SimWE: From In-House Solvers to a Physical-AI Engine, the Domestic CAE Leader and What Comes Next

Executive Summary

SimWE (Beijing Simulation World Engine Co., Ltd.), short for Simulation World Engine, is a Chinese engineering-software company founded in March 2014 in Haidian, Beijing, by Qu Kaifeng, a Tsinghua University PhD in engineering. It develops Simdroid, a general-purpose multi-physics simulation PaaS covering solid mechanics, fluid dynamics, electrodynamics and thermodynamics, and since 2025 has extended the same physics engine into Sim-PI, a physical-AI development platform for embodied-intelligence training data.

Four facts frame the case.

First, the technology is genuinely self-owned. The company states that Simdroid grew from its own kernel work rather than from wrapped commercial solvers: the founders explicitly say that supporting no-code secondary development and independently compiling a simulation APP away from the platform required rebuilding the underlying kernel. Simdroid 1.0 shipped in April 2017; version 6.0 shipped in 2025. The platform carries more than 200,000 digital-twin component models and a unified multi-physics architecture on GPU rather than the scattered-solver architecture typical of incumbent CAE suites.

Second, it has a measurable, unusually specific market win. In IDC's "China Design and R&D Class Industrial Software (CAE) Market Share, 2025" report, the company took first place in China's electronics thermal simulation market with a 14.9% share, ahead of Cadence, Ansys and Siemens. This was the first time a domestic Chinese CAE vendor outranked the global incumbents in a core sub-segment. The company separately reports a 59.3% year-on-year revenue growth across the overall CAE market, and separately states that Simdroid-EC ranked first in China by licensed-sales revenue for electronics thermal simulation in 2025.

Third, the financing is real but bounded. Public records show seven rounds from a 2016 angel round through an RMB 700M round in March 2024 led by CCB Trust and the Greater Bay Area Homeland Development Fund, followed by a C+ round in October 2024. Investors include Tencent (two entities), Sequoia China (Hongshan), Huawei Hubble, Shenzhen Capital Group, Zhongguancun Science City, China Merchants and others. No valuation has ever been disclosed, and no public IPO tutoring filing has been found.

Fourth, the strategic pivot is from selling tools to supplying data. The company's own framing is that the next generation of AI is physical AI and the core of physical AI is simulation, so the same engine that sells thermal analysis licenses is now being used to generate synthetic training data, run interactive reinforcement-learning training, and validate policies before deployment. The "zero real machines, all simulation" pitch has landed in two named industrial deployments: data-centre thermal control with Susu Technology, and an automotive line-side warehouse picking platform with EFORT's Qizhi unit.

Our view: this is a technically credible, well-capitalised domestic CAE vendor that has won a narrow but high-value sub-segment and is betting the company on an adjacent market. The strengths are real - owned solvers, a unified GPU multi-physics architecture, a productised vertical module that actually sells, and a rare combination of research credibility (a national key R&D programme completed in 2026, an academic advisory panel of five academicians) with a 400-person engineering organisation. The weaknesses are equally real: the physical-AI platform has no disclosed commercial traction, the general-purpose CAE market still shows a 3-to-5-year technology gap against international leaders, and the flagship "thousand-fold speedup" claim is a marketing figure whose error bound is not independently reproducible.

Company Overview

The company was registered on March 7, 2014 as Beijing Internet Based Engineering Co., Ltd. and was renamed Beijing Simulation World Engine Co., Ltd. (a joint-stock company) in a brand upgrade announced for 2026, with the registry re-approval dated April 30, 2026. Its English identity was rebranded in the same upgrade: the English name became Simulation World Engine, and the new brand logo SimWE was formally unveiled at the 2026 user ecology conference, replacing the former Internet Based Engineering name used under the old Chinese brand.

Qu Kaifeng, founder, chairman and legal representative, earned his engineering PhD from Tsinghua University in 2009, worked on two national key R&D projects between 2008 and 2011, joined Fisher & Paykel in New Zealand in 2011 and resigned in April 2013 to return to China. The stated origin of the business model is a 2012 insight drawn from the iPhone 4S app store: a "simulation PaaS platform plus simulation APP" arrangement in which expert-developed APPs can be compiled into small standalone programs, decoupling intellectual property from the platform runtime.

The company's own current materials describe close to 400 full-time employees, with R&D and technical staff approaching 70% of the headcount and roughly two-thirds holding a master's degree or above, backed by a postdoctoral research workstation and an advisory panel of five academicians.

Technology

Simdroid is a general multi-physics simulation platform, first released as version 1.0 in April 2017, with version 6.0 shipped in 2025. The technical claim that distinguishes it from incumbent CAE suites is architectural: rather than shipping a set of disconnected solvers, it uses a unified multi-physics kernel on GPU, so that genuinely coupled problems - chip thermal deformation with stress, wafer-level package thermal-stress coupling, forced-draft motor cooling, GIS breaker electromagnetic-fluid-thermal coupling - are solved on one computational model. Coverage spans solid mechanics (implicit structural, explicit dynamics, material point method, multibody dynamics), fluid dynamics, electrodynamics (low- and high-frequency electromagnetics) and thermodynamics, plus meshing and a rendering engine.

On hardware efficiency, the company reports a 20x to 40x simulation speedup on GPU relative to CPU across a set of benchmarked cases, with published case figures of 23x for chip water cooling, 39x for mainboard shock loading and 40x for water-impact flow. On top of that, a large-language-model layer accepts natural-language instructions to drive the simulation workflow and generate simulation APPs, and an AI surrogate model produces second-level predictions that are passed off to a "simulation agent", shortening solution time from hours to seconds in the error-bounded regime. The company publishes one accuracy example: on a thermal case the surrogate predicted a maximum temperature of 62.95 degrees against a numerical result of 62.93 degrees, an absolute error of 0.02 degrees, and matched the minimum temperature of 34.36 degrees exactly.

The physical-AI platform Sim-PI sits on the same engine. It generates diverse simulation scenarios, runs interactive training with built-in imitation-learning and reinforcement-learning frameworks, and supports multi-agent cooperative training for bipedal, quadruped and robotic-arm agents with real-machine linkage for closed-loop validation. The industrial design principle the company calls "zero real machines, all simulation" is meant to remove the cost, cycle time and safety exposure of collecting data on physical hardware.

The commercial logic of the model is a two-sided platform. Expert users develop parameterised simulation APPs in a no-code environment and can compile them into small standalone programs; ordinary engineers consume them from the Simapps cloud store without installing any CAE package, or download them for local repeated use. The Simwe community, described in 2021 reporting as China's largest simulation community, was said to gather close to 1.1M simulation engineers.

Product and Commercial Model

The commercial story is uneven across the product line, and the split matters. Simdroid-EC, the electronics thermal module, is the proven line: the company states it achieved full commercial validation and large-scale deployment in 2024, ranked first in China by licensed-sales revenue for electronics thermal simulation in 2025, and won a 2026 award for outstanding electronics thermal solution innovation at the China International Data Center Liquid Cooling Key Technologies conference in Shenzhen. The 2025 milestone disclosed on the company timeline is a standard product sales contract of considerable size with a leading Chinese ICT company.

The wider vertical module family - Simdroid-IC for chip multi-physics, Simdroid-Emag for two- and three-dimensional electromagnetics, LNG storage tank design, hydrogen fuel cell thermal analysis, electrochemistry, crystal growth - is real and shipped, but the disclosed scale of individual customer adoption outside electronics thermal is thin.

Sim-PI is the strategic bet and the least disclosed part. It has been recognised rather than monetised in public: the "AI-accelerated electronics thermal design simulation" case, jointly delivered with Susu Technology, was listed first among the ten benchmark application scenarios of the national AI application pilot base in the manufacturing field in September 2026, and won a 2026 IDC China Artificial Intelligence Innovation Award in August 2026 for AI industry application. IDC's Market Glance on Chinese industrial AI and agents in 2Q26 places the company in three sub-markets: embodied-intelligence data simulation and development platforms, design-simulation agents, and pilot-plant agents. No revenue, contract value or customer count has been disclosed for Sim-PI.

The clearest external validation of Sim-PI's technical value comes from the industry itself. At the 2026 World Robot Conference, EFORT's Qizhi unit and the company demonstrated a physical-AI simulation training platform for automotive line-side warehouse picking, where more than a thousand part types with different masses, centres of gravity and friction coefficients are modelled 1:1 in simulation so that a robotic arm can practise before deployment. Separately, the company signed ecosystem agreements with Geek+, Yunji Technology and EFORT, and strategic GPU agreements with Moore Threads, MetaX, Dawning and Iluvatar CoreX.

Competitive Landscape

SimWE operates in two overlapping markets, and it leads in one of them.

In electronics thermal simulation it is first with a 14.9% share in 2025, ahead of Cadence, Ansys and Siemens. This is the segment where AI datacentre build-out has made multi-level thermal simulation a mandatory step rather than a discretionary one, from chip packaging and PCB level up to rack and room, and where domestic advanced packaging iteration has opened a window for electro-thermal coupled simulation at the micrometre scale. IDC's analyst Du Yanze characterises the growth as structural rather than cyclical, driven by long-cycle AI infrastructure investment.

In the general-purpose CAE market it is a small participant in a market where Ansys, Siemens and Dassault together hold roughly 40% of the Chinese market. Third-party industry research puts the overall domestic share of Chinese CAE vendors at 21.7% in 2025, up from 9.4% in 2020, with roughly 110 active vendors in China of which fewer than 15 have complete solver kernels. The same research states that high-end CAE retains a 3-to-5-year technology gap against international levels, with import dependence still above 65% in some specialised scenarios.

Within the domestic cohort, the company is distinguished by breadth of owned physics rather than by size. Suochen Technology (A-share 688507) has comparable or broader in-house solver coverage and the only listed comparable; Anshi Asia-Pacific leads domestic structural and electromagnetic simulation with about 3% global share; Holleywell leads domestic electromagnetic simulation; Emmi AI and PhysicsX are AI-native challengers approaching the same problem from the surrogate-model and agent side rather than from a solver foundation. The distinguishing claim of the company is that it owns a physics kernel, runs it on GPU, and is now monetising the same kernel twice - once as analysis software and once as a data engine for embodied AI.

Funding

The financing history is long, well-documented and never accompanied by a disclosed valuation.

• 2016: angel round, including the Haidian Innovation and Entrepreneurship Center.
• March 2019: early strategic financing from Zhongguancun Development Group and others.
• April 2021: Tencent Investment.
• July 2021: strategic round of nearly CNY 300M led by Sequoia China, with strategic industrial capital participating.
• 2022: Zhongguancun Science City.
• June 2023: Shenzhen Capital Group, Shuncheng Capital, Keli Ventures and others.
• March 2024: C round of CNY 700M, jointly led by CCB Trust and the Greater Bay Area Homeland Development Fund, with Zhongguancun Science City, Shuncheng Capital, Galaxy Securities, Keli Ventures, Shenzhen Investment Holdings, Shunjiang Hongshi and China Merchants International participating. The proceeds were earmarked mainly for expanding R&D headcount, refining products and building the ecosystem, including strengthening the solid, fluid and electrodynamic solvers and developing advanced-packaging chip multi-physics modules.
• October 2024: C+ round, including Guangdong Southern Industrial Investment Fund and China Ordnance Equipment Technology Innovation Investment. Amount not disclosed.

Total capital raised is not published on any consistent basis, and the company has never disclosed a valuation. The company is legally a joint-stock company but is not listed, and no IPO tutoring filing was found in the public pre-IPO supervision system as of the report date.

DISCREPANCIES

Four source conflicts were identified and adjudicated item by item.

1. "First in electronics thermal" measured two different ways. IDC's market-share report gives 14.9% share and first place ahead of Cadence, Ansys and Siemens. The company's own wording is first place in China by licensed-sales revenue for electronics thermal simulation in 2025. Both stand, on different measurement bases - market share versus licensed-sales revenue - and they are reported separately rather than merged into a single claim.

2. The "nearly CNY 300M from Sequoia" headline is not the latest round. A company news page titled "Yundao Zhizao (Beijing Internet Based Engineering) completes nearly CNY 300M strategic financing led by Sequoia China" remains live under the former Chinese brand and is easily misread as the most recent raise. Adjudicated: that is the July 2021 round. The most recent publicly disclosed rounds are the CNY 700M C round of March 2024 and the C+ round of October 2024.

3. Round labels differ across databases while amounts agree. One funding database labels March 2024 as a "strategic financing" and July 2021 as an "A round"; another labels them C and A respectively; a third labels the same March 2024 event a "B round". Adjudicated: amounts and investor lists are consistent and reliable; round-letter labels are not standardised and should be cited by date and amount rather than by letter.

4. Speedup: 20x to 40x versus a thousand-fold. The company reports a 20x to 40x GPU-versus-CPU speedup for the solver itself. Third-party industry research quotes the company as achieving up to a near-thousand-fold efficiency gain once AI surrogate models are layered on. Adjudicated: these are different layers and both are retained - the solver-level speedup is the measurable one, while the thousand-fold figure is a compound marketing claim whose end-to-end error bound is not independently reproducible.

Risks and Diligence Gaps

Revenue scale is small and unaudited. Measured against that, Ansys, Siemens and Dassault each run tens of billions of dollars, and the nearest listed domestic comparable, Suochen Technology, posted CNY 466M for 2025. A CNY 100M-scale business with a 400-person engineering organisation is structurally tight, and the burn-rate question that applies to every pre-scale domestic CAE vendor applies here.

Sim-PI's commercial traction is undisclosed. The platform's recognition - a national pilot-base benchmark scenario ranked first, an IDC innovation award, an IDC Market Glance placement in three sub-markets, a named joint delivery with EFORT - is real. None of it is revenue. The pivot to a physical-AI data business is a bet that a new market will pay for the same engine, and the company has not yet shown that a customer will sign a large Sim-PI contract the way a Simdroid-EC licence is sold.

The simulation-to-reality gap is an industry-wide problem, and the company's own framing concedes it. A policy that succeeds in simulation commonly degrades sharply on real hardware, because friction, clearance, material deformation, end-effector characteristics, motor hysteresis, camera angle, lighting and on-site material variation all compound. The company's own marketing line is "zero real machines, all simulation", and it is selling simulation-derived data as a second growth curve on that basis. That claim currently has no public, independently reproducible error benchmark behind it. A company selling simulation-derived data into this market is selling into a measurement problem it has not itself solved.

The general-purpose CAE gap is not closed. Overseas CAE holds roughly 40% of the Chinese market, domestic share overall is 21.7%, and high-end specialised scenarios still show 65%-plus import dependence and a 3-to-5-year technology gap. Winning electronics thermal does not generalise to the structural, electromagnetic and materials domains where Ansys and Siemens are strongest, and the vertical modules beyond electronics thermal lack disclosed scale.

Key-person and multi-founder dependency. The technology narrative, the "simulation PaaS plus simulation APP" model and the industry position all trace to one founder's specific 2012 insight and to a bench of simulation experts. The advisory panel of five academicians and the postdoctoral workstation are unusual assets but are not contractual retention mechanisms.

Valuation opacity. Seven rounds, a CNY 700M lead round and a C+ round, and still no disclosed valuation, no disclosed revenue trajectory after 2024, and no public IPO tutoring filing. An investor or partner cannot mark this position from public information.

Brand transition risk. The company changed both its Chinese and English names and converted to a joint-stock company within roughly a year. Renaming a company that sells to state-owned and large-ICT buyers is operationally sensitive, and the stale "nearly CNY 300M from Sequoia" news page still sitting under the old brand is a live instance of the resulting confusion.

Overall Assessment

• Technology 8.0. Owned kernels rather than wrapped solvers, a unified GPU multi-physics architecture instead of a scattered solver suite, four physics domains with multi-body and material-point methods, more than 200,000 digital-twin models, a published 20x-40x solver speedup with named benchmark cases, an LLM-driven workflow layer, a surrogate-model prediction path with a disclosed accuracy example, and a full physical-AI platform with RL and imitation-learning integration plus real-machine linkage. Deducted for the absence of an independently reproducible public benchmark, the fact that the thousand-fold figure is a compound marketing claim, and the 3-to-5-year gap in high-end general CAE.
• Commercialisation 7.5. One vertical line is genuinely proven: Simdroid-EC at first place in China by licensed-sales revenue, with large-scale deployment, a large standard-product sales contract with a leading ICT company, and an award in the liquid-cooling market. Deducted because the vertical modules other than electronics thermal show no disclosed scale, Sim-PI has recognition but no disclosed revenue, and pricing and recurring-revenue structure are undisclosed.
• Funding 7.5. Seven rounds, a CNY 700M round jointly led by a trust and a national fund, a follow-on C+ round, and a shareholder register including Tencent (two entities), Hongshan, Huawei Hubble, Shenzhen Capital Group, Zhongguancun Science City, China Merchants and Kunming's state-linked funds. Deducted for the absence of any disclosed valuation, the undisclosed amount of the most recent round, and the lack of a public listing path.
• Product maturity 7.5. A coherent ladder: a general platform at version 6.0, a family of shipped vertical modules, an APP store and cloud simulation platform, a physics community, and now a physical-AI platform. The APP model that lets an expert's model be compiled into a standalone program is a genuine architectural answer to the IP objection that slows CAE adoption. Deducted for the early stage of Sim-PI, the thin disclosure of non-thermal vertical adoption, and reliance on partner announcements rather than contract data for the physical-AI narrative.
• Market position 7.5. First place in electronics thermal simulation ahead of Cadence, Ansys and Siemens, which was the first time a domestic Chinese CAE vendor outranked the global incumbents in a core sub-segment, plus a 59.3% revenue growth rate leading the overall CAE market, a national key R&D programme completed in 2026, and a recognised national pilot-base scenario. Deducted for a revenue scale against a market where the top three overseas vendors hold roughly 40%, the "3-to-5-year gap" in high-end CAE, and the fact that a sub-segment lead is not the same as platform dominance.

References

1. IDC, "China Design and R&D Class Industrial Software (CAE) Market Share, 2025" (released August 2026), as cited by Haidian Bao and by the company.
2. SimWE official site, company profile and development timeline, ibe.cn and simwe.com.
3. SimWE news releases: Simdroid-EC 2026 electronics thermal innovation award; completion of the national key R&D programme on multi-physics simulation for high-voltage power equipment; national AI application pilot base benchmark scenario with Susu Technology; IDC China AI Innovation Award; East Lake Forum, October 2026.
4. SimWE and EFORT Qizhi joint announcement, World Robot Conference 2026, line-side warehouse picking physical-AI simulation training platform.
5. Beijing News, 2021 reporting on the Sequoia-led round and the simulation PaaS plus simulation APP model.
6. Taibo industry database, company profile and financing history for Beijing Internet Based Engineering Co., Ltd.
7. Zero2IPO (PEDATA) and Qichamao company records, shareholder and investment change history.
8. Industry research report, "Domestic substitution of simulation applications: the breakout path from catch-up to running alongside" (sgpjbg.com.cn), 2025-2026 CAE market share and technology gap data.
9. Public filings of Suochen Technology (A-share 688507) for listed domestic comparable financial data.

This report is compiled from public sources. It is for reference only and does not constitute investment advice. Data is current as of October 22, 2026; please refer to official company announcements and filings.