MetaX

MetaX (沐曦, also written 沐曦集成电路 / MetaX Integrated Circuits) is a Chinese fabless semiconductor company that designs general-purpose graphics processing units (GPGPUs) and related accelerator products aimed at artificial intelligence training and inference, data-center computing, high-performance computing and graphics workloads. The company was established in Shanghai in 2020 by a team with backgrounds in GPU architecture and chip design, and it operates as a chip designer that outsources wafer fabrication and packaging to foundry and OSAT partners rather than owning fabs. MetaX positions itself in the domestic Chinese GPU market, where demand for AI accelerators and for alternatives to imported high-end GPUs has grown sharply. Its product line is built around the company's own GPU architecture, with product families marketed under the MetaX brand and supported by software stacks, drivers and developer tooling intended to make porting AI frameworks and models onto its hardware practical. The company targets both training clusters and inference deployment, and it works with server OEMs, system integrators and cloud/data-center operators to bring complete accelerator cards and systems to market. As a privately held startup, MetaX has been funded through venture and strategic investment rounds and has been reported among the more closely watched Chinese GPU startups. Its commercial story is tied to the broader build-out of China's AI computing infrastructure and to domestic semiconductor self-sufficiency efforts. The brand is a business-to-business technology brand: it sells chips, accelerator cards and related software to enterprise and institutional customers rather than to consumers, and its reputation rests on performance, software maturity and supply capability rather than on consumer marketing.

MetaX was founded in Shanghai in 2020 as a fabless GPU design company, at a time when Chinese demand for AI accelerators and for domestic alternatives to imported high-end GPUs was accelerating. The founding team brought GPU architecture and chip design experience, and the company set out to build a general-purpose GPU architecture rather than a narrow, single-purpose accelerator. In its early years MetaX focused on architecture development, chip tape-outs and building the software stack needed to run mainstream AI frameworks on its hardware. Because it is fabless, the company relies on foundry and packaging partners for manufacturing, which makes it sensitive to foundry capacity and to export-control conditions affecting advanced process access. The company raised venture and strategic funding across multiple rounds and became one of the more closely watched Chinese GPU startups as AI computing demand surged. Its product roadmap expanded from initial GPU silicon toward complete accelerator cards, server-level solutions and developer tooling, with partnerships with server OEMs and system integrators used to reach data-center customers. MetaX's growth trajectory is tied to China's build-out of AI computing infrastructure and to policy emphasis on semiconductor self-sufficiency. The brand remains a privately held, business-to-business technology supplier whose commercial progress depends on silicon performance, software ecosystem maturity and volume supply capability.

• 2020

  Founded in Shanghai as a fabless GPU design company

Domestic Chinese GPGPU supplier for AI training/inference and data-center computing, competing in the market for alternatives to imported high-end GPUs.

• MetaX general-purpose GPU (GPGPU) product families

  GPU chips and accelerator cards for AI training and inference, data-center and high-performance computing workloads.

• MetaX software stack and developer tools

  Drivers, libraries and tooling intended to port and run mainstream AI frameworks on MetaX hardware.

1. 12.4万亿参数模型发布当天，智源FlagOS让九大芯片同步“开箱即用”163.com