Legal name: IAT Automobile Technology Co., Ltd.
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
China's first listed independent automotive design company (300825), founded in 2007; it has pushed AI across R&D since 2023, launched the Yuanfang, Yufeng and Taiyi 3.0 product system and partnered with NVIDIA Omniverse; 2025 revenue was RMB 975M with a net loss attributable of RMB -250M, while revenue for the first nine months of 2026 reached RMB 925M, up 25.74 percent, with attributable net profit of RMB 36.19M as the company returned to profit.
Executive Summary
IAT Automobile Technology Co., Ltd., listed on the ChiNext board of the Shenzhen Stock Exchange under the ticker 300825, is China's first listed independent automotive design company, founded in 2007 and headquartered in the Beijing Economic-Technological Development Area. Its traditional business is full-process vehicle research and development, core-component development and manufacturing, and new-energy vehicle development, delivered to more than eighty OEM customers across roughly four hundred to five hundred vehicle programmes, with a team exceeding two thousand people, over eighty percent of them in research and development, and five R&D centres across the United States, Japan, Italy, Beijing and Shanghai.
Since 2023 the company has reoriented the entire R&D organisation around AI, and the clearest output is the AI-plus-digital-creation product system launched on 23 December 2025. The system comprises the Yuanfang automotive R&D knowledge large model for drawing parsing, scheme generation and intelligent review, the Fangcun AI-plus-digital-mock-up product for automated clearance and interference detection on a cloud architecture, the Yufeng aerodynamics AI that predicts drag in real time during early design and reverse-generates compliant contours with a claimed two-month shorter aero development cycle, and the Taiyi 3.0 styling AI with 3D generation, video generation, model training and an infinite collaboration canvas. The launch event, co-hosted with the Modular World AI community under the guidance of the Beijing ETDZ information technology industry bureau, also started an ecosystem matrix whose first ten partners include Xirun Digital, Baidu Famou, the Shanghai AI Laboratory, the Institute for AI Industry Research at Tsinghua, Siemens, Stardust Data and Minqiao.
The physics-AI substance sits on two pillars: an NVIDIA stack and an internal surrogate-model programme. The company has built a digital-twin and real-time collaboration environment on NVIDIA Omniverse and OpenUSD, with unified USD asset management through Nucleus, and is developing a 3D high-precision vehicle drag-prediction model on NVIDIA Modulus trained on data from one hundred vehicle CAS models, with stated plans to extend into aeroacoustics, pressure distribution, cooling management and structural physics fields. Its own 2025 presentation at the China Industrial AI Conference reported a geometric neural network trained on large volumes of 2D graphics and 3D CAD/CAE data whose predictions deviate from simulation results by an average of three to five percent. The Qingdao entity has joined the NVIDIA Partner Network as a Solution Advisor, and at the August 2026 NVIDIA Physical AI and Robotics Partner Day the company released a hexapod special-robot concept video as a Solution Advisor: Consultant.
The robotics move is the second curve and is explicitly pre-revenue. The company has established an AI business unit and the controlled subsidiary Artreit, replicating automotive engineering capability into robot development; in July 2026 it released a Foxconn-partnered Turn-Key full-stack platform for robot manufacturers spanning R&D, production, training, sales and investment, with both parties to build a joint robot company; a joint venture named Beijing Futekang Intelligent Robot Technology was registered in September 2026 with Foxconn and IAT among the shareholders; and at the August 2026 World Robot Conference the company signed ecosystem partnerships including with Lightwheel Intelligent. The hexapod release is explicitly a concept video rather than a physical demonstration.
The financial picture is the counterweight to the AI narrative, and the first nine months of 2026 reversed its direction. Revenue was 857.95 million RMB in 2023, 970.98 million in 2024 and 975.15 million in 2025, essentially flat for two years; net profit attributable swung from 35.97 million in 2023 to minus 132.85 million in 2024 and minus 250.29 million in 2025, with gross margin at 6.17 percent and operating cash flow at minus 98.74 million in 2025. R&D spend fell from 530.12 million in 2023 and 219.98 million in 2024 to 53.08 million in 2025, a 75.87 percent decline, which is a materially different story from an AI build-out and is one of the most important facts in this file. Overseas revenue reached 181 million RMB in 2025, up 63.24 percent, at 18.52 percent of total.
The third quarter report of October 29, 2026 changed the financial line of this file. Revenue for the first nine months was 925 million RMB, up 25.74 percent. Attributable net profit was 36.19 million RMB, up 123.99 percent, turning the company profitable; adjusted net profit was 45.06 million, up 128.51 percent; operating cash flow turned positive; gross margin was 24.75 percent, up 11.56 percentage points; and R&D expense grew 52.45 percent, reversing direction. Overseas revenue exceeded 300 million, about 37 percent of the total. The three pressures of 2025, flat revenue, a widening loss and a collapsing R&D line, all eased at once in the first nine months of 2026: revenue returned to double-digit growth, profit turned positive, gross margin recovered to 24.75 percent and R&D expanded again. It should be noted that a cumulative return to profit implies the losses were concentrated in the first half of 2026 and that the third quarter alone turned positive, which is consistent with the order-driven delivery pattern of the YAMATO contract.
Several company-level developments in the same period bear directly on AI commercialisation: the Yufeng aerodynamic agent and the TAI styling tool have received orders from vehicle manufacturers, which is the first order-side signal the company has disclosed for its AI products, though no amount or segment revenue is given; the testing centre obtained CNAS accreditation; Wuhu IAT Power opened on 14 October 2026 with designed capacity of 600,000 electromagnetic clutches per year; and the Foxconn joint venture Beijing Futekang Intelligent Robot Technology was registered on 21 September 2026 with registered capital of 50 million RMB.
The anchor contract is the YAMATO electric-system-kit purchase agreement signed on 25 December 2023, under which YAMATO INDUSTRY CO., LTD., listed on the Tokyo Stock Exchange Standard market under code 7886, purchases Isuzu ELF and Mitsubishi Canter electric system kits for diesel-to-electric conversions of light commercial vehicles, at 155,000 and 153,000 RMB per kit respectively, totalling no more than 5.383 billion RMB including tax across 35,000 kits scheduled from 2024 to 2026, valid until 31 March 2027. The contract is disclosed as a major operating contract and related-party transaction, and the master list's characterisation of it as an electric heavy-truck project is a mislabel corrected in the discrepancies below.
Company Overview
IAT Automobile Technology was founded in 2007 and listed on ChiNext in March 2020, becoming the first listed automotive R&D company on China's A-share market. It is headquartered at the IAT Automotive Science Park in the Beijing Economic-Technological Development Area, is among the first batch of national high-tech enterprises, operates a Ministry of Industry and Information Technology-certified industrial design centre and a Ministry of Human Resources-approved postdoctoral research workstation, and states that R&D personnel exceed eighty percent of headcount across eleven provincial subsidiaries. The company describes itself as the pioneer of turnkey whole-vehicle design and engine and powertrain development services for China's independent automotive design industry, with cumulative cooperation with more than eighty domestic and international OEMs, more than ten of them for over a decade.
Chairman Xuan Qiwu's biography is the company's origin story. Born in Changchun, he entered the Tsinghua University automotive engineering department at sixteen, worked as an engineer at FAW Group, studied at Kyushu University in Japan from 1992, received his doctorate in 1998, and then worked at Mitsubishi Motors where he rose to head of the development headquarters. He returned to China to start a business in 2002 under the national Thousand Talents introduction programme, founded IAT in 2007, and led its ChiNext listing in 2020. His 2026 framing of the company is that twenty years of work with more than eighty OEMs and nearly five hundred vehicle programmes has produced a large R&D database, from styling and interior through engineering to trial production and validation, which he calls the core asset for entering the AI era.
The AI reorganisation is company-wide and began in 2023. The company states that since 2023 it has comprehensively advanced AI across the full R&D process, actively exploring the organic coordination of knowledge engineering, design data and AI models, and building an AI-native automotive digital design system centred on a vehicle user-value and attribute large model, with the overall architecture of a whole-vehicle R&D digital system planned and multiple automotive-design AI models and agent tools developed, some already in verification in partner OEM projects. The work is led by vice-president Liu Yabin, head of the AI.X Lab.
The product surface now spans styling, engineering, performance verification and knowledge. The Taiyi styling AI moved from a text-to-image and image-to-image creation tool into commercial promotion and then into the 3.0 platform; a CAD intelligent modelling assistant accepts text, sketches and CAD files and generates parametric 3D models through natural-language instruction; the performance-verification work uses the NVIDIA Modulus framework to build geometric neural network models; and an automotive design knowledge large model underpins vertical agents for R&D assistance, review, scheme generation and regulatory parsing. In 2026 the company co-published the Agent-Empowered Automotive R&D Design White Paper with the China Association of Automobile Manufacturers, with FAW, Changan, BAIC, Chery, Great Wall, Li Auto and NIO participating in its preparation.
The organisation is deliberately distributed along the automotive value chain. Five R&D centres across the United States, Japan, Italy, Beijing and Shanghai, subsidiaries in eleven provinces, and a Japanese subsidiary that developed the diesel-to-electric conversion vehicles now being adopted by Japan's SBS Holdings, whose fleet of 4,500 trucks included only thirty electric vehicles as of the relevant reporting. The overseas business in 2025 covered R&D, technology licensing and core components under an R&D-plus-manufacturing-plus-ODM model, starting from Japan and radiating to Southeast Asia, the Middle East, Europe and the United States.
No cash balance, order-book total, per-product AI revenue or robot revenue has been disclosed. The only quantitative business signals are the annual-report financials, the YAMATO contract terms and the overseas-revenue share.
Common Misreadings: Four Errors That Change the Analysis
The 5.383 billion RMB YAMATO figure is a contract ceiling, not revenue. The signed agreement caps the total at no more than 5.383 billion RMB including tax, states that actual purchase quantities are determined by effective purchase orders, and requires a supplementary agreement for any order value beyond the cap. The master list's description of the project as an electric heavy-truck programme is also a mislabel: the subject matter is electric system kits for Isuzu ELF and Mitsubishi Canter light commercial vehicles converted from diesel, not heavy trucks.
The physics-AI order narrative circulating in self-media conflates three different things. Retail-investor posts describe more than eight billion RMB of orders on hand with physical-AI-related work at forty to fifty percent, and attach the YAMATO, AIM and HDI contracts plus AI-tool subscriptions into one physics-AI bucket. The company itself has disclosed only the YAMATO-class vehicle and component contracts and has stated that its AI tools are commercialising with orders landing; no order-book total or physical-AI revenue share has been officially disclosed, and the self-media figures are not used in this report.
The widening loss is not explained by an AI spending surge. R&D investment fell 75.87 percent to 53.08 million RMB in 2025 from 219.98 million in 2024 and 530.12 million in 2023, so the loss expansion from minus 132.85 million to minus 250.29 million attributable profit occurred while R&D spend was contracting sharply, alongside a 6.17 percent gross margin and negative operating cash flow. A reader who maps the loss onto an AI investment story has the causality backwards.
But the opposite misreading should also be avoided: the 2025 R&D contraction has already reversed in 2026. Research expense grew 52.45 percent over the first nine months of 2026, gross margin rose 11.56 points to 24.75 percent, and attributable net profit turned positive at 36.19 million RMB. The earlier collapse-in-R&D conclusion of this file holds for 2025 alone and is out of date as a description of the company's current state. The correct framing is that 2023 to 2024 was an investment peak, 2025 a contraction trough and the first nine months of 2026 a recovery period.
The NVIDIA relationship is advisor-level and application-level, not a joint-development or exclusivity agreement. The documented facts are that the Qingdao entity joined the NVIDIA Partner Network as a Solution Advisor, that the company builds on Omniverse, OpenUSD, RTX GPUs and Modulus, and that it appeared at the August 2026 Physical AI and Robotics Partner Day as a Solution Advisor: Consultant. Claims of Thor-ecosystem participation and Isaac Sim platform cooperation circulating in self-media are not confirmed by official sources in this review.
Technology
The technical claim is that twenty years of vehicle-programme data, unified into USD-based digital-twin pipelines and trained into domain models, can compress automotive R&D iteration from weeks to hours and drag development by two months, and that the same engineering capability transfers to robots and aircraft.
The NVIDIA stack is the backbone of the physics-AI claim. On Omniverse and OpenUSD the company has built a digital-twin and real-time collaboration environment, unifying multi-department design data into USD through Nucleus as the asset store, enabling parallel rather than serial workflows across disciplines, using DeepSearch for 3D asset retrieval and Action Graph for component-explosion video production, with a virtual vehicle-review system planned. On Modulus it is developing a 3D high-precision vehicle drag-prediction model trained on one hundred vehicle CAS datasets, with stated extension paths into aeroacoustics, pressure distribution, cooling management and structural physics fields.
The internal surrogate-model programme is the self-developed layer. The 2025 China Industrial AI Conference presentation describes a geometric neural network built on the Modulus framework, trained on large volumes of the company's own 2D graphics and 3D CAD/CAE data, whose predictions deviate from simulation data by an average of three to five percent, which the company presents as engineering-usable and as enabling rapid iteration from concept design to performance evaluation. This is the most technically concrete in-house claim in the file, and its evidence is a conference presentation rather than a published benchmark.
The four products divide the R&D chain cleanly. Yuanfang handles knowledge, drawings and review; Fangcun handles digital-mock-up clearance and interference on cloud HPC; Yufeng handles aerodynamics prediction and inverse contour generation at the design stage; Taiyi handles styling generation and has extended into 3D and video generation, model training and an infinite collaboration canvas, with demonstrated reach into aircraft, robots and a concept pet robot. The chain coverage is genuinely full, which is the company's core argument against point-solution AI vendors.
The physics connection is real but mediated. The company is not a solver vendor: it does not sell a CAE kernel, and its aerodynamic prediction is trained on simulation data generated elsewhere, including through the Modulus framework and third-party ecosystems such as Xirun Digital's Aerocae. Its contribution is in the data asset, the workflow integration and the domain models, which places it closer to an AI-native engineering-services company than to a simulation-software company.
The simulation world-model framing is aspirational and early. The 2026 white-paper cycle and the agent-empowered R&D positioning describe a shift from tool assistance to agent-native design, and the company's chairman frames the goal as reshaping the underlying logic of automotive R&D rather than drawing a few more pictures. The technical content of that ambition, as publicly documented, is the product system above; the world-model layer itself has no published artifact.
Product and Commercial Model
The commercial model has four layers: the traditional vehicle-design and component business that generates the revenue, the overseas ODM and technology-licensing business that is growing fastest, the AI product system that is beginning to commercialise, and the robotics platform business that is pre-revenue.
The traditional business is flat and low-margin. Revenue of 975.15 million RMB in 2025 at a 6.17 percent gross margin, with the vehicle-design and component base essentially unchanged for two years. The strategic response, overseas expansion, is working on growth but not yet on margin structure at scale: overseas revenue of 181 million RMB, up 63.24 percent, now 18.52 percent of the total, with the company stating that overseas business skews to higher-value technology licensing and ODM with better margins than domestic R&D services.
The AI product system is commercialising through multiple channels. The company's vice-president has stated that AI commercialisation must first achieve survival through diversified revenue, including the Taiyi product, integrated appliances, self-built platforms and national key projects, and that AI commercialisation should focus on the core design business while extending to non-design scenarios. Publicly, the Taiyi styling AI is described as in commercial promotion, the CAD assistant and the knowledge-model agents are deployed internally and in partner projects, and the December 2025 launch formalised the productised system with an ecosystem matrix.
The robotics business is a platform play with partners, not a product play. The Turn-Key full-stack platform with Foxconn spans R&D, production, training, sales and investment, and the joint company and the September 2026 registered joint venture are the vehicles for it; the hexapod special robot is a concept video for high-altitude, underground and tunnel inspection and maintenance; ecosystem signings at the World Robot Conference 2026 with partners including Lightwheel Intelligent cover real-scenario deployment, data collection, robot bodies, models, compute and integration services; and the chairman has acknowledged the robot business is still in a cultivation stage focused on platform building and product optimisation, with industrialisation partnerships signed with Shoucheng Holdings and Galaxy General.
The pricing of the AI products is undisclosed. No per-seat, per-project or subscription pricing, no AI-segment revenue and no robot-segment revenue has been published, which means the market cannot yet price the AI transition from the filings alone.
The buyer set is the company's real asset. More than eighty OEM customers, nearly five hundred vehicle programmes, ten million-plus cumulative vehicle sales from its designed models, and a 2026 white-paper coalition spanning FAW, Changan, BAIC, Chery, Great Wall, Li Auto and NIO give the company a distribution and data position that pure AI vendors lack.
Competitive Landscape
IAT competes in three distinct markets and its position differs sharply in each.
In automotive design services, it is the domestic leader by listing status and scale, against other independent design houses and against OEM in-house teams, with the moat being twenty years of programme data, the overseas R&D-centre footprint and the component manufacturing base in Sichuan and elsewhere. The constraint is that design services is a people-and-hours business with structurally low margin, as the 6.17 percent gross margin shows.
In AI-for-R&D tooling, it competes against both international incumbents and AI natives. Against Siemens, Ansys, Altair and Dassault, which are adding AI-assisted exploration, calibration and review to their suites, IAT's differentiation is that it is a user of those ecosystems turned product vendor, with its own programme data and a design-services channel. Against AI natives such as Neural Concept, NAVASTO, now Autodesk's, PhysicsX and Monumo, the comparison is domain-data breadth against algorithmic depth: IAT has the data and the customer relationships, they have the method. The ecosystem-matrix approach, pulling in Siemens, Baidu, the Shanghai AI Laboratory and Tsinghua AIR as partners rather than fighting them all, is the pragmatic read of that asymmetry.
In robotics, it is a new entrant using an automotive-engineering transfer thesis, against robot-native companies and against other automotive-to-robot transfers. The Foxconn partnership is the strongest card, pairing IAT's R&D and engineering methodology with Foxconn's manufacturing and supply-chain management, and the joint venture gives the platform a credible industrialisation path. The hexapod concept and the cultivation-stage framing mean the competitive claim here is optionality, not position.
The physical-AI classification question is the honest limitation. Under the NVIDIA definition of physical AI, IAT's Omniverse digital twins, Modulus-trained drag prediction, dynamics simulation and robot simulation qualify, and financial media now list it among physical-AI names. But the company's own filings frame the same activity as AI plus digital creation in R&D, and the physical-AI label is largely an external classification of an automotive-AI business.
Financing and Financial Position
The company is public, so the financing question is a balance-sheet question. Total assets were 3.1518 billion RMB at the end of 2025 against liabilities of 1.0637 billion, an asset-liability ratio of 33.75 percent, with interest-bearing debt of 559.13 million and current and quick ratios of 1.76 and 1.43. Shareholder count was 28,500 with the top ten holders at 22.50 percent.
The profit trajectory is the central fact. Net profit attributable was plus 35.97 million RMB in 2023, minus 132.85 million in 2024 and minus 250.29 million in 2025, with the 2025 loss 116.96 million wider year on year; non-GAAP deductions were minus 276.69 million; gross margin was 6.17 percent; weighted ROE was minus 11.80 percent, with a second reported figure of minus 12.31 percent on a different basis; earnings per share were minus 0.51; and operating cash flow was minus 98.74 million, down 141.29 percent year on year.
R&D spend collapsed in 2025. From 530.12 million RMB in 2023 and 219.98 million in 2024 to 53.08 million in 2025, a 75.87 percent decline, taking R&D intensity from roughly 22.7 percent in 2024 to 5.44 percent in 2025. Across 2023 to 2025 cumulative R&D was 803.19 million RMB at 28.64 percent of cumulative revenue. The company has paid no cash dividend in 2023, 2024 or 2025, and disclosed a buyback-and-cancellation of 49.98 million in 2023; the 2025 annual general meeting will consider a no-distribution plan, with the company citing its 2026 operating plan and funding needs.
The contract assets are large but conditional. The YAMATO agreement caps at 5.383 billion RMB including tax across 2024 to 2026, valid to 31 March 2027, with quantities set by effective purchase orders and any excess requiring a supplementary agreement; it is disclosed as a major operating contract and related-party transaction. Public records also show a Yazaki joint venture landing in Yizhuang in August 2024, a flying-car joint laboratory unveiled in December 2025, a terminated private placement in June 2024, and shareholder activity in 2026 including an 8.18-million-share reduction by Junwen Yinbao in August and partial pledge release and re-pledge of controlling-shareholder shares in September.
The first nine months of 2026: return to profit and R&D reflow. The third quarter report of October 29, 2026 records revenue of 925 million RMB for the first nine months, up 25.74 percent; attributable net profit of 36.19 million, up 123.99 percent and a return to profit; adjusted net profit of 45.06 million, up 128.51 percent; operating cash flow turning positive; gross margin of 24.75 percent, up 11.56 percentage points; and research expense up 52.45 percent. Overseas revenue exceeded 300 million, about 37 percent of revenue.
These figures make two substantive corrections to the earlier version of this file. First, the conclusion that R&D collapsed applies to 2025 only; in 2026 R&D has expanded again and the direction has reversed. Second, the framing of a widening loss as the dominant risk needs its weight reduced, since the first nine months turned profitable on a cumulative basis and the adjusted figure turned positive too, at 45.06 million, above the attributable figure, which indicates the profit is not driven by non-recurring items.
Four structural items need to be recorded. The Yufeng aerodynamic agent and the TAI styling tool have received orders from vehicle manufacturers, the first order-side disclosure for the AI products, though still without an amount or segment revenue; the testing centre obtained CNAS accreditation; Wuhu IAT Power opened on 14 October 2026 with designed capacity of 600,000 electromagnetic clutches per year, which is component capacity expansion rather than an AI business; and Beijing Futekang Intelligent Robot Technology was registered on 21 September 2026 with registered capital of 50 million RMB, which corroborates the September 2026 registration noted in the earlier version.
The implicit question is what funds the AI and robotics build-out. With losses widening, R&D contracting, no dividend and negative operating cash flow, the AI product system and the robotics platform are being funded from a shrinking base, and the company's own answer, diversified AI commercialisation including appliances and national key projects, is a services-flavoured answer rather than a software-margin answer. The return to profit in the first nine months of 2026 eases this problem, but does not change the underlying state in which AI revenue cannot be priced from disclosure.
DISCREPANCIES
1. YAMATO project characterisation. Convention A: an electric heavy-truck project of 5.383 billion RMB signed in 2023 (master list and self-media). Convention B: a purchase contract for Isuzu ELF and Mitsubishi Canter electric system kits for diesel-to-electric conversion of light commercial vehicles, totalling no more than 5.383 billion RMB including tax (company announcements of 27 October 2023 and 25 December 2023). Ruling: the company announcements are authoritative; the contract subject matter is light-commercial-vehicle electric system kits, not heavy trucks, and the master-list characterisation is recorded here as a mislabel. The amount is a ceiling, not confirmed revenue.
2. Contract amount. Convention A: 5.383 billion RMB (headline figures). Convention B: no more than 5.383 billion RMB including tax, with actual quantities set by effective purchase orders and excess requiring a supplementary agreement (contract terms). Ruling: the figure is a cap on an order-driven contract, and this report does not treat it as booked or scheduled revenue.
3. Vehicle-programme count. Convention A: more than four hundred vehicles developed (company site). Convention B: nearly five hundred (chairman interview, 2026). Convention C: more than two hundred (company statement, 2021). Ruling: the figures are sequential disclosures at different dates, not conflicts; the report uses more than four hundred with the nearly-five-hundred figure attributed to the 2026 interview.
4. OEM customer count. Convention A: more than eighty (company site). Convention B: more than sixty (company statement, 2021). Ruling: sequential disclosures; the current figure is adopted and the 2021 figure is treated as the earlier state.
5. NVIDIA relationship level. Convention A: Omniverse and Modulus technology application with the Qingdao entity as an NVIDIA Partner Network Solution Advisor (NVIDIA official account and company statements). Convention B: a strategic cooperation based on the Omniverse platform for vehicle and robot dynamics simulation (industry press, 2026). Convention C: Solution Advisor: Consultant partner with Thor-ecosystem participation and Isaac Sim platform cooperation (self-media). Ruling: A is official and adopted; B is reported as industry-press framing of the same relationship; C is not confirmed by any official source in this review and is not used.
6. Order book. Convention A: more than eight billion RMB of orders on hand at end-2025 with physical-AI-related work at forty to fifty percent (self-media research note). Convention B: no disclosed order-book total or segment split (absence of company disclosure). Ruling: the self-media figure has no official basis and is not used; the company is treated as not having disclosed an order-book total.
7. Hexapod robot status. Convention A: released at the NVIDIA Physical AI and Robotics Partner Day as a milestone in special robots (company channels). Convention B: explicitly a concept video for future scenarios, not a physical demonstration, per the release's own disclaimer. Ruling: the product is treated as a concept, and the disclaimer is quoted in substance rather than the milestone framing.
8. Related-party nature of the YAMATO contract. Convention A: disclosed as a major operating contract and related-party transaction (announcement title). Convention B: no public explanation of the related-party basis. Ruling: the announcement's own classification is recorded, and no reason for the related-party attribute is inferred.
9. Return on equity. Convention A: minus 11.80 percent weighted (one financial-data provider). Convention B: minus 12.31 percent (another provider). Ruling: both are reported as different calculation bases for the same 2025 annual report, and the weighted figure is used in the text with the second noted.
10. The collapsing R&D line and the first-nine-months return to profit cover different periods. Convention A: 2025 R&D investment of 53.08 million RMB, down 75.87 percent, with attributable net profit of minus 250.29 million (annual-report basis, the original basis of this file). Convention B: first nine months of 2026 revenue of 925 million RMB, up 25.74 percent, attributable net profit of 36.19 million, up 123.99 percent and a return to profit, research expense up 52.45 percent and gross margin of 24.75 percent (third quarter report basis). Ruling: both are true and there is no conflict, since they describe the same company in different periods; this report retains the 2025 figures as historical record and uses the third quarter report basis for any statement about current operating state, R&D direction and risk weighting.
Risks and Due Diligence Gaps
The durability of profitability is now the primary risk. The first nine months of 2026 turned profitable, with attributable net profit of 36.19 million RMB and adjusted net profit of 45.06 million, but this is a cumulative figure, meaning the losses were concentrated in the first half and the third quarter alone turned positive. A gross margin of 24.75 percent against 6.17 percent in 2025 is a one-off step change whose sustainability still needs a full year to validate, and no dividend has been paid cumulatively over the first nine months. Due-diligence action: whether the fourth quarter sustains the return to profit, the cost structure behind the 24.75 percent gross margin in terms of one-off project closeout versus a delivery mix change, and the funding plan for the robotics joint company.
The AI product revenue is unquantified. No pricing, no AI-segment revenue, no renewal data and no robot-segment revenue has been disclosed, so the AI transition cannot be priced from filings. Due-diligence action: segment disclosure or management accounts for the Taiyi, Yuanfang, Yufeng and Fangcun lines, and the revenue mix behind the diversified-commercialisation statement.
The YAMATO contract's execution risk is contractual, not hypothetical. Quantities are order-driven, the cap requires supplementary agreements for excess, the contract expires 31 March 2027, and the underlying demand, diesel-to-electric conversions of used light trucks at Japan's SBS Holdings and similar fleets, depends on a conversion market that is itself policy- and price-sensitive. Due-diligence action: delivered order volumes by year against the 5,000, 10,000 and 20,000 kit schedule, and the concentration of revenue in this contract.
The physics-AI claims are presentation-grade, not benchmark-grade. The three-to-five-percent deviation of the geometric neural network against simulation, the two-month aero-cycle reduction and the hundred-fold-class efficiency framings are company statements at conferences and launch events, with no published third-party benchmark or error methodology. Due-diligence action: a defined benchmark on a standard aero and a standard structural case with error tolerances fixed in advance.
The NVIDIA relationship is advisor-level and therefore replicable. Solution Advisor status and Omniverse and Modulus application are available to many partners, and the company's differentiation is its data and channel rather than exclusive technology access. Due-diligence action: the actual scope and term of the cooperation, and whether any joint-development or exclusivity exists beyond the partner-network status.
The robotics bet is pre-revenue and concept-grade. A concept hexapod video, a Turn-Key platform, a joint venture registered in September 2026 and cultivation-stage framing mean the second curve is optionality, and the chairman has said so. Due-diligence action: the joint company's funding, order pipeline and first product timeline.
Governance and shareholder signals are mixed. A controlling shareholder with partial pledge release and re-pledge in September 2026, a shareholder reduction of 8.18 million shares in August 2026, a terminated private placement in 2024, and a chairman who also serves as president concentrate both direction and risk. Due-diligence action: pledge coverage ratios, reduction plans and the related-party framework around the YAMATO contract.
Overall Assessment
- Technology 6.5. A genuinely full-chain AI product system spanning knowledge, digital mock-up, aerodynamics and styling, an Omniverse and OpenUSD digital-twin pipeline with USD unification and real-time collaboration, a Modulus-based drag-prediction programme trained on one hundred vehicle datasets, and an in-house geometric neural network with a stated three-to-five-percent deviation from simulation. Deducted points: no self-developed solver, performance claims that are presentation-grade without published benchmarks, and a physics-AI layer that is externally classified rather than company-filed.
- Commercialisation 6.5. More than eighty OEM customers, nearly five hundred vehicle programmes, a formalised AI product system with an ecosystem matrix including Siemens, Baidu and Tsinghua AIR, Yufeng and TAI receiving orders from vehicle manufacturers, overseas revenue above 300 million RMB at about 37 percent of revenue in the first nine months of 2026, and a white-paper coalition spanning most major Chinese OEMs. Deducted points: revenue returned to double-digit growth only in 2026 after two flat years, no AI-segment or robot-segment revenue disclosure, and traction that is channels and relationships rather than quantified bookings.
- Financing 6.5. A public company with 3.1518 billion RMB of assets, a 33.75 percent asset-liability ratio and a real contract asset in the YAMATO agreement, with attributable net profit of 36.19 million RMB over the first nine months of 2026 marking a return to profit, adjusted profit turning positive in the same period, operating cash flow turning positive and research expense up 52.45 percent. Deducted points: the 2025 loss had widened to 250.29 million RMB with R&D down 75.87 percent, no dividend cumulatively over the first nine months, pledge and reduction activity in 2026, and AI and robotics revenue contributions that still cannot be priced from disclosure.
- Product maturity 6.5. Four shipped AI products with versioned iteration, a delivered Omniverse-based digital-twin environment, a Turn-Key robotics platform with Foxconn and a registered joint venture, and a concept hexapod with industrial-grade positioning. Deducted points: the robot product is a concept video by the company's own disclaimer, AI pricing and segment revenue are undisclosed, and the physics-AI performance claims lack published error methodology.
- Market position 6.5. A unique A-share position as the first listed independent automotive design company, with a data asset and customer base that pure AI vendors lack, a pragmatic ecosystem strategy that partners with Siemens and Baidu rather than fighting them, and a credible robotics industrialisation path through Foxconn. Deducted points: a structurally low-margin core business, an AI transition whose revenue cannot yet be priced, incumbent and AI-native competition on both flanks, and a physical-AI label that is largely external classification.
References
- IAT company site and about page, including the 2007 founding, the ChiNext listing, the first-listed-independent-automotive-design-company positioning, the five R&D centres, the eleven provincial subsidiaries, the eighty-percent R&D personnel share, the sixty-plus and eighty-plus OEM cooperation counts and the four-hundred-plus vehicle programmes
- IAT company site chairman profile for Xuan Qiwu, including the Tsinghua automotive engineering entry at sixteen, the FAW engineer period, the Kyushu University doctorate in 1998, the Mitsubishi Motors development headquarters role, the 2002 return under the national talent programme, the 2007 founding and the 2020 ChiNext listing
- IAT 2025 annual report financial data as reported by financial media on 29 April 2026, including revenue of 975.15 million RMB, attributable net loss of 250.29 million RMB, gross margin of 6.17 percent, operating cash flow of minus 98.74 million, asset-liability ratio of 33.75 percent, R&D investment of 53.08 million RMB down 75.87 percent, and the three-year cumulative R&D of 803.19 million RMB at 28.64 percent of cumulative revenue
- IAT 2025 no-distribution announcement, including the 2023 to 2025 attributable profit, R&D and revenue figures, the absence of cash dividends and the 49.98 million RMB buyback-and-cancellation in 2023
- IAT announcements on the YAMATO purchase contract of 27 October 2023 and 25 December 2023, including the Isuzu ELF and Mitsubishi Canter electric system kits, the unit prices, the 35,000-kit schedule across 2024 to 2026, the 5.383 billion RMB cap including tax, the validity to 31 March 2027 and the order-driven quantity terms
- China Securities Journal coverage of the SBS Holdings diesel-to-electric introduction context, including the Japanese subsidiary's development role, the fleet composition and the conversion-market framing
- IAT news on the China Industrial AI Conference 2025, including the AI.X Lab presentation by Liu Yabin, the Taiyi styling tool, the CAD intelligent modelling assistant, the Modulus-based geometric neural network with three-to-five-percent deviation, and the automotive design knowledge large model and vertical agents
- IAT news and Shanghai Securities News, China Industrial News and Beijing ETDZ government coverage of the 23 December 2025 AI-plus-digital-creation product launch, including the Yuanfang, Fangcun, Yufeng and Taiyi 3.0 products, the claimed efficiency gains and the first ten ecosystem partners
- NVIDIA official account coverage of IAT's Omniverse, OpenUSD, RTX and Modulus application, including the Nucleus USD unification, DeepSearch, Action Graph, the hundred-vehicle CAS training plan and the Qingdao entity's Solution Advisor status
- China Times coverage of the IAT intelligent-transformation interview, including the Omniverse strategic cooperation and Solution Advisor status, the AI business unit and Artreit subsidiary, the Shoucheng Holdings and Galaxy General industrialisation signings, and the 2025 overseas revenue of 181 million RMB up 63.24 percent at 18.52 percent of revenue
- Securities Times and company channel coverage of the July 2026 Foxconn Turn-Key platform, the August 2026 hexapod special-robot concept video at the NVIDIA Physical AI and Robotics Partner Day with its concept-video disclaimer, and the September 2026 registration of the Foxconn and IAT joint robot company
- China Business Daily and China Association of Automobile Manufacturers coverage of the 2026 Agent-Empowered Automotive R&D Design White Paper and the chairman's data-asset and agent-training framing
- IAT 2026 third quarter report, disclosed October 29, 2026: revenue for the first nine months 925 million RMB up 25.74 percent, attributable net profit 36.19 million up 123.99 percent turning to profit, adjusted net profit 45.06 million up 128.51 percent, operating cash flow turning positive, gross margin 24.75 percent up 11.56 points, research expense up 52.45 percent, overseas revenue above 300 million RMB at about 37 percent of revenue; Yufeng aerodynamic agent and TAI styling tool receiving orders from vehicle manufacturers; testing centre obtaining CNAS accreditation; Wuhu IAT Power opening on 14 October 2026 with designed capacity of 600,000 electromagnetic clutches per year; Beijing Futekang Intelligent Robot Technology registered on 21 September 2026 with registered capital of 50 million RMB
This report is compiled from public sources. It is for reference only and does not constitute investment advice. Data is current as of October 24, 2026; please refer to official company announcements and filings.
Related papers9
- Transolver: A Transformer Solver for Large-Scale Industrial Geometries
- Fourier Neural Operator (Original Paper)
- DeepONet (Original Paper)
- KATOsuper: Surrogate-accelerated neural topology optimization with sensitivity-consistent Fourier neural operators
- Deep-Learning Solvers and Surrogates for Infinity and p-Laplace Problems
- DeepFEAv2: Deep Learning for Transient Finite Element Analysis Beyond Structured Meshes
- Sparse-Observation Atmospheric Thermal Forecasting with Physics-Informed Neural Networks for Climate-Aware Digital Twins
- Analysis of trade-offs in urban heat mitigation using a Bayesian Optimization framework for an urban canopy layer model
- A Fully Implicit Electromagnetic Energy- and Charge-Conserving PIC Model for High-β Plasma Simulations
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
- Wind Turbine Aeroelastic Simulation and Wake Control: A Verifiable Chain from OpenFAST to Reinforcement Learning
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
- 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
- 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.