1011.AI Deep Dive: A Gordon-Bell-Prize-Led Startup Shipping the Xiewu Physics Engine and the Fuyao Edge Inference Framework as 2026 Physical-AI Infrastructure
1011.AI was founded by Peking University professor Yang Chao, a Gordon Bell Prize laureate, began operations in June 2026, and raised an angel round above CNY 100 million; it ships the Xiewu physics simulation engine spanning rigid, flexible, cloth, cable, fluid and particle bodies and the Fuyao edge inference framework, with no disclosed revenue.
Executive Summary
1011.AI (Changsha 1011 Technology Co., Ltd.) is a physical-AI infrastructure startup founded by Professor Yang Chao of Peking University, a member of the team that won China's first Gordon Bell Prize for a ten-billion-core implicit atmospheric simulation on the Sunway TaihuLight supercomputer. The company began operations in June 2026 and raised an angel round above CNY 100 million from a mix of early-stage, industrial and government-guided funds. Its two product lines are Xiewu, a self-built physics simulation engine spanning rigid, flexible, cloth, cable, fluid and particle bodies with in-house contact, friction and deformation solvers, and Fuyao, an edge inference framework using operator fusion, mixed precision and cross-architecture deployment to run large models on edge devices. No revenue has been disclosed.
Company Overview
1011.AI is registered in Changsha and started operating in June 2026. The defining fact of the company is its founder and chief scientist, Yang Chao, a professor at the School of Mathematical Sciences of Peking University and a Gordon Bell Prize laureate. Yang was part of the Sunway TaihuLight team that, in a globally first achievement, ran a ten-billion-core implicit atmospheric simulation with roughly 77 billion degrees of freedom; the work is frequently cited as the first Gordon Bell Prize for a Chinese-led team. The company name, 1011, reads as "one zero one one" in Chinese and is presented as a numerical brand rather than an acronym.
The founding team is described as combining high-performance computing pedigree with engineering delivery. Beyond Yang, the company is reported to have recruited from supercomputing, game-engine and AI-infrastructure circles, though the precise composition of the leadership beyond the founder has not been disclosed in primary filings. The registered entity is a limited liability company; the founding date of the legal person and the start of commercial operations are reported as the same mid-2026 window.
Technology
The technical thesis of 1011.AI is that a physics engine, not a data-driven surrogate, should be the substrate for simulation, training-data generation and digital twins, and that the engine's value lies in solver fidelity rather than in visual plausibility. Xiewu is presented as a unified framework covering rigid-body, soft-body, cloth, cable, fluid and particle assets, with self-built solvers for contact, friction and deformation and with trajectories constrained by dynamics rather than by keyframed animation.
The differentiation the company claims against mainstream game engines such as PhysX and Havok is industrial-grade accuracy and controllability, including the intent to expose the solver in a way that is usable for robot training and for engineering-grade digital twins rather than only for real-time entertainment. The companion product, Fuyao, addresses the deployment side: it applies operator fusion, mixed-precision computation and cross-architecture targeting so that large models can be compressed and run on edge devices such as robots, smart terminals and industrial cameras. The two products are positioned as a stack, simulation on the one side and efficient inference on the other.
A second technical claim is around synthetic data. Because Xiewu produces physically constrained trajectories rather than recorded or scraped data, the company frames the engine as a synthetic-data generator for downstream model training, a positioning that aligns it with the broader physical-AI data-infrastructure theme. The maturity of these solvers in published benchmarks has not been independently verified.
Product and Commercial Model
The product portfolio has two lines. Xiewu is the physics simulation engine, aimed at game and film production, robot training and industrial digital twins; the company has not publicly disclosed whether it is monetised by seat licences, private deployment or cloud API. Fuyao is the edge inference framework, aimed at edge deployment of large models across robotics, smart devices and industrial vision. No paying customers or revenue figures have been disclosed.
The commercial narrative is broad rather than narrow: the same engine is proposed to serve entertainment, embodied-AI training and industrial simulation, which gives a large addressable market but also diffuses the go-to-market focus. As of this writing the company is in the seed stage with an angel round closed and no documented commercial traction beyond product announcements.
Competitive Landscape
The domestic physics-engine field includes in-house engines from large internet and hardware companies and a number of simulation startups, while the global field is defined by NVIDIA's Warp and Omniverse, Unity's PhysX and Havok, with open-source projects such as Taichi providing a strong baseline. 1011.AI's claimed differentiation is the academic-grade solver lineage tied to Yang Chao's supercomputing record and the in-house contact, friction and deformation solvers. The difficulty is that the engine market is either free (open source, game engines bundled) or dominated by platforms with existing distribution, so a standalone engine must demonstrate accuracy and integration advantages that customers will pay for.
Financing and Capital Structure
The angel round exceeded CNY 100 million and was backed by a blend of early-stage, industrial and government-guided investors, including Fenghe Capital, Huakong Fund, Beijing Information Industry Fund, Shanghai Future Industry Fund, Daohe Longterm, Linge Venture Capital, Chengkai Fund, Jinke JunChuang, Zhongguancun Science City and Edge Partners. No subsequent round has been disclosed, and the company remains privately held at the seed stage.
DISCREPANCIES
1. Operating start date. Sources describe the company as beginning operations in June 2026, while the legal-person registration date is reported in the same period without a single precise day being consistently cited. Source grade: OFFICIAL versus MEDIA.
2. Funding amount. The angel round is described as above CNY 100 million, but no exact figure or per-investor allocation has been disclosed. Source grade: MEDIA.
3. Gordon Bell attribution. Yang Chao is described both as a Gordon Bell Prize laureate and as a member of the winning team; the award is a team prize, so the individual-versus-team framing differs across sources. Source grade: OFFICIAL versus MEDIA.
4. Product form. Xiewu is described alternately as a physics engine and as a platform or toolkit, with the precise packaging not yet fixed in public materials. Source grade: MEDIA.
5. Revenue status. The company has disclosed no revenue, no paying customers and no deployment figures; some coverage implies early commercial interest without quantification. Source grade: OFFICIAL, non-disclosure.
6. Valuation. No post-money valuation has been disclosed; any valuation figure circulating is a media estimate. Source grade: MEDIA, estimate only.
Risks and Diligence Gaps
The central gap is the absence of commercial evidence. With no disclosed revenue, no named paying customers and a broad multi-market positioning, the company's traction rests on announcements rather than audited data. A second gap is technical validation: the solver claims, including contact, friction and deformation accuracy and the synthetic-data value proposition, have not been confirmed by independent benchmarks. A third gap is the crowded substrate: physics engines are either free or bundled by platforms with distribution, raising the bar for a standalone entrant. A fourth gap is team disclosure beyond the founder, which limits assessment of execution capacity.
Overall Assessment
Technology scores 7.5. The academic pedigree is the strongest asset: a Gordon Bell-winning supercomputing lineage and an in-house solver stack are hard to replicate quickly, and the Xiewu-plus-Fuyao pairing covers simulation and edge deployment. Deducted points: no independent benchmark, early product maturity, and unproven generalization beyond the founder's track record.
Commercialisation scores 5.5. A large addressable market across entertainment, robotics and industrial simulation is offset by a diffuse go-to-market and zero disclosed revenue. Deducted points: no paying customers, no pricing model, no deployment figures.
Financing scores 7.0. An angel round above CNY 100 million from a diversified investor base is a solid seed-stage outcome. Deducted points: only one round disclosed, no valuation disclosed, government-guided capital implies a long-horizon thesis ahead of commercial proof.
Product maturity scores 6.5. Two distinct products exist and are described with concrete technical mechanism, yet public validation is thin. Deducted points: no benchmark, no customer reference, packaging still evolving.
Market position scores 7.0. The physical-AI infrastructure theme is large and the founder's profile is differentiated, but the engine layer is contested by free and platform-bundled alternatives. Deducted points: crowded substrate, unclear moat against NVIDIA and open source.
The composite view is 6.7. 1011.AI is a founder-pedigree bet on a self-built physics engine as physical-AI infrastructure, with genuine technical provenance but the typical early-stage deficits of no revenue, no independent validation and a diffuse commercial story. The most informative next disclosures will be the first paying customer, an independent solver benchmark, and a clearer decision about which market Xiewu serves first.
References
- 1011.AI official materials and coverage, 2026: Xiewu physics engine spanning rigid, flexible, cloth, cable, fluid and particle bodies; Fuyao edge inference framework with operator fusion and mixed precision.
- Peking University news and academic profiles of Professor Yang Chao, 2026: Gordon Bell Prize participation with the Sunway TaihuLight ten-billion-core implicit atmospheric simulation team, roughly 77 billion degrees of freedom.
- 36Kr and investment-media coverage of the angel round, 2026: above CNY 100 million raised, investors including Fenghe Capital, Huakong Fund, Beijing Information Industry Fund and Shanghai Future Industry Fund.
- Changsha Xiangjiang New Area and local-government coverage, 2026: registration and operations in Changsha, technology-entrepreneurship support context.
- Industry analysis of domestic physics engines, 2026: comparison with PhysX, Havok, NVIDIA Warp and Omniverse, and open-source Taichi as baseline.
- Coverage of edge inference frameworks, 2026: operator fusion, mixed precision and cross-architecture deployment as the dominant efficiency techniques for edge AI.
- Coverage of synthetic-data generation for embodied-AI training, 2026: physics-constrained trajectory generation as an alternative to recorded or scraped data.
- Startup databases and company filings, 2026: legal-person registration, investor roster and seed-stage status.
This report is compiled from public sources and does not constitute investment advice. The company is private with no audited financials; the funding amount, the Gordon Bell attribution, the product packaging and the valuation are drawn from media coverage with no company-disclosed totals, and the solver accuracy and synthetic-data claims have not been independently benchmarked; uncertainties are flagged throughout. Data is current as of 5 November 2026.
Related tools
Related cases
Related reports
- NVIDIA Deep Dive: Building the Full-Stack Physical-AI Infrastructure with PhysicsNeMo, Warp, Cosmos World Foundation Models and Omniverse
- PhysicsX: Pretrained Geometry and Physics for Industrial Engineering at a $2.4B Mark
- 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
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.