Anshi Pacific PERA Series C

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

Anshi Pacific (PERA) is a Beijing domestic-CAE leader founded in 2003; its PERA SIM general simulation, AIROM reduced-order models, AgentX multi-agent system and Jingzhi iGPT/Whale stack move it from industrial simulation to physics AI, with four financing rounds of roughly RMB 120M total and no disclosed valuation, privately held across multiple rounds.

Executive Summary

PERA Global is one of the longest-established Chinese CAE and industrial software companies, starting in 1996 as a reseller of international simulation technology, moving to in-house development in 2003, launching the proprietary PERA platform in 2008, beginning PERA SIM development in 2016 and releasing PERA SIM on December 29, 2020. Headquartered in Beijing, the company was added to the US Entity List in June 2023, among the first Chinese industrial software companies sanctioned.

On technology assets, the company reports more than 550 patents and participation in over 130 international and domestic standards, with its three-dimensional model framework adopted into the international MBSE standard. Customers span aviation, aerospace, shipbuilding, electronics, ordnance, nuclear and rail transit, and the company says all top 100 enterprises in the relevant domestic industries are customers, having served more than 5,000 customers historically.

At the eighth Simulation Technology Application Conference in 2026 the company released three product groups. A new PERA SIM with substantially strengthened structural and fluid solvers and, for the first time, a multiphysics product PERA SIM Multiphysics achieving two-way strong fluid-structure coupling between its own structural and fluid software. PERA SIM AIROM, an intelligent reduced-order model product training fast predictive models on high-fidelity simulation data for parameter evaluation, design space exploration and real-time optimization. And PERA AgentX, an industrial simulation multi-agent system where users describe requirements in natural language and the agent executes tasks, bound to enterprise private knowledge bases to reduce hallucination. The company also released the Jingzhi iGPT industrial large model, certified by the China Academy of Information and Communications Technology at the highest level five and described alongside Huawei and Volcano Engine as one of China's three trusted industrial large models, plus Jingzhi Whale, an enterprise-grade agentic operating system.

In physical AI terms PERA Global represents the agentification of an established CAE vendor. Senior vice president Tian Feng frames it clearly: AI is bringing traditional industrial software giants back to the same starting line, China holds world-factory-scale industrial scenarios and massive data, industrial software is moving from toolization through platformization to agentification, and future software will not wait for human operation but will put physical laws as a constraint on AI so it understands human language while obeying engineering rules.

One framing point matters. This is the Chinese CAE company with the deepest accumulation, the fastest transformation and the least financial and governance disclosure. Its strengths are thirty years of industry accumulation, coverage of top 100 customers and the certainty of domestic substitution created by the Entity List. Its weaknesses are undisclosed revenue and profitability, unquantified AI products and the distance between agent narrative and engineering deployment.

Company Overview

Three Transformations in Thirty Years

The history has three phases. From 1996 to 2003 the company resold international simulation technology as an early mover in Chinese simulation. 2003 was the first turning point, abandoning reseller dependency for in-house development. In 2008 it released the PERA platform, its first proprietary flagship, moving from point tools to full-process collaboration.

2016 was the second turning point, beginning PERA SIM development based on two decades of service to more than 5,000 customers. PERA SIM launched in December 2020 covering structures, thermal, fluids, electromagnetics, acoustics and multiphysics. After 2023 came the third turning point, with Jingzhi iGPT and a shift from simulation tool provider to industrial AI platform provider.

That path is unusual in Chinese industrial software, where most companies start from one product and expand. PERA Global moved from reseller to platform to general simulation to industrial AI, with each jump landing on an industry inflection.

Entity List and Domestic Substitution

Being added to the Entity List in June 2023 is essential context. As one of the first sanctioned Chinese industrial software companies it faces two sides: constraints on technology cooperation and supply chain, with possible restrictions on foreign components, cloud services and development tools, and a sharp increase in the certainty of domestic substitution, since defense and aerospace customers now treat domestic software as mandatory rather than optional.

The company's response was to increase research investment, which is commercially rational, since the Entity List strengthens its domestic position and directing resources to self-controlled technology yields the highest return.

Organization and Ecosystem

Headquartered in Chaoyang district, Beijing, the company co-funded the Beijing Industrial Software Industry Innovation Center in December 2023 under the Beijing municipal economy and information technology bureau, building a parallel verification platform, an application development service platform and an incubation platform for domestic industrial software. In 2024 it was designated among Beijing's first pilot demonstration units for integrating advanced manufacturing and modern services.

On talent it has run Ministry of Education industry-academia cooperation projects for eight consecutive years and built laboratories with Shenzhen Technology University, Shandong University and Wuhan Institute of Technology to bring domestic CAE into classrooms.

Technology

Positioning of PERA SIM

PERA SIM is the technical foundation, positioned as a general simulation platform covering structures, thermal, fluids, electromagnetics, acoustics and multiphysics. Two changes in the 2026 release matter.

First, structural and fluid solver capability was substantially strengthened. Solvers are the core of CAE, and improvement implies handling harder nonlinearities, larger meshes and tougher convergence. The company discloses no specific metrics such as maximum mesh size, parallel efficiency or accuracy on typical cases, so external assessment of the magnitude is impossible.

Second, PERA SIM Multiphysics achieves two-way strong fluid-structure coupling between its own structural and fluid software. Two-way strong coupling carries technical substance, since most domestic CAE offers only one-way coupling, passing fluid pressure to structures, or weak partitioned iteration, while strong coupling requires a unified solution framework and stable coupling algorithms. INTESIM also works in this area, showing it is contested technical ground among domestic vendors.

AIROM and the Value of Reduced-Order Models

PERA SIM AIROM trains fast predictive models on high-fidelity simulation data, placing it in the surrogate and reduced-order model family alongside Vinci, Ansys SimAI and Supreium's DPB.

Its engineering value depends on three factors: training data coverage, since a surrogate is credible only inside the covered parameter space; accuracy retention, where key physical quantities typically must stay within five percent deviation; and inference speed, since reducing two hours to ten minutes is marginal while reducing to seconds enables design loops.

The company discloses no quantified data on these three, only that repeated computation time falls substantially and that future uses include parameter evaluation, design space exploration and real-time optimization.

AgentX and Engineering Design

PERA AgentX is an industrial simulation multi-agent system aimed at letting junior engineers run simulations directly, targeting two obstacles: difficult software operation and lack of engineering experience.

Two design details deserve credit. Standard engineering specifications are built in to help users avoid common setup errors, encoding expert knowledge into the product rather than relying on prompts. And deep integration with enterprise private knowledge bases grounds every analysis in real specifications and historical data, the correct approach to hallucination by constraining generation to verifiable knowledge.

The company adds that customer historical project data including success cases, failure lessons and specific parameter settings can be uploaded so the agent internalizes and reuses them. If realized, this accumulation creates stickiness, since the longer a customer uses it the richer the knowledge base and the higher the switching cost.

The boundary must be stated: the agent does not validate physical results, it orchestrates tools, checks specifications and calls knowledge, so wrong geometry or boundary conditions still produce wrong output. Engineering judgment about whether a simplification is acceptable or a result plausible remains human.

iGPT and Whale as Platform Layer

Jingzhi iGPT is the industrial large model, certified by CAICT in January 2026 at the highest level five and described by the company as one of China's three trusted industrial large models alongside Huawei and Volcano Engine. Jingzhi Whale is an enterprise agentic operating system supporting companies to create, configure, test and manage agents with security, tool invocation, skill management, workflow orchestration and runtime monitoring.

Both are positioned as platforms rather than tools, with iGPT providing industry knowledge and reasoning and Whale providing the runtime and management environment. The commercial logic upgrades the company from selling simulation software to selling industrial AI infrastructure.

Claims of a level five rating and three trusted industrial models need caution: the rating is a compliance assessment under CAICT standards rather than evidence of customer value, and the grouping of three is company or media framing rather than an official ranking.

Technical Shortfalls

First, missing quantification of solver performance, since substantially strengthened capability has no metrics for comparison with Ansys, Siemens or domestic peers.

Second, no validation for AI products, with no public data on AIROM accuracy and speedup or AgentX efficiency.

Third, unproven depth in multiphysics, where two-way fluid-structure coupling is a first step and thermal-fluid-structural-electromagnetic coupling plus convergence in strongly nonlinear scenarios are not disclosed.

Fourth, uncertain platform openness, since whether Whale can connect third-party tools determines whether it is an open platform or a closed ecosystem.

Product and Commercial Model

Product Matrix

Four lines exist. PERA SIM is the general platform with component versions including Fluid 2025 and Mechanical 2025. AIROM is the reduced-order product. AgentX is the multi-agent system. iGPT and Whale are the industrial model and agentic operating system.

Two horizontal products complement them: HySim, a multidisciplinary joint simulation platform with an open heterogeneous plug-in framework integrating CAX tools and data coupling interfaces, recognized as an influential product in China's software city development in 2024; and IFCDB, described as China's first proprietary BIM database management system and a digital twin database.

Customers and Markets

Customers span aviation, aerospace, shipbuilding, electronics, ordnance, nuclear and rail transit, with the company saying all top 100 enterprises in relevant industries are customers. That means near-full coverage of leading customers, though coverage is not revenue, since defense procurement follows budgets and approval cycles and domestic software prices are typically below foreign tools.

A new direction in 2025 is urban governance, with a simulation platform covering urban flooding early warning, gas leak safety simulation, aviation emergency rescue decisions and carbon management. The value is extending simulation from industry to public governance, with local governments as customers under project-based payment.

Business Model and Pricing

The model mixes software licenses, project services and platform subscriptions, with defense customers on licenses plus implementation, enterprises possibly moving to subscription and governments project-based.

Pricing is undisclosed. Domestic CAE typically prices at 30 to 50 percent of foreign tools plus local services. Whether AI products are billed separately is unstated, which directly affects whether AI generates incremental revenue.

Financing and Capital

Public sources indicate four rounds totaling about 120 million renminbi with undisclosed valuation. That is small relative to the company's revenue scale and thirty-year history, suggesting growth financed mainly by operating cash flow and a possibly dispersed shareholder structure.

Competitive Landscape

The Domestic First Tier

The domestic first tier includes PERA Global with the broadest portfolio and longest history, INTESIM with multiphysics coupling, Tenfong in fluids, Yundao with Simdroid and ZWSOFT extending from CAD.

PERA Global's relative advantage is breadth: five physics fields plus coupling plus multidisciplinary optimization plus data management plus an industrial large model plus agents. Breadth attracts large customers wanting one supplier for most needs while risking insufficient depth in each.

Against International Tools

In single-physics capability it faces Ansys, Siemens, Dassault, COMSOL and Altair, with gaps in robustness at extremes, validation case accumulation and ecosystem depth in material libraries, customization and certification.

The Entity List ironically provides exemption from those gaps for defense customers, who cannot buy foreign tools and must use domestic ones even with a gap.

Competition in Industrial AI

In industrial AI it faces two groups: AI features from traditional CAE vendors such as Ansys SimAI and Siemens Simcenter PhysicsAI, and AI-native companies such as PhysicsX, Vinci and Neural Concept plus domestic players Supreium and ShonCloud.

Its differentiation is industry knowledge and customer data, with thirty years of scenarios and knowledge engineering from thousands of enterprises constituting data assets AI-native companies lack.

Competition in Agent Platforms

Jingzhi Whale competes with platform offerings from Alibaba, Baidu and Tencent and with vertical agent platforms. Its differentiation is industrial scenarios and engineering specifications, with a disadvantage in platform engineering such as multi-tenancy, permissions, audit and service levels.

Financing and Capital Structure

History

Four rounds totaling about 120 million renminbi is low for Chinese CAE, where INTESIM alone raised a Series B in the 100 million renminbi range, indicating growth funded by operations rather than financing.

Revenue and Profit

Revenue and profit are undisclosed. Given thirty years, top 100 coverage and more than 5,000 customers, revenue is likely in the hundreds of millions of renminbi, with profitability unknown since domestic CAE generally carries high research spending and heavy project delivery.

Listing Prospects

No listing plan is disclosed. As one of the first sanctioned domestic industrial software companies with a CAICT-certified industrial model and a complete portfolio, it meets basic conditions for the STAR Market or Beijing Stock Exchange, though historical equity held by multiple rounds and employees may need consolidation.

DISCREPANCIES

1. Two framings of the founding date. The company describes starting in 1996, while registration and business launch dates may differ and some material treats 2003, when in-house development began, as the real starting point. This report uses starting in 1996 with the shift to in-house development in 2003. Source tier: OFFICIAL (company statements).

2. Definitions of customer counts. The company cites more than 5,000 customers as of 2016 and full coverage of top 100 enterprises. The first is cumulative customers and the second a coverage rate, which are not directly comparable. This report labels both. Source tier: OFFICIAL.

3. The three trusted industrial models claim. The company describes iGPT alongside Huawei and Volcano Engine as one of China's three trusted industrial large models, which is company framing based on CAICT assessment rather than an official ranking. This report retains the statement and labels its nature. Source tier: company statement relayed by media.

4. Meaning of the CAICT level five rating. The company says iGPT received the highest level five in CAICT's trusted AI industrial model assessment. The rating is a standards compliance assessment rather than evidence of customer value or technical leadership. This report separates rating from effect. Source tier: OFFICIAL (rating fact).

5. Missing quantification of solver improvements. The company says the new release substantially strengthened structural and fluid solvers without disclosing mesh size, parallel efficiency, convergence or accuracy. This report labels it qualitative. Source tier: OFFICIAL (company statement).

6. Missing effect data for AIROM and AgentX. Neither product has public data on accuracy, speedup or efficiency gains. This report labels them clear in direction with metrics absent. Source tier: OFFICIAL (products exist) with effect data absent.

7. Funding rounds and amounts. Public sources report four rounds totaling about 120 million renminbi without disclosing rounds, investors or valuation. This report uses the public framing and notes its incompleteness. Source tier: MEDIA (third-party compilation).

8. Assessment of Entity List impact. The company says it increased research investment after sanctions without disclosing specific effects on supply chain, cloud services or development tools. This report treats sanctions as both increased substitution certainty and supply chain risk. Source tier: OFFICIAL (sanction fact) plus inference (impact assessment).

Risks and Diligence Gaps

First, financial opacity. Revenue, gross margin and profit are undisclosed, so revenue scale can be estimated from history and coverage while earnings quality cannot be judged.

Second, validation gaps in AI products. AIROM, AgentX, iGPT and Whale have no quantified effect data, while industrial AI value must be proven by engineering time saved and error rates reduced rather than ratings and awards.

Third, risk of over-broad portfolio. Five physics fields plus multiphysics plus optimization plus data management plus a large model plus an agent platform divides resources against specialists in each.

Fourth, platform engineering capability. An enterprise agentic operating system requires multi-tenancy, permissions, audit and service levels, capabilities different from traditional CAE.

Fifth, customer concentration and budget cycles, with defense customers subject to defense budgets and approval pacing and government customers to local fiscal conditions.

Sixth, long-term effects of the Entity List, which aids substitution near term while potentially limiting international recruiting, academic exchange and overseas markets.

Seventh, equity structure and listing, since multiple rounds and employee holdings may disperse ownership and complicate listing applications.

Composite Assessment

Technology scores 7.5. PERA SIM covers five physics fields plus multiphysics with genuine two-way strong fluid-structure coupling, AIROM follows the correct reduced-order route, AgentX's built-in specifications plus private knowledge base correctly counters hallucination, and iGPT holds a CAICT level five rating. Deductions for unquantified solver performance, unvalidated AI products and unproven multiphysics depth.

Business scores 7.0. Top 100 coverage, thirty years of accumulation and Entity List driven substitution certainty form a solid base, and urban governance opens a new scenario. Deductions for undisclosed revenue and profit, unclear AI pricing and unproven platform capability.

Finance scores 6.0. Four rounds of about 120 million renminbi indicate growth funded by operating cash flow. Deductions for complete financial opacity and unknown profitability.

Product scores 7.5. PERA SIM, AIROM, AgentX, iGPT, Whale, HySim and IFCDB form a complete matrix from tools to platform to agents across the simulation workflow. Deductions for portfolio breadth, unknown depth per product and unquantified synergy between AI features and the core business.

Market scores 8.0. Domestic substitution in defense and aerospace is rigid demand, the Entity List strengthens irreplaceability and industrial AI is a policy priority. Deductions for competition in commercial markets and cyclical customer budgets.

Overall 7.2. This is the deepest accumulation and most aggressive transformation in domestic CAE: thirty years from reseller to in-house development, Entity List certainty for substitution, and iGPT, Whale and AgentX pushing industrial software toward the agent era. Its weaknesses are financial opacity and unvalidated AI. The judgment on physical AI is that its physical base in five physics fields and multiphysics coupling is real and Tian Feng's image of putting physical laws as a constraint on AI is among the most accurate metaphors for industrial AI, while moving from tools and platforms to an agentic operating system requires clearing platform engineering and effect validation.

References

  1. PERA Global website and a 2026 profile in the Chaoyang district publication: starting in 1996 as a reseller, in-house development from 2003, PERA platform in 2008, PERA SIM development from 2016 and release on December 29, 2020; added to the US Entity List in June 2023; more than 550 patents and participation in over 130 international and domestic standards with the 3D model framework adopted into the international MBSE standard; more than 5,000 customers historically and full coverage of top 100 enterprises in relevant domestic industries
  2. Eighth Simulation Technology Application Conference in 2026, from company and media reporting: a new PERA SIM with substantially strengthened structural and fluid solvers and the first PERA SIM Multiphysics achieving two-way strong fluid-structure coupling; PERA SIM AIROM training fast predictive models on high-fidelity data for parameter evaluation, design space exploration and real-time optimization; PERA AgentX as an industrial simulation multi-agent system using natural language, built-in standard specifications and enterprise private knowledge bases, with customer historical project data uploadable for experience accumulation
  3. Jingzhi iGPT and Whale from company and media reporting: iGPT certified by the China Academy of Information and Communications Technology in January 2026 at the highest level five and described alongside Huawei and Volcano Engine as one of China's three trusted industrial large models; Jingzhi Whale as an enterprise agentic operating system supporting creation, configuration, testing and management of agents with security, tool invocation, skill management, workflow orchestration and runtime monitoring
  4. Tian Feng, senior vice president, in a company interview: AI bringing traditional industrial software giants back to the same starting line; China holding world-factory-scale scenarios and massive data; industrial software moving from toolization through platformization to agentification; future software putting physical laws as a constraint on AI so it understands human language while obeying engineering rules
  5. Cui Liang, product manager, in a company interview: PERA SIM must be both self-controlled and genuinely usable; current projects face complex assemblies, massive meshes, multiphysics coupling and extensive design iteration, with traditional efficiency at its ceiling
  6. Luo Jiahua, research engineer, in a company interview: PERA AgentX lets junior engineers run simulations directly, targeting difficult software operation and lack of engineering experience
  7. Horizontal products from public sources: HySim as a multidisciplinary joint simulation platform with an open heterogeneous plug-in framework integrating CAX tools, data coupling interfaces and modular joint simulation workflows, recognized in 2024 as an influential product in China's software city development; IFCDB as China's first proprietary BIM database management system and a digital twin database
  8. Product versions from public sources: PERA SIM Fluid 2025 and PERA SIM Mechanical 2025 released in 2025
  9. Urban governance direction from public reporting: a 2025 platform covering urban flooding early warning, gas leak safety simulation, aviation emergency rescue decisions and carbon management
  10. Ecosystem and talent from public reporting: co-founding of the Beijing Industrial Software Industry Innovation Center in December 2023 in Chaoyang under the Beijing municipal economy and information technology bureau; designation in 2024 among Beijing's first pilot demonstration units for integrating advanced manufacturing and modern services; eight consecutive years of Ministry of Education industry-academia cooperation projects and laboratories with Shenzhen Technology University, Shandong University and Wuhan Institute of Technology
  11. Competitive landscape from public sources: the domestic first tier includes PERA Global with the broadest portfolio and longest history, INTESIM with multiphysics coupling, Tenfong in fluids, Yundao with Simdroid and ZWSOFT extending from CAD; international tools include Ansys, Siemens, Dassault, COMSOL and Altair; industrial AI competitors include AI features from traditional vendors such as Ansys SimAI and Siemens Simcenter PhysicsAI and AI-native companies such as PhysicsX, Vinci and Neural Concept
  12. Industry background from public sources: two-way strong fluid-structure coupling requires a unified solution framework and stable coupling algorithms while most domestic CAE offers one-way or weak coupling; reduced-order and surrogate model value depends on training data coverage, accuracy retention typically within five percent on key quantities and inference speed; industrial AI hallucination requires private knowledge bases and engineering specifications as constraints
This report is compiled from public sources for informational purposes only and is not investment advice. The company is private and revenue and profitability are undisclosed; effect metrics for AIROM, AgentX, iGPT and Whale are not public, and related judgments are analytical and labeled as uncertain.

Related reports66

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.