MetaX

MetaX Integrated Circuits (Shanghai) Co., Ltd., branded as MetaX, is a China-based fabless GPU designer founded in Shanghai in September 2020. It was founded by former AMD executives and is led by CEO Chen Weiliang. MetaX has built a multi-site footprint with subsidiaries and R and D centers across cities including Beijing, Nanjing, Chengdu, Hangzhou, Shenzhen, Wuhan, and Changsha. On December 17, 2025, MetaX listed on the Shanghai Stock Exchange STAR Market under stock code 688802.

‍

MetaX focuses on general-purpose GPU compute for AI workloads plus graphics acceleration, positioning its offering as a full-stack platform. Its product roadmap spans N-series GPUs for inference, C-series GPUs for general-purpose computing, and G-series GPUs for graphics rendering, supported by the MXMACA software stack. Alongside accelerator cards, the company sells GPU servers and larger “supernode” style systems aimed at data center deployments.

‍

MetaX’s traction has been tied to China’s domestic AI infrastructure build-out and the push for locally sourced compute. The company highlights large cluster deployments in its milestones, including 1,000-card clusters and a total installed base that exceeded 10,000 cards during 2024. In its IPO, MetaX raised about $600 million and its shares surged on the first trading day, underlining strong investor demand for domestic AI chip names.

MetaX Integrated Circuits is a China-based, fabless GPU designer that sells GPU chips and packaged systems into domestic data center and enterprise workloads. It listed onANext on the Shanghai STAR Market on December 17, 2025 under stock code 688802.

‍

How MetaX makes money

‍

1. GPU silicon built around proprietary IP
   MetaX designs GPUs across three product lines: N-series (inference), C-series (general-purpose compute and training), and G-series (graphics and rendering). In its STAR Market listing document, MetaX states that N100 entered mass production in April 2023 and C500 entered mass production in February 2024, anchoring the shift from prototypes to shipments.
   ‍
2. Accelerator cards, servers, and “supernode” system sales
   MetaX sells PCIe accelerator cards and server systems, then scales deployments with multi-GPU interconnect and system-level packaging. Examples include the C500 PCIe GPGPU that highlights MetaXLink multi-GPU interconnection, and the C550 3D Mesh Supernode, positioned for up to 8 servers and 64 cards in a single supernode topology.
   ‍
3. MXMACA software stack as the adoption layer
   MXMACA is the software stack that MetaX markets as compatible with mainstream GPU workflows, with a focus on migration of existing workloads. The company frames ecosystem compatibility as a core part of the value proposition, since GPU buyers usually standardize on a software platform once clusters reach scale.
   ‍

Market position in China
‍

• Fast revenue ramp, still loss-making
  MetaX’s STAR Market listing document reports revenue of RMB 0.43 million (2022), RMB 53.02 million (2023), and RMB 743.07 million (2024). It reports net profit attributable to owners of -RMB 776.97 million (2022), -RMB 871.16 million (2023), and -RMB 1,408.88 million (2024), plus -RMB 232.51 million in Q1 2025.
  ‍
• Policy tailwind and early share
  Reuters reports MetaX at around 1% of China’s AI chip market, shaped by Beijing’s push to replace foreign suppliers. Reuters also reports management guidance that sales should more than double and that the company targets break-even soon after listing.
  ‍
• Competes under a clear global hierarchy
  In its STAR Market listing document, MetaX describes a market structure dominated by Nvidia and AMD, with Nvidia’s CUDA ecosystem cited as a key barrier. That framing matches how investors should think about MetaX’s position: a domestic supplier building share inside China while closing the software and scale gap that decides long-run adoption.

China is MetaX’s core market, with sales focused on domestic cloud and enterprise deployments that need local GPU supply. The company points to scale-out adoption through OAM-style clusters, including a 3,000-card OAM compute cluster in Beijing and a liquid-cooled 3,000-card OAM cluster in Hangzhou. MetaX also states that total compute clusters deployed exceeded 10,000 cards in 2024.

‍

Key traction signals include:

‍

1. Cluster scale: 3,000-card OAM deployments in Beijing and Hangzhou (2024).
   ‍
2. Installed base: over 10,000 cards deployed across compute clusters (2024).
   ‍
3. Revenue ramp: CNY 1.2 billion revenue in the first three quarters of 2025, with net loss narrowing to CNY 345.5 million over the same period.
   ‍

MetaX also uses ecosystem partnerships to move beyond pilots. It showcased its N100 inference GPU in China Mobile partner events and states the product is in mass production with “key customers and partners” building application solutions.

MetaX’s near-term trajectory is shaped by two forces: China’s push to expand domestic AI compute, and MetaX’s execution on a competitive GPU plus software platform. In December 2025, MetaX listed in Shanghai and raised about CNY 4.2 billion, earmarked for R&D in high-performance general-purpose GPUs, next-generation AI inference GPUs, and related technologies.

‍

The addressable China market is expanding fast. Reuters, citing Frost and Sullivan, forecasts China AI chip sales rising to $189 billion by 2029 from $54 billion in 2026.  MetaX also entered public markets with a clear profitability milestone. Reuters reports management expects break-even in 2026, after scaling sales from a small base with about 1% share of China’s AI chip market.

‍

Key growth drivers include

‍

1. Product cadence moving up the stack
   MetaX’s C-series targets data center training and inference, with features like MetaXLink multi-GPU interconnect highlighted on product pages.
   ‍
2. C600 ramp as a 2026 inflection point
   The C600 is positioned as the next-generation flagship GPGPU. SCMP reports a July launch and company commentary pointing to mass production in the first half of 2026.
   ‍
3. Software stack expansion
   MXMACA claims adaptation support for 6,000+ applications and 1,000+ models, framing migration cost as the core adoption lever.
   ‍

Challenges ahead

‍

• Technology gap and supply-chain exposure tied to U.S. technology restrictions, flagged in the IPO risk discussion reported by Reuters.
  ‍
• Ecosystem friction in mainstream LLM tooling, visible in community work on vLLM integration and PyTorch version constraints for the MetaX backend.
  ‍
• Profitability pressure from heavy R&D, with rapid revenue growth but large losses typical for early-stage GPU platforms.