Jingshuo

One-stop AI platform for physical-AI data, world models, and reusable robot skills.

Cover Block

Open sources

Field	Value
Name	Jingshuo
Tagline	One-stop AI platform for physical-AI data, world models, and reusable robot skills. [JINGSHUO]
Headquarters	Kowloon, China [Yicai Global, October 2026]
Founded	2024 [Yicai Global, October 2026]
Business model	B2B [Yicai Global, October 2026]
Industry	Deeptech [Yicai Global, October 2026]
Technology	AI / Machine Learning [JINGSHUO]
Geography	East Asia [Yicai Global, October 2026]
Growth profile	Venture Scale [Yicai Global, October 2026]
Founding team	Corporate spinout [TechNode, October 2026]
Funding label	$10M+ [Yicai Global, October 2026]
Total disclosed	$10,500,000 [Yicai Global, October 2026]

What an Investor Needs First

PUBLIC Jingshuo builds infrastructure for physical AI, combining data collection, dataset services, world models, and reusable robot-skill tooling for autonomous-driving and embodied-AI developers, and it merits attention now because its first outside financing put a newly formed 2024 spinout at a disclosed RMB2.07 billion valuation less than three years after launch [Yicai Global, October 2026]. The company was founded in May 2024 as a WeRide subsidiary, and public reporting indicates it still operates as a majority-owned subsidiary whose results remain consolidated into WeRide's accounts, which cuts both ways for investors: it gives Jingshuo lineage, operating context, and customer adjacency, while leaving some questions around independence and transferability of that advantage [TechNode, October 2026].

Its product stack, as described in company materials and October 2026 coverage, spans EGOK for first-person data capture, large-scale data acquisition and customized data services, WorldEngine for data and model workflows, and GENESIS-Robotics for generative world modeling; the differentiating claim is not a single model layer but an attempt to tie the full data-to-model workflow into one platform for physical-world AI development [Yicai Global, October 2026] [JINGSHUO] [Autonomous Vehicle International, October 2026]. Public evidence on commercial traction is thinner than the product narrative, although the company has said it works with "dozens" of foundation-model and embodied-AI companies and has more than 100 customers and ecosystem partners, claims that should be treated as company-reported until independently corroborated [Yicai Global, October 2026].

The founding bench is more legible than many companies at this age. Founder and CEO Da Huo was WeRide's first employee and helped build engineering teams across China, the United States, and Singapore, while Ming Han has been identified in a WeRide filing as co-founder, president, COO, and director [Yicai Global, October 2026] [stocktitan.net]. That background matters because Jingshuo is selling into a category where operational know-how in autonomy data pipelines can be as important as model research.

On financing, Jingshuo disclosed a RMB70 million, or about US$10.5 million, pre-Series A in October 2026 from Vision Knight Capital, Hong Kong Science and Technology Parks Corporation, and Sirius Capital, with the company framed as a B2B venture-scale infrastructure bet rather than an application-layer software play [Yicai Global, October 2026]. Over the next 12 to 18 months, the key things to watch are whether Jingshuo can convert its WeRide-adjacent origins into repeatable third-party demand, whether named customer evidence begins to appear, and whether the platform evolves from a broad product architecture into a measurable commercial system with validated usage outside company and affiliate claims [TechNode, October 2026] [JINGSHUO, About].

Partially corroborated -- Core funding, valuation, and founder background are supported by Yicai Global and TechNode, but most product breadth and traction claims remain company-reported or single-source.

Taxonomy Snapshot

Axis	Value
Stage	Pre-Series A
Business Model	B2B
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Geography	East Asia
Growth Profile	Venture Scale
Founding Team	Corporate Spinout
Funding	$10M+ disclosed, including RMB70 million pre-Series A [Yicai Global, October 2026]

Inside the Company

PUBLIC Jingshuo presents as a new physical-AI infrastructure company, but the more material fact is where it came from. ] [JINGSHUO, About]. The public framing is B2B and infrastructure-led, with emphasis on autonomous driving, robot control, and broader embodied-AI workflows rather than a consumer application layer [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [JINGSHUO, 自动驾驶 | 景烁科技].

The available public record is thin on legal-entity detail, but the operating narrative is clearer than the corporate paperwork. Jingshuo says it has built a platform that unifies data management, model training, and model evaluation, and its website places that offering in the context of physical-world AI development across autonomous driving and robotics use cases [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [JINGSHUO, Generative AI | JINGSHUO].

A short chronology is visible from public materials. The company was founded in 2024, then by October 2026 had announced its first external financing, a RMB70 million pre-Series A round that valued the business at RMB2.07 billion, according to Yicai Global; that event sits outside the narrow source set requested for this section, so it is better treated here only as context rather than a fully developed milestone [JINGSHUO, About]. Within company-controlled sources, the clearest milestone is product scope expansion: the website now positions Jingshuo as covering the full path from high-quality data through model development to deployment support, and says it operates more than 10 AI data centers worldwide with more than 3,000 experts, claims that remain company-only [JINGSHUO, About] [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.].

Claim stands unchecked -- This section relies primarily on company website materials, with no corroborating state filing or Crunchbase source provided in the research set.

Under the Hood

Reported and inferred

Jingshuo is positioning itself as an infrastructure supplier for physical AI, not simply a dataset vendor. Public materials describe a stack that spans data capture, data management, model training, evaluation, and reusable skill generation for autonomous driving and embodied AI use cases [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [Yicai Global, October 2026]. The most concrete named products in press coverage are EGOK, described as a high-precision first-person data-collection device, WorldEngine, described as a data-and-model platform, and GENESIS-Robotics, described as a generative world model [Yicai Global, October 2026] [Autonomous Vehicle International, October 2026].

The company website pushes the same integrated thesis, stating that the platform unifies data management, model training, and model evaluation, and that it supports customer fine-tuning on proprietary data [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.]. Website copy also says Jingshuo serves the full development cycle for L2 to L4 autonomous driving and offers bespoke datasets for generative models, with expert review and validation in the workflow [JINGSHUO, 自动驾驶 | 景烁科技] [JINGSHUO, Generative AI | JINGSHUO]. Those claims are useful for understanding product scope, but they remain company-described capabilities rather than independently verified deployment evidence.

A useful way to read the stack is as a closed loop between real-world data acquisition and synthetic or generative modeling. Third-party and company-linked coverage refers to WorldEngine, GENESIS-Robotics, and a data asset engine called SkillForge, while one award writeup describes the model as a "real data × generative enhancement" flywheel supporting embodied-intelligence research and commercialization [robothot.com] [10100.com]. The technical ambition is clear, but the public record still leaves several operating questions open, including which modules are broadly commercial versus customized service work, and how much of the value rests on software, proprietary data operations, or WeRide-derived know-how [TechNode, October 2026] [10100.com].

Partially corroborated -- Product names and broad platform scope are corroborated by company materials and October 2026 press coverage, but most capability claims remain company-described rather than independently verified.

Market Research

PUBLIC The market matters now because physical AI developers are running into the same constraint across robotics and autonomous systems: models improve only as fast as teams can gather, label, organize, and reuse real-world data at scale, and Jingshuo is positioning itself in that bottleneck rather than at the application edge [Yicai Global, October 2026] [TechNode, October 2026].

Public evidence on formal market size is thin in the source set, so any TAM, SAM, or SOM statement would overreach. What can be said with confidence is narrower: Jingshuo is targeting budgets tied to autonomous driving, embodied AI, and foundation-model development for physical-world tasks, with offerings that span data acquisition, custom datasets, model infrastructure, and world models [Yicai Global, October 2026] [Autonomous Vehicle International, October 2026] [JINGSHUO]. That places it in an adjacent layer to both data infrastructure software and technical services, rather than in a single cleanly bounded software category.

Demand signals in the reporting are specific enough to frame the opportunity, even if they do not size it. Yicai Global reported that Jingshuo works with “dozens of foundation model and embodied AI companies” and has more than 100 customers and ecosystem partners, though both figures remain company-supplied and are not independently enumerated [Yicai Global, October 2026]. The same report and Jingshuo's website describe a product stack built around first-person data capture, large-scale data acquisition, model training and evaluation workflows, and reusable skill assets, which suggests buyers are looking for fewer point solutions and more integrated tooling across the development cycle [Yicai Global, October 2026] [JINGSHUO] [JINGSHUO, About].

A second tailwind is that the market boundary is widening beyond self-driving. Jingshuo's own positioning spans autonomous driving, robot control, embodied intelligence, and generative AI data services, while third-party coverage repeatedly describes the company as physical-AI infrastructure rather than only an AV supplier [JINGSHUO] [JINGSHUO, 自动驾驶 | 景烁科技] [JINGSHUO, Generative AI | JINGSHUO] [TechNode, October 2026]. If that framing holds, the relevant spend pool includes not just AV developers but also robotics model builders and research-to-commercial teams that need simulation, world models, and task-specific data pipelines.

The adjacent and substitute markets are clearer than the core market size. Jingshuo could be selling into budgets that might otherwise go to in-house data operations, outsourced annotation and dataset vendors, simulation and synthetic-data tooling, model-ops platforms, or vertically integrated robotics software teams [Yicai Global, October 2026] [Autonomous Vehicle International, October 2026] [JINGSHUO]. The strategic question is whether buyers want a bundled infrastructure layer, which is Jingshuo's thesis, or prefer to assemble separate providers for capture, labeling, model training, and deployment.

Macro and regulatory context also help explain the timing. Autonomous systems and embodied AI both face a practical compliance burden: training data quality, operating-environment coverage, and evaluation workflows become more important as systems move closer to real-world deployment, especially in transport and robotics settings [JINGSHUO, 自动驾驶 | 景烁科技] [TechNode, October 2026]. At the same time, the company's roots inside WeRide give it proximity to an operator with nearly a decade of commercialization experience in physical AI, according to company-linked coverage, which may matter in a market where buyers are skeptical of purely theoretical tooling [10100.com] [TechNode, October 2026].

Cited market signal	Figure	Scope and caveat
First external financing	RMB70 million	Confirms investor appetite for Jingshuo's category positioning, not end-market size [Yicai Global, October 2026]
Post-money valuation disclosed	RMB2.07 billion	A pricing signal on future expectations, not a TAM proxy [Yicai Global, October 2026]
Customers and ecosystem partners	More than 100	Company-supplied adoption signal, not independently enumerated [Yicai Global, October 2026]
AI data centers worldwide	Over 10	Company-supplied capacity signal, not independently audited [JINGSHUO, About]

The table underscores the main limitation in this market read: there are public operating signals and capital-market signals, but no named third-party market-sizing study in the provided evidence. The opportunity looks real at the workflow level, especially where physical-world data is scarce and expensive, but the size of the addressable spend still needs to be triangulated from external industry reports before conviction can move beyond a qualitative view.

Claim stands unchecked -- This section relies on named publisher coverage for financing and positioning, but the available market-demand and operating-scale signals are materially company-supplied and lack independent third-party market-sizing corroboration.

Competition and Substitutes

MIXED Jingshuo is competing less as a model lab and more as an infrastructure supplier to physical-AI builders, which places it in a narrower lane than general-purpose AI vendors but a more crowded one than the branding suggests [Yicai Global, October 2026] [JINGSHUO] [TechNode, October 2026].

The clearest way to map the field is by function, because the public record does not yet name direct startup rivals. In data collection and labeling, Jingshuo is up against incumbent data-service firms and internal tooling teams at autonomous-driving and robotics companies, even if those alternatives are not identified by name in the available reporting [JINGSHUO, 自动驾驶 | 景烁科技] [JINGSHUO, Generative AI | JINGSHUO]. In model infrastructure, the substitute set broadens again, from in-house world-model stacks to general machine-learning platforms that can be adapted for robotics workflows, while at the application layer many customers may choose to build reusable skills internally rather than procure them from a third party [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [Yicai Global, October 2026].

That leaves Jingshuo with a position that is interesting but still partly unproven in public. The company says it offers a one-stop stack spanning data management, model training, evaluation, and reusable skill assets, with products including EGOK, WorldEngine, GENESIS-Robotics, and, in secondary coverage, SkillForge [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [Yicai Global, October 2026] [robothot.com]. If that integration works in production, it can matter because embodied-AI teams often face handoff friction between raw data capture, simulation or world modeling, and downstream deployment. The evidence for commercial differentiation, though, is still mostly company-led or repeated through funding coverage rather than independent customer case studies [Yicai Global, October 2026] [Autonomous Vehicle International, October 2026].

The edge visible today is distribution through WeRide and the operating history that comes with that lineage. Jingshuo was founded as a WeRide subsidiary in 2024, and TechNode reported in October 2026 that it remained majority-owned by WeRide, operated independently, and continued to have its financial results consolidated into WeRide's accounts [TechNode, October 2026]. Yicai Global also reported that founder Da Huo was WeRide's first employee and helped establish engineering teams in China, the United States, and Singapore, which suggests the company did not start from a standing stop on technical hiring or customer access [Yicai Global, October 2026]. The durability of that advantage is mixed: parent-company credibility can accelerate pilots and product feedback, but it is perishable if outside customers come to view Jingshuo primarily as a captive toolchain for WeRide-related use cases rather than a neutral supplier.

The exposure is the mirror image of that strength. A buyer that wants vendor neutrality, broad third-party ecosystem support, or evidence across many production deployments may prefer internal builds or independent infrastructure vendors over a recently financed spinout that still sits inside a listed parent structure [TechNode, October 2026]. Jingshuo also does not appear, from the public record here, to own a customer channel outside physical-AI developers themselves. That matters because adjacent platform companies with stronger cloud, developer-tooling, or simulation distribution could still intercept budget even if Jingshuo's underlying data assets are useful [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [JINGSHUO, About].

The most plausible 18-month scenario is a sorting between integrated specialists and internal platform efforts. WeRide is the most obvious winner if physical-AI customers increasingly value a supplier with operating exposure to autonomous systems and a closed-loop data pipeline, because Jingshuo's positioning and credibility are tightly linked to that parentage [TechNode, October 2026] [Yicai Global, October 2026]. Jingshuo is the likeliest loser if customers decide the highest-value layers are proprietary world models and task policies built in-house, leaving third-party vendors to compete mainly on data services and tooling, where differentiation tends to compress absent named reference accounts or independently verified performance gains [JINGSHUO, One-stop AI Platform: Build Smarter, Launch Faster.] [Yicai Global, October 2026].

Opportunity

Upside case

PUBLIC The prize here is straightforward: if Jingshuo can become the infrastructure layer that physical-AI developers use to collect real-world data, train world models, and operationalize reusable robot skills, it could sit in a part of the stack that expands alongside autonomous driving and embodied AI rather than depending on any single application winner [Yicai Global, October 2026] [JINGSHUO] [TechNode, October 2026].

The headline opportunity is not that Jingshuo becomes another point solution for annotation or simulation. The more ambitious, and still source-grounded, case is that it becomes a default operating layer for physical-AI development in China and potentially broader East Asia: data capture through EGOK, large-scale acquisition and custom data services, a data-and-model platform in WorldEngine, and a generative world model in GENESIS-Robotics are already described as parts of one stack rather than separate products [Yicai Global, October 2026] [Autonomous Vehicle International, October 2026]. That matters because infrastructure vendors usually gain more durable positions when they touch multiple steps in a customer's workflow. Jingshuo's reachability case rests on three public facts: it emerged from WeRide's operating history in autonomous systems, it remains linked to that parent through majority ownership and financial consolidation, and it has already attracted external capital at a RMB2.07 billion valuation in its first outside round [TechNode, October 2026] [Yicai Global, October 2026].

The growth paths are easier to see when separated into distinct scenarios rather than treated as one broad AI bet.

Scenario	What happens	Catalyst	Why it's plausible
AV infrastructure standard	Jingshuo becomes a repeat supplier of data, tooling, and model infrastructure for autonomous-driving teams that need full-cycle support from data collection to evaluation	Expansion of its L2-L4 autonomous-driving tooling and broader adoption of WorldEngine inside WeRide-adjacent and third-party programs [JINGSHUO, 自动驾驶 \| 景烁科技] [Yicai Global, October 2026]	The company already positions itself around autonomous-driving development workflows, and founder Da Huo helped build WeRide engineering teams across China, the US, and Singapore [Yicai Global, October 2026]
Embodied-AI picks-and-shovels layer	Jingshuo becomes a common backend for robotics and embodied-AI groups that need real-world datasets, generative world models, and skill assets	Commercial uptake of GENESIS-Robotics and the broader "data-to-model-to-skill" stack as robotics companies move from labs to deployment [Yicai Global, October 2026] [10100.com]	Public reporting says Jingshuo serves "dozens of foundation model and embodied AI companies," and its stated products are already organized around that customer set, though the customer names remain undisclosed [Yicai Global, October 2026]
Data-center network scale-up	Jingshuo turns its global data-center footprint into a supply-side advantage for hard-to-source physical-AI data and faster iteration cycles	Better utilization of its reported network of more than 10 AI data centers and its expert workforce to expand data coverage across tasks and environments [JINGSHUO, About] [Yicai Global, October 2026]	If the company-operated footprint is directionally accurate, it suggests an attempt to solve a real bottleneck in physical AI: collecting and processing domain-specific data at operational scale [JINGSHUO, About]

What compounding looks like, if it works, is a closed loop rather than a one-time services sale. The company's own product framing ties data management, model training, and evaluation into one platform, while outside reporting adds a skill layer and cites a "real data × generative enhancement" flywheel [JINGSHUO] [10100.com] [Yicai Global, October 2026]. In practical terms, each additional customer workflow could improve three things at once: broader data coverage, better tuned models for specific operating conditions, and more reusable task libraries. That kind of accumulation can matter in physical AI because the hard part is often not model access but operationalizing enough high-quality domain data to make systems work in varied environments. The early signs are still mostly company-described, but the combination of reported customer breadth, product breadth, and a parent with nearly a decade of commercialization experience in the category gives the flywheel a plausible starting point rather than a purely theoretical one [Yicai Global, October 2026] [10100.com] [TechNode, October 2026].

The size of the win is best framed from the private-market marker already on the table, because the source set here does not include a defensible public comparable or third-party market-size study. Jingshuo's first external financing valued it at RMB2.07 billion, or about US$311 million, in October 2026 [Yicai Global, October 2026]. If the embodied-AI infrastructure scenario plays out and the company proves that its platform can serve as a cross-program system of record for data, models, and skill assets, a multi-billion-dollar outcome is conceivable from this base, particularly because infrastructure businesses that become embedded in customer development workflows often compound across programs and renewal cycles. That said, any statement above the last disclosed valuation should be read as a scenario, not a forecast, because public evidence here is still thin on revenue, retention, and named customer depth [Yicai Global, October 2026] [TechNode, October 2026].

Reasoned from indirect evidence -- Core funding and ownership facts are confirmed by Yicai Global and TechNode, but much of the upside case rests on company-described product breadth, reported customer counts without named accounts, and scenario analysis rather than independently corroborated operating metrics.

Sources