatial-Intelligence Unicorn, Acquired by AMD for $8.2B
- XVERSE Deep Dive: A Simulation-World Route to Physical AI Built on the XVERSE Model Family, an Edge-Cloud 3D Engine and the Physics-Based MotionGen Model
- SimWE: From In-House Solvers to a Physical-AI Engine, the Domestic CAE Leader and What Comes Next
- Yunjing Zhifang Deep Dive: A Peking-University-Incubated Cloud-Native GPU CAE Company with the Simversus Platform and the YJ-DORGP Small-Sample Intelligent Solver, Reporting 275 Percent Order Growth, 160 Percent Revenue Growth and Breakeven in 2025
- Zhongke Diwiji (Five Ages) Deep Dive: A CAS-Institute-of-Automation and Tsinghua Spin-out Whose FAM Embodied Foundation Model Uses Action Optical Flow for Cross-Embodiment Generalisation, Reaching 97% Success from 3 to 5 Demos, with EraBrain and Jiling Robots Serving Sinopec, Leapmotor and Midea
- ZWSOFT (688083) Deep Dive: Overdrive Kernel and ZW3D Wukong 2027 Underpin an Integrated CAD/CAE/CAM Stack, with a CAx Agent Platform as the Practical Route to Engineering AI
This report is compiled from public sources. It is for reference only and does not constitute investment advice. Data is current as of the publication date; please refer to the company’s official filings.