MetaX
MetaX is a Shanghai-based fabless GPU designer building proprietary GPGPU chips and the MXMACA CUDA-compatible software platform for AI training, inference, graphics, and scientific computing, serving Chinese government, cloud, and AI enterprises as a domestic Nvidia alternative.
- Company typePublic
- Founded2020
- HeadquartersShanghai, China
- Headcount501–1,000
- GTM typeB2B
- OfferingHardware or Manufacturing
What MetaX does
MetaX (沐曦股份/MetaX Integrated Circuits (Shanghai) Co., Ltd., ticker 688802) is a Shanghai-based fabless GPU and heterogeneous computing company founded in September 2020 by former AMD and Nvidia engineers. The company designs and sells a four-segment matrix of proprietary general-purpose GPU (GPGPU) chips — the N-series for AI inference (N100, N260, N300), C-series for AI training and general compute (C500, C550, C500X, C588, C600), G-series for graphics rendering (G100), and X-series for scientific AI (X206, X301, X302) — supported by the MXMACA heterogeneous computing software platform that provides CUDA-compatible programming interfaces, support for 40+ AI frameworks including PyTorch, TensorFlow, vLLM, and SGLang, and Day 0 adaptation for 500+ AI models. MetaX also ships GPU servers, workstations, and supernode cluster systems (including liquid-cooled and optical-interconnect configurations) for hyperscale AI deployments.
The company's go-to-market is primarily direct enterprise sales to Chinese government entities, state-owned enterprises, and large domestic internet/cloud/AI companies, supplemented by a channel distribution partnership with Qidian Guofeng, co-selling with Lenovo (DeepSeek all-in-one machine), and integration into major deployments such as the China Unicom $390M Qinghai data center. MetaX generates the majority of revenue from hardware sales — FY2025 revenue was RMB 1.6 billion (~US$232 million), up from RMB 743 million in 2024, with Q1 2026 revenue of RMB 561.9 million (+75% YoY) — supplemented by software/platform services revenue tied to MXMACA deployments and developer ecosystem support. Pricing is not publicly disclosed and is negotiated on a quote basis under multi-year enterprise contracts. The company remains unprofitable with a FY2025 net loss of RMB 779 million, narrowing 43-54% from the prior year as heavy R&D spending continues; it listed on the Shanghai STAR Market in December 2025 and filed for a Hong Kong H-share listing in 2026 to support international expansion.
MetaX firmographics
Firmographics- Name
- MetaX
- Legal name
- 沐曦集成电路（上海）股份有限公司
- Website
- https://metax-tech.com
- Company type
- Public
- Founded year
- 2020
- Operating status
- Operating
- Headcount range
- 501–1,000 employees
- Short description
- MetaX is a Shanghai-based fabless GPU designer building proprietary GPGPU chips and the MXMACA CUDA-compatible software platform for AI training, inference, graphics, and scientific computing, serving Chinese government, cloud, and AI enterprises as a domestic Nvidia alternative.
- Ownership category
- akta.pro rank
MetaX industry classification
Industry- Product category
- GPU Chips and Accelerators
- NAICS
- Computer Systems Design and Related Services (5415), Custom Computer Programming Services (541511)
- SIC
- Electronic Computers (3571), Services-Computer Integrated Systems Design (7373)
- akta.pro primary industry
- AI Accelerators (GPUs/TPUs/NPUs/ASICs) (HDAAAAAA)
- akta.pro secondary industries
- Compute & Acceleration ICs (CPU/GPU/AI/FPGA/ASIC) (HDAHABAC), GPU-Accelerated & AI Training/Inference Servers (HDACABAG), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI)
Keywords
Where MetaX is headquartered
LocationHeadquarters
- HQ city
- Shanghai
- HQ country
- China
- HQ region
- Asia
Offices8 records
Markets served
MetaX business model
Business model- GTM type
- B2B
- Offering type
- Hardware or Manufacturing
- Cost components
- Technology or R&D, Personnel, Operations, Supply Chain, Marketing or Sales, Infrastructure
Revenue model
- GPU Chip and Accelerator Sales: MetaX generates revenue primarily from selling GPU chips (N-series, C-series, G-series, X-series), GPU accelerator cards, GPU servers, workstations, and supernode systems. Revenue more than doubled to RMB 1.6 billion (~$232 million) in full-year 2025 from RMB 743 million a year earlier, driven by robust domestic AI demand and Beijing's domestic substitution policies. The company remains unprofitable with a net loss of RMB 779 million in 2025, narrowing from prior year, as heavy R&D spending continues.
- Software and Platform Services: Revenue from MXMACA software platform, developer community services, and software ecosystem support tied to hardware deployments. PyTorch operator support across all operators and 92% plug-and-play compatibility with ~4,500 active open-source repositories drive software ecosystem revenue.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Other | Multi-year contract | Enterprise GPU hardware — N-series, C-series, G-series, X-series accelerator cards |
Go-to-market motion2 records
Distribution channels4 records
Marketing channels7 records
MetaX product offering
Product offeringCore offering
MetaX designs and sells proprietary GPGPU (General-Purpose GPU) chips and computing platforms across four product lines: N-Series for AI inference (N100/N260/N300), C-Series for training and general-purpose workloads (C500/C550/C500X/C588/C600), G-Series for graphics rendering (G100), and X-Series for AI-for-Science (X206/X301/X302). The chips power GPU servers, workstations, and supernode clusters (liquid-cooled, optical interconnect, 3D Mesh), all running on the proprietary MXMACA heterogeneous computing platform with CUDA compatibility. The offerings serve Chinese government entities, state-owned enterprises, cloud providers, and AI foundation model developers for AI training, inference, graphics rendering, and scientific computing workloads.
Product overview
MetaX (沐曦股份, stock code 688802) is a Shanghai-based GPU designer developing full-stack high-performance GPU chips and computing platforms. Its product portfolio spans four GPU accelerator lines — the N-Series (AI inference), C-Series (general training/inference), G-Series (graphics rendering), and X-Series (scientific AI) — complemented by GPU servers, workstations, and supernode clusters, all supported by the proprietary MXMACA heterogeneous computing platform. The N-Series and C-Series are in mass production. MetaX went public on the Shanghai STAR Market in December 2025 and is pursuing a Hong Kong H-share listing. Founded September 2020 by former AMD/Nvidia engineers.
Differentiator
Problem solved
Functional benefit
Products and services
- 曦思N系列 (N-Series) GPU Accelerator AI inference-focused GPU product line including N100 (with up to 128-channel encoding and 96-channel decoding, supporting 8K resolution, HEVC/H.264/AV1/AVS2 codecs), N260 for large model inference, and N300 AI inference accelerator, delivering high AI compute performance with MXMACA software integration for intelligent video processing, video transcoding, and LLM inference acceleration.
- 曦云C系列 (C-Series) GPU Accelerator General-purpose training and inference GPU line including C500, C550 (next-gen GPGPU with proprietary architecture and domestic advanced process), C500X (optical interconnect general GPU), C588, and C600 (next-gen GPGPU with fully domestic supply chain), supporting large-scale AI model training and deployment for enterprise and cloud customers.
- 曦彩G系列 (G-Series) GPU Accelerator
Quantifiable outcome
- Revenue more than doubled to RMB 1.6B in FY2025 (+115-129% YoY)
- +5 more outcomes
Companies that use MetaX
Customer profileNamed customers7 records
Segments4 records
Ideal customer profiles3 records
MetaX technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
Integration10 records
AI capability9 records
Feature6 records
MetaX partnerships and signals
Strategic signalPartnerships
Ten partnerships are on record, tiered key, potential and minor.
- Multiple Open Source Communities (PyTorch, vLLM, SGLang, Red Hat, CNCF, DragonLzhi/Anolis, Mulan, Moligu Fangan, Dynamia/HAMi, MinerU, FlagOS)keyMetaX partners with 10+ open source communities for its open source zone at WAIC 2026, showcasing AI Infra open collaboration chain from chip compute to operating systems, inference engines, and cloud-native scheduling. Partners include PyTorch Foundation, vLLM, Red Hat, CNCF, DragonLzhi, Mulan Open Source Community, and others.
- UBTECH RoboticskeyMetaX and UBTECH signed a strategic partnership in Nanjing on June 12, 2026, establishing a joint venture Xixuan Chuangzhi Technology (Wuxi) Co., Ltd. with 100 million yuan registered capital to manage R&D and mass production of chips for embodied intelligence. MetaX and UBTECH each hold 35.01% as largest shareholders. This joint venture targets the rapidly growing embodied AI humanoid robot market.
- HKUST & CalmCar (Physical AI Innovation Center)keyMetaX is an inaugural industry partner in the HKUST-CalmCar Physical AI Innovation Center, alongside Tsinghua University, Microsoft Research Asia, Arm China, and HSBC. The center focuses on full-stack physical AI technologies for autonomous driving, robotics, and smart manufacturing, and includes both an Industry-Research Alliance and Capital Alliance.
- Alibaba (T-Head)keyChina Unicom deployed a $390 million data center in Qinghai using chips from Alibaba T-Head, MetaX, Biren Tech, and Zhonghao Xinying, with 20,000 petaflops planned computing capacity. MetaX chips are part of this major domestic chip deployment reducing reliance on foreign technology.
- ByteDancepotentialByteDance is accelerating its shift toward domestic AI chips and is in talks with MetaX Integrated Circuits along with Biren Technology, Moore Threads Technology, and Enflame Technology for potential cloud infrastructure orders, as Beijing pushes for greater use of homegrown semiconductors.
- LenovokeyLenovo and MetaX jointly released a DeepSeek all-in-one machine solution covering 10+ core industries including finance, healthcare, education research, and other sectors, demonstrating a strategic co-selling/partnership relationship.
- Baidu (PaddlePaddle)keyMetaX's 曦云C500 GPU completed Level II compatibility testing with Baidu PaddlePaddle deep learning platform, enabling joint AI ecosystem integration for enterprise customers.
- Zhejiang UniversitykeyIn December 2020, MetaX and Zhejiang University established the 'Zhejiang University-MetaX GPU Chip Design and Application Joint Research Center' to pursue core GPU technology research. In May 2022, MetaX partnered with Tsinghua University's School of Integrated Circuits for reconfigurable architecture design for homomorphic cryptography and genomic sequencing.
- Tsinghua UniversitykeyMetaX partnered with Tsinghua University's School of Integrated Circuits for reconfigurable architecture design for homomorphic cryptography and genomic sequencing applications. Tsinghua University students have conducted multiple industry practice sessions at MetaX as part of the university's talent development programs.
- Qidian Guofeng Holdings (Distributor)minorQidian Guofeng Holdings (1280.HK) pivoted toward AI computing hardware through a distribution partnership with MetaX (688802.SH), distributing MetaX GPU computing products in Hong Kong market.
Scale indicators10 records
Recent moves8 records
Expansion highlights6 records
MetaX competitors and assessment
Company assessmentDirect peers
- Biren Technology: Biren Technology is a Chinese fabless GPU startup developing AI training and inference chips. It is one of MetaX's closest domestic competitors, competing for the same Chinese hyperscaler and government customers with overlapping GPGPU product positioning.
- Moore Threads: Moore Threads is a Chinese GPU design company founded by former Nvidia engineers, focusing on graphics and AI compute GPUs. It directly competes with MetaX in the domestic Chinese GPU market across both consumer and enterprise AI workloads.
- Iluvatar CoreX: Iluvatar CoreX is a Chinese GPGPU/AI chip designer focused on training and inference for data center workloads. It is a direct competitor to MetaX, named alongside MetaX in discussions with ByteDance and China Unicom for domestic AI chip orders.
- Enflame Technology: Enflame Technology is a Chinese AI training chip designer targeting cloud and data center customers. It directly competes with MetaX's C-series training GPUs for the same Chinese cloud provider and enterprise AI training workloads.
- Cambricon Technologies: Cambricon is a publicly listed Chinese AI chip designer with broader coverage including edge AI and cloud training chips. It competes with MetaX across Chinese government, enterprise, and AI inference/training segments, though with a more diversified product portfolio.
- Alibaba T-Head (Pingtouge): Alibaba's chip design arm develops RISC-V and AI inference chips for cloud and edge workloads. T-Head is both a co-deployed partner with MetaX in the China Unicom Qinghai data center and a competitor for internal Alibaba cloud and external Chinese hyperscaler AI chip orders.
Broad incumbents
- Huawei (Ascend / HiSilicon): Huawei's Ascend AI accelerator line and HiSilicon chip design arm constitute a major Chinese AI compute platform widely deployed in government and enterprise data centers. It is a broader incumbent offering full-stack AI compute (chips, software, systems) competing with MetaX across most of MetaX's target segments.
- Nvidia: Nvidia is the global GPU/AI accelerator leader and the dominant incumbent in AI training and inference. MetaX explicitly positions its CUDA-compatible MXMACA platform as an Nvidia replacement in the Chinese market; US export controls on Nvidia create MetaX's core opportunity.
- AMD: AMD is a global GPU and CPU manufacturer with a growing AI accelerator portfolio (Instinct MI series). Several MetaX founders (including CTO Peng Li, an AMD Fellow) built their careers at AMD, and AMD competes with MetaX for GPU/AI compute workloads globally.
- Intel (Habana Labs): Intel is a global semiconductor incumbent with AI accelerator offerings through Habana Labs (Gaudi). It is a broad incumbent competing for AI training/inference workloads globally, though with limited presence in the China-domestic chip market MetaX primarily targets.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
MetaX social profiles
Digital presenceMetaX compliance and trust
Trust signalCompliance1 record
MetaX financial estimates
Financial estimateRevenue estimate
Valuation estimate
MetaX leadership team
Management profileNumber of profiles
Profiles1 record
MetaX subsidiaries and ownership
Company hierarchySubsidiaries1 record
MetaX funding detail
Funding detailFunding overview
Funding rounds4 records
Investors20 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
MetaX M&A and investment
M&A and investmentM&A
Investments3 records
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about MetaX
What does MetaX do?
MetaX designs and sells proprietary GPGPU (General-Purpose GPU) chips and computing platforms across four product lines: N-Series for AI inference (N100/N260/N300), C-Series for training and general-purpose workloads (C500/C550/C500X/C588/C600), G-Series for graphics rendering (G100), and X-Series for AI-for-Science (X206/X301/X302). The chips power GPU servers, workstations, and supernode clusters (liquid-cooled, optical interconnect, 3D Mesh), all running on the proprietary MXMACA heterogeneous computing platform with CUDA compatibility. The offerings serve Chinese government entities, state-owned enterprises, cloud providers, and AI foundation model developers for AI training, inference, graphics rendering, and scientific computing workloads.
Is MetaX a public or private company?
MetaX is a public company. It is classified as public and is currently operating.
When was MetaX founded?
MetaX was founded in 2020. It employs 501 to 1,000 people.
Where is MetaX based?
MetaX is headquartered in Shanghai, China, in the Asia region.
How does MetaX make money?
Two revenue lines are on record. GPU Chip and Accelerator Sales are the primary driver. The others are software and Platform Services.
Who are MetaX's main competitors?
Direct peers on record are Biren Technology, Moore Threads, Iluvatar CoreX, Enflame Technology, Cambricon Technologies and Alibaba T-Head (Pingtouge). Broad incumbents are Huawei (Ascend / HiSilicon), Nvidia, AMD and Intel (Habana Labs).
Does MetaX have an API?
No public API is recorded for MetaX.
What industry is MetaX in?
MetaX's product category is GPU Chips and Accelerators. Its primary akta.pro industry code is HDAAAAAA, AI Accelerators (GPUs/TPUs/NPUs/ASICs), with a secondary code of HDAHABAC, Compute & Acceleration ICs (CPU/GPU/AI/FPGA/ASIC). Its NAICS code is 5415 and its SIC code is 3571.