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
Maniformer (Mifeng) was founded in February 2026 as an Agibot-spun physical-AI data-service platform; its bodyless MEgo capture hardware is in mass production, it launched the Mifeng Pai crowdsourcing platform in October 2026, closed seed, angel and angel-plus rounds within half a year with the latest led by China Telecom, and made a strategic investment in Pimax.
Executive Summary
Maniformer, formally Shanghai Maniformer Embodied Intelligence Technology Co., Ltd., was founded in February 2026 in Shanghai with Yao Maoqing as legal representative, registered capital of 5.29 million renminbi and a 70.92 percent stake held by Agibot Innovation Shanghai Technology. Yao is simultaneously Agibot partner, senior vice president and head of the embodied business unit, and Maniformer chairman and CEO. That ownership and staffing structure means Maniformer is a data subsidiary incubated inside the Agibot system rather than an independent startup.
The company positions as a one-stop physical AI data service platform building platform-based supply infrastructure for embodied intelligence data: collection through MEgo body-free capture devices, data governance, dataset production, closed-loop evaluation and a data circulation network called the Hive Data Co-Creation Initiative.
On August 17, 2026 the company announced a new round in the hundreds of millions of renminbi led by China Telecom with Zhangjiang Group following and existing investors Sequoia China and Yuanqi Innovation adding above pro rata. Proceeds build platform-based data supply infrastructure, expand MEgo production, strengthen the collection, governance and evaluation stack and accelerate tens of millions of hours of physical interaction data capacity. This was the third round since founding in early 2026.
Commercial progress has two checkable milestones. On August 31, 2026 the 20,000th MEgo body-free data capture device rolled off the line and was delivered to JD.com for real business scenarios, and the company announced a body-free data partnership with Tencent Robotics X to support Tencent's embodied model development. Twenty thousand units indicates hardware is in mass production and customers have extended from labs to internet majors and industrial players.
October 2026 added two new lines. On October 12, at the 26th China International Industry Expo, the company showed the MEgo series of body-free data capture terminals, the MEgo Engine data governance engine and Mifeng Pai, a physical-AI data crowdsourcing service platform that the company calls the first full-category high-quality physical-AI data crowdsourcing platform in the world; users take tasks in the app, record real grasping, carrying and tidying operations in the field and submit the data, and the platform settles pay by the effective data duration approved after review. In the same period the MEgo series was selected as a Forbes China Yangtze River Delta New Intelligence Award case. On October 19 the company made a strategic investment in the VR headset maker Pimax, with the amount undisclosed, while China Renaissance acted as Pimax's exclusive financial adviser and long-term strategic adviser; Pimax had already joined the Hive Data Co-Creation Initiative led by the company as a hardware partner.
Yao made two judgments at WAIC 2026 worth recording. First, scaling physical intelligence requires breaking three walls: a data wall since real interaction data is extremely scarce and expensive, a representation wall since unified physical representations across tasks, scenes and embodiments have not formed, and a closed-loop wall since real trial and error is costly and slow. Second, world models are essentially AI systems understanding how the physical world operates, whose most important function is predicting the next state, while most current work remains at visual generation with two problems: data type mismatch between internet video distributions and the physical world data robots need, and insufficient data scale.
In physical AI terms Maniformer occupies a distinctive position: not models, not robots, but data. Its value depends on one judgment, that the embodied bottleneck is data rather than algorithms or compute. If true, data suppliers sell picks and shovels; if not, their ceiling is compressed.
One framing point matters. This is a company with excellent parentage, fast pace and hard data assets but opaque financials and customer structure. Its strengths are Agibot scenarios and technology, China Telecom cloud and government channels, Zhangjiang campus scenarios, Sequoia backing and 20,000 delivered devices. Its weaknesses are undisclosed revenue and margin, unknown customer concentration and the feasibility of the tens of millions of hours target.
Company Overview
Parentage and Ownership
Founded in February 2026 with 5.29 million renminbi of registered capital and 70.92 percent held by Agibot, the structure of parent majority control plus separate legal entity plus independent fundraising is increasingly common in embodied intelligence: the parent provides technology, scenarios and orders while the subsidiary raises independently, holds its own valuation and faces customers including the parent's competitors.
The commercial implications cut both ways. The subsidiary can serve Agibot's competitors, since a customer that is another robot company may not buy data directly from Agibot but may buy from a nominally neutral Maniformer, while the credibility of that neutrality is questionable given 70.92 percent ownership.
Team
Yao's dual role means he can mobilize Agibot resources including robots, scenarios and algorithm teams, while his time is divided. Public information on other team members is limited, and the business implies three groups: hardware engineers for MEgo, data engineers for collection, governance and labeling, and commercial staff for industrial customers.
Funding Pace
Founded in early 2026 and already through multiple rounds including the hundreds of millions in August, the pace is typical for embodied intelligence where capital enthusiasm is high and data infrastructure is seen as the most certain beneficiary, since whichever robot company wins needs data.
The latest round's investor structure is notable: China Telecom leading with cloud, government customers and data compliance qualifications, Zhangjiang Group following with campus scenarios, industrial policy and space, and Sequoia China and Yuanqi Innovation adding above pro rata for market endorsement.
Technology
What Body-Free Collection Means
MEgo's core concept is collecting data without a robot body, using independent devices to record human manipulation or environment interaction and mapping that to robot action spaces.
The approach solves a concrete problem: real robot collection is expensive, requiring a robot costing tens to hundreds of thousands of renminbi, teleoperation equipment, operators and space, with low efficiency since robots move slowly and fail often. Body-free collection uses cheaper devices such as gloves, motion capture and first-person cameras to let humans operate directly, faster and cheaper.
The technical difficulty is the domain gap. Human hands and robot grippers differ in morphology, degrees of freedom and mechanics, so human data cannot train robots directly and requires retargeting into joint space, which introduces error and differs per robot.
Body-free collection is therefore not free: it trades low collection cost for mapping error and generality challenges. Whether the trade is worthwhile depends on retargeting maturity.
The Three Walls
Yao's framework of data, representation and closed-loop walls is a clear account of embodied bottlenecks.
The data wall is scarcity and cost of real interaction data, an industry consensus since language models scrape trillions of tokens while robot data must be collected with people, devices, space and time.
The representation wall is the absence of unified physical representations across tasks, scenes and embodiments, meaning the same action such as grasping a cup requires different joint commands on different robots and models struggle to learn embodiment-agnostic representations, a fundamental reason embodied intelligence is harder than language.
The closed-loop wall is costly and slow real-world trial and error, where hardware wear, safety risk and time far exceed the cost of text generation and human labeling, preventing large-scale trial-based training.
The framework's value is showing that data is only one wall and not the hardest, since representation and closed-loop problems cannot be solved by more data. That means Maniformer's business solves one third of the problem, which is both an opportunity and a risk.
Judgment on World Models
Yao argues world models are essentially systems understanding physical laws whose core function is predicting the next state, that most current work remains at visual generation, and that two problems are data type mismatch and insufficient scale.
The judgment aligns with LeCun's position that world models predict state transitions rather than pixels, and contrasts with World Labs' generation-first route. As a data supplier, the stance is clear: if world models need physical data rather than internet video, the value of a data supplier holds.
Technical Shortfalls
First, retargeting accuracy is undisclosed, and it determines data usability.
Second, there is no shared standard for embodied data quality across diversity, labeling precision, physical realism and task coverage, the same gap as the absence of accepted physical AI benchmarks.
Third, MEgo technical parameters such as capture precision, degrees of freedom, battery life, weight and price are unpublished.
Fourth, the definition of tens of millions of hours is unclear, whether capture duration or effective interaction duration.
Product and Commercial Model
Product Matrix
Five parts exist. MEgo series hardware, already at 20,000 units. A data platform covering collection, governance, labeling and dataset production. Closed-loop evaluation cycling data into training, testing in simulation or real robots and feeding improvements back. The Hive Data Co-Creation Initiative, a circulation network resembling a data alliance. And Mifeng Pai, the physical-AI data crowdsourcing service platform first shown at the October 12, 2026 industrial expo.
Mifeng Pai addresses the scale of the collection side. For industrial settings such as machine shops, electronics workshops, printing shops and garment plants, users take tasks in the Mifeng Pai app, record people performing grasping, carrying and tidying operations in the real environment and submit the data; after quality verification and processing the data is used for embodied model training and evaluation, and the platform settles pay by the effective data duration approved after review. The company calls it the first full-category high-quality physical-AI data crowdsourcing platform in the world, a first that depends on how full-category and high-quality are defined and that this report keeps as company wording.
The relation to the other three is division of labor: proprietary devices and enterprise purchases cover major accounts and high-standard scenarios, while the crowdsourcing platform covers long-tail and scattered scenarios, and the two carry different data quality criteria and compliance requirements.
Customer Structure
Three customers are public: JD.com receiving the 20,000th MEgo into real business scenarios, Tencent Robotics X in a body-free data partnership, and Agibot itself as parent and largest data consumer.
Their natures differ, with JD.com an industrial player in logistics where repetitive operations suit embodied data, Tencent a research organization, and Agibot both major customer and related party. Concentration may be high but is undisclosed.
The collection network is the more checkable scale figure here: the company reports more than 20 city innovation centers in China and more than 5 overseas nodes, with strategic cooperation with leading domestic and overseas internet firms, embodied intelligence companies and AI research institutions, supplying training data for foundation models such as Ant Group's LingBot-VLA 2.0. These nodes are the operating base for crowdsourced collection and scenario task organization; their definition, whether self-built centers, regional partners or site alliances, and their per-unit output are undisclosed, so the figure indicates coverage rather than capacity.
Business Model
Three models are possible: selling devices at hardware margins of 20 to 40 percent, selling data through dataset licensing or hourly billing at potentially higher but hard-to-standardize margins, and selling services through custom collection, governance and evaluation at project margins.
China Telecom adds a fourth: a bundled cloud-network plus embodied data solution for humanoid robot companies, smart manufacturing, research institutes, smart parks and smart home customers, whose value lies in channel access to enterprise customers.
Capacity Target
Tens of millions of hours is repeatedly cited. Feasibility depends on three variables: device capacity, where 20,000 units at two hours per day yields about 14.6 million hours annually; collectors, requiring many operators or crowdsourcing; and governance, since cleaning, labeling and structuring are more laborious than collection.
At 20,000 units and 1.5 effective hours daily, annual capacity is about 10.9 million hours, matching the target, so the goal is reachable at the device level with real bottlenecks in staffing and governance.
Competitive Landscape
Competitors in Embodied Data
Three groups exist: robot companies building in-house teams such as Agibot, Unitree and Robot Era; specialized data providers such as Maniformer and Guanglun Intelligent plus annotation firms moving in; and simulation vendors providing synthetic data including the NVIDIA Isaac Sim ecosystem and domestic simulation companies.
Maniformer sits in the second group with a peculiarity: Agibot scenarios and technology backing combined with a separate legal entity attempting to serve third parties, an insider-outsider position that is both advantage and disadvantage.
Comparison With Guanglun Intelligent
Guanglun Intelligent is another major domestic player emphasizing simulation, synthetic data and evaluation, contrasting with Maniformer's body-free real collection. The industry debate continues between real data, higher quality but costlier and limited in scale, and synthetic data, cheaper and scalable but with a sim-to-real gap. The rational answer combines both, while early companies usually pick one.
Relationship With Cloud Vendors
China Telecom's investment shows carriers value data plus cloud-network. Cloud vendors such as Alibaba, Huawei and Tencent also offer storage, annotation and training platforms for embodied data, so binding to one carrier brings government channels and compliance credentials while risking exclusion from other carriers' customers.
Financing and Capital Structure
History
Founded in February 2026 with multiple rounds in the same year and hundreds of millions in August, cumulative funding is likely in the several hundred million to one billion renminbi range, though totals are undisclosed.
Ownership
Agibot holds 70.92 percent with external investors holding about 29 percent led by China Telecom, Zhangjiang Group, Sequoia China and Yuanqi Innovation. Agibot has absolute control while external investors combine financial and strategic motives with limited voice.
Valuation and Revenue
Valuation is undisclosed. Revenue is undisclosed, though with 20,000 devices delivered and three benchmark customers it is likely in the tens of millions to hundreds of millions of renminbi range, with margin structure across hardware, data and services unknown.
Outbound Investment
Alongside taking outside investment the company also makes strategic investments. On October 19, 2026 Maniformer invested in Pimax, the Hangzhou maker of immersive headsets, and the two parties established a strategic cooperation, with the amount undisclosed; China Renaissance served as Pimax's exclusive financial adviser and long-term strategic adviser.
The industrial logic is traceable. Pimax was founded in 2015 and makes VR headsets, and its decade of accumulated capability is time-space alignment across multiple sensors and sampling rates, used to record human intent and full-body motion in real scenes at high fidelity; Pimax had already joined the Hive Data Co-Creation Initiative launched by Maniformer together with the Shanghai research institute, the National Data Standards Committee and the CCID Research Institute, acting as a hardware partner, so this transaction upgrades an existing ecosystem relationship into a capital tie. Maniformer chairman and CEO Yao Maoqing framed the gap as real scenes being hard to cover, collection being expensive and quality being hard to unify, all requiring firmer underlying technology.
For Pimax this extends spatial computing into AI data. For Maniformer it brings upstream capture capability into its own data quality system. Because the amount is undisclosed, the scale of the transaction cannot be assessed and its effect on the financial statements cannot be judged.
DISCREPANCIES
1. Founding date. Registration was February 2026 while some coverage says early 2026. This report uses February 2026. Source tier: OFFICIAL (registry).
2. Number of funding rounds. The company says multiple rounds since early 2026 while media describe seed, angel and angel plus. Rounds, amounts and valuations are incompletely disclosed. This report uses multiple rounds with the latest in the hundreds of millions. Source tier: MEDIA crossed with OFFICIAL.
3. The latest round amount. The company says hundreds of millions of renminbi without a precise figure while some media write nearly 100 million. This report uses hundreds of millions. Source tier: OFFICIAL (company statement).
4. Timing of the Agibot stake. Registry data shows 70.92 percent at a point in time, subject to dilution in later rounds. This report labels it as of the public information date. Source tier: third-party registry (MEDIA).
5. Definition of tens of millions of hours. The company targets tens of millions of hours without defining whether hours are capture duration, effective interaction or labeled data. This report estimates on capture duration. Source tier: OFFICIAL (target) plus inference.
6. Nature of customer relationships. JD.com receives devices, Tencent partners on data business and Agibot is parent and related party, yet some coverage calls all three customers. This report labels each. Source tier: OFFICIAL (partnership facts) plus inference.
7. Yao's dual role. He is both Agibot senior vice president and Maniformer chairman and CEO, while some coverage mentions only one title. This report lists both. Source tier: OFFICIAL.
8. Boundaries of the one-stop positioning. The company calls itself a one-stop physical AI data service platform while actual capability covers collection, governance and evaluation, with model training, synthetic generation and data trading unclear. This report limits the scope to collection, governance and evaluation per public information. Source tier: OFFICIAL (positioning) plus inference.
9. Settlement basis for crowdsourced data on Mifeng Pai. The platform settles pay by the effective data duration approved after review, so the pricing basis is verified effective duration, and submitted volume or device online time are different bases. The criteria for what counts as effective, meaning verification pass rate, exclusion rules and per-task duration caps, are unpublished, and neither the unit price nor collector earnings are disclosed. This report uses the wording that settlement follows approved effective data duration and does not derive collector income or platform margin from it. Source tier: OFFICIAL (company and media wording) with unit price and criteria absent.
10. Statistical basis of the collection network scale. The company reports more than 20 city innovation centers in China and more than 5 overseas nodes. The figure appears in financing coverage from August 2026 without a node-by-node list, start dates or output data, and whether a city innovation center means a self-built operating site, a regional partner or a site alliance is unclear. This report treats it as a public statement of coverage and does not convert it into collection hours or capacity. Source tier: OFFICIAL (company statement in financing coverage) with node details absent.
11. Outbound investment and neutrality. The controlling shareholder is Agibot at 70.92 percent, and in October 2026 the company made a strategic investment in Pimax with the amount undisclosed, with China Renaissance acting as Pimax's exclusive financial adviser and long-term strategic adviser. An outbound investment and an inbound one are different things, and the former does not change the ownership structure, but it puts the company in three roles at once in the data ecosystem, namely platform, capital provider and related party, alongside its positioning as a neutral data platform for third-party customers. Source tier: MEDIA crossed with OFFICIAL.
Risks and Diligence Gaps
First, customer concentration and related-party transactions, with Agibot both parent and major customer and JD.com and Tencent benchmarks of perhaps small share, so revenue dependent on the related party would raise independence and valuation questions.
Second, credibility of neutrality, since 70.92 percent Agibot ownership makes external robot customers question handing data needs to a competitor's subsidiary.
Third, unvalidated data quality, with no public metrics on retargeting accuracy, task coverage or labeling quality, while data quality is the core asset of a data supplier.
Fourth, mismatch between capacity and governance, since 20,000 devices can yield tens of millions of raw hours but governance capability is undisclosed and labor intensive.
Fifth, margin across models, where devices bring hardware margins, data brings content margins and services bring project margins, and a mix may yield modest overall margin.
Sixth, policy and compliance covering privacy in human operation video, cross-border data for overseas customers and data trading under the Hive initiative, where China Telecom credentials help without resolving everything.
Seventh, competition from in-house robot data teams, synthetic data vendors and cloud providers, creating both substitution and squeeze pressure.
Composite Assessment
Technology scores 6.5. Body-free collection has a cost advantage, 20,000 MEgo units show engineering usability, the three walls framework is clear and the world model judgment aligns with the frontier. Deductions for undisclosed retargeting accuracy, no quantified data quality standard and unpublished device parameters.
Business scores 7.0. Three benchmark customers span industrial, research and parent, China Telecom channels and Zhangjiang scenarios offer scale paths and 20,000 deliveries prove mass production. Deductions for undisclosed revenue and margin, unknown concentration and questionable neutrality.
Finance scores 7.0. Multiple rounds with the latest in the hundreds of millions led by China Telecom provide ample capital and Agibot control provides resources. Deductions for undisclosed valuation and revenue and the asset-heavy nature of hardware manufacturing.
Product scores 7.0. MEgo hardware, data platform, closed-loop evaluation and the Hive initiative form a chain from collection to circulation, and the cloud-network plus data bundle has channel value. Deductions for unverified governance capability and unpublished parameters.
Market scores 7.5. Embodied data demand is certain since whichever robot company wins needs data, policy and capital both support it and China Telecom and Zhangjiang resources are scarce. Deductions for competition, possible substitution by synthetic data and customer budgets tied to funding cycles.
Overall 6.9. This is the best-parented and fastest-moving company in embodied data: it used Agibot technology and scenarios, China Telecom cloud and government channels and Zhangjiang campus resources to reach 20,000 device deliveries and three benchmark customers within half a year. Its structural problems are equally clear: the controlling shareholder is a competitor, neutrality is hard to prove, revenue and margin are opaque and data quality lacks quantified metrics. The judgment on physical AI is that data is indeed one of the three walls and Maniformer's body-free collection with platform supply is a viable path to solving it, while data addresses only one third of the problem and representation plus closed loops require model and engineering breakthroughs that cap the ceiling for data suppliers.
References
- Financing announcement of August 17, 2026 reported by Beijing News, Securities Times and China Business Journal: a new round in the hundreds of millions of renminbi led by China Telecom with Zhangjiang Group following and Sequoia China and Yuanqi Innovation adding above pro rata; proceeds for platform-based embodied data supply infrastructure, MEgo production expansion, the collection, governance and evaluation stack and tens of millions of hours of physical interaction data capacity
- Registry and ownership information from Tianyancha and Qichacha: Shanghai Maniformer Embodied Intelligence Technology founded February 2026, legal representative Yao Maoqing, registered capital 5.29 million renminbi, Agibot Innovation Shanghai Technology holding 70.9217 percent
- Personnel information from public reporting: Yao as Agibot partner, senior vice president and head of the embodied business unit, and Maniformer chairman and CEO
- Product delivery and partnership of August 31, 2026 reported by Securities Times: the 20,000th MEgo body-free data capture device off the line and delivered by Yao to JD.com for real business scenarios; a body-free data partnership with Tencent Robotics X announced the same day
- WAIC 2026 remarks of July 19: Yao on the three walls of data, representation and closed loop, and on physical AI moving from demonstration to real work requiring generalizable, optimizable and evolvable embodied systems
- WAIC 2026 world model discussion of July 17: Yao describing world models as systems understanding how the physical world operates with the core function of predicting the next state, noting most work remains at visual generation and citing data type mismatch and insufficient scale
- Company positioning from the website and reporting: a one-stop physical AI data service platform building platform-based supply infrastructure for systematic, standardized and scaled embodied intelligence data, and the Hive Data Co-Creation Initiative for a data circulation network
- Collaboration with China Telecom from public reporting: combining 5G networks, cloud storage, compute scheduling and scenario connectivity with Maniformer collection, dataset production, governance and integrated hardware and software to build standardized cloud-network plus embodied data solutions for humanoid robot firms, smart manufacturing, research institutes, smart parks and smart home customers
- Collaboration with Zhangjiang Group from public reporting: opening campus, industrial and commercial scenarios plus policy, space and supply chain resources, jointly building a public service carrier for embodied data, co-developing high-quality industry datasets and promoting data standards
- Technical background from public sources: body-free collection records human manipulation or interaction with independent devices and retargets into robot action space; advantages are low cost and speed, difficulties are domain gaps between human hands and robot grippers in morphology, degrees of freedom and mechanics
- Industry background from public sources: embodied data comes from real robot teleoperation, body-free collection and simulation; real data is high quality but costly and synthetic data is cheap but carries a sim-to-real gap, with industry consensus favoring a combination
- Competitive landscape from public sources: participants include in-house robot data teams such as Agibot, Unitree and Robot Era, specialized providers such as Maniformer and Guanglun Intelligent, simulation vendors including the NVIDIA Isaac Sim ecosystem and domestic companies, and cloud vendors offering storage, annotation and training platforms
- CLS, October 19, 2026: Pimax received a strategic investment from Maniformer, the Agibot-spun physical-AI data platform, and the two parties established strategic cooperation, with the amount undisclosed; Pimax had already joined the Hive Data Co-Creation Initiative launched by Maniformer together with the Shanghai research institute, the National Data Standards Committee and the CCID Research Institute, alongside the Beijing Humanoid Robot Innovation Center, the Shanghai Guodi Center, Lingchu Intelligence and Haitian Ruisheng
- China City News, October 13, 2026, and industrial expo coverage: Maniformer showed the MEgo series, MEgo Engine and the Mifeng Pai crowdsourcing platform at the 26th China International Industry Expo; Mifeng Pai settles pay by the effective data duration approved after review; the MEgo series was selected as a Forbes China Yangtze River Delta New Intelligence Award case
- Xinhua Finance, August 17, 2026: a new round in the hundreds of millions of renminbi led by China Telecom, with more than 20 city innovation centers in China and more than 5 overseas nodes, strategic cooperation with domestic and overseas internet, embodied intelligence and AI companies, and professional training data supplied for foundation models such as Ant Group's LingBot-VLA 2.0
- Forbes China 2025 New Intelligence Award Yangtze River Delta case list and the industrial expo forum report: Shanghai Maniformer Embodied Intelligence Technology was selected as a Yangtze River Delta New Intelligence Award case with the MEgo series body-free data capture terminal, which Forbes China describes as original innovation from zero to one
This report is compiled from public sources for informational purposes only and is not investment advice. The company is private and revenue, margin and customer structure are undisclosed; capacity targets and data quality metrics are company statements without third-party verification, and related judgments are labeled as uncertain. Data is current as of 2 November 2026.
Related papers
Related tools
Related cases
- DrivAerML High-Fidelity Road-Car Aerodynamics Dataset and Surrogates
- AirfRANS High-Fidelity Airfoil Surrogate Benchmark Case
- Physics-Informed CNN for Battery Pack Temperature Fields
- IASC-ASCE Structural Health Monitoring Benchmark and Physics-Informed Damage Identification
- OpenFOAM Industrial Multiphysics Pipeline: From Meshing to Data Supply
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
- 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
- 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
- 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.