NEWS  /  Analysis

China GPU Maker Metax Shares Surge 700% on Debut Amid China’s AI Chip Push

By  xinyue  Dec 16, 2025, 10:50 p.m. ET

Founded by former AMD executive Chen Weiliang, MetaX raised approximately $600 million in an initial public offering last week, just days after larger rival Moore Threads surged 400% in its own debut.

MetaX Integrated Circuits saw its shares soar 700% in their market debut on Wednesday, as investors looked to capitalize on China’s government-driven efforts to reduce reliance on U.S. AI chip giants Nvidia and AMD.

Founded by former AMD executive Chen Weiliang, MetaX raised approximately $600 million in an initial public offering last week, just days after larger rival Moore Threads surged 400% in its own debut.

The company, which trades under the stock code 688802.SH, priced its initial public offering at 104.66 yuan per share, raising 4.2 billion yuan ($590 million). Proceeds will be used primarily to fund research and industrialisation of next-generation high-performance general-purpose GPUs, artificial intelligence inference chips, and advanced GPU technologies for emerging applications, Metax said in its prospectus.

Founded in 2020, Metax is one of China’s leading developers of high-performance general-purpose graphics processing units, focusing on products for AI training and inference, general-purpose computing and graphics rendering. The company develops and sells GPU chips alongside supporting software stacks and computing platforms, positioning itself as a full-stack provider in a market long dominated by U.S. firms.

Metax said its products rank among the most advanced domestically in terms of versatility, core performance, commercial readiness and “independent controllability”, a term widely used in China to describe technologies that are not reliant on foreign suppliers.

The listing comes amid strong policy support for technology innovation, particularly in areas regarded as strategically critical, such as integrated circuits and computing infrastructure. Under recommendations tied to China’s 15th Five-Year Plan, authorities have called for decisive breakthroughs in core technologies across the entire semiconductor value chain, while elevating computing power infrastructure to a national strategic priority.

Moore Threads, often called “China’s Nvidia” by analysts, raised $1.1 billion in its Shanghai IPO in late November.

The momentum continues as AI chip startup Biren Technology received approval to sell shares publicly in Hong Kong on Monday. Rival Kunlunxin, owned by Baidu, has also announced plans for a Hong Kong IPO.

Meanwhile, Enflame, another Chinese AI chipmaker, has engaged Citic Securities to prepare for a potential public listing.

Recent reforms to the STAR Market, including a so-called “1+6” policy package covering listing standards, review mechanisms and post-listing supervision, are designed to improve capital access for technology firms engaged in fundamental innovation. Metax’s IPO process took about 170 days from application acceptance to listing, highlighting the faster-track approach regulators have promoted for priority sectors.

Analysts said the successful flotation will allow Metax to accelerate product iteration, expand market share and strengthen its long-term technology pipeline, at a time when Chinese companies are under pressure to localise advanced computing hardware.

Metax has assembled what it describes as a rare, stable and fully structured technical team, with capabilities spanning GPU intellectual property, system-on-chip design, high-speed interconnects and GPU software. The company says it has independently mastered key technologies such as instruction sets, microarchitecture and computing platforms, overcoming bottlenecks that have historically constrained domestic GPU development.

As of the end of March 2025, Metax held 245 domestic invention patents, three integrated circuit layout design rights and 25 software copyrights, according to company filings. It has also participated in multiple government-backed research programmes, building one of the deepest technical reserves among China’s homegrown GPU developers.

Leveraging its self-developed GPU IP and a unified computing architecture, Metax has launched the Xisi N series for AI inference and the Xiyun C series for integrated training, inference and general-purpose computing. In 2023, the Xiyun C500 series achieved what the company described as leading domestic performance levels.

In October 2025, Metax released the Xiyun C600, a GPU based entirely on its proprietary core IP, completing what it called a fully domestic supply chain loop covering chip design, manufacturing, packaging and testing. The product is aimed at providing “secure and controllable” computing power for China’s digital economy. By the end of March 2025, cumulative shipments of Metax GPUs had exceeded 25,000 units.

A key differentiator for the company is its MXMACA software stack, built around its proprietary instruction set and GPU parallel computing engine. Metax is among a small number of Chinese firms able to offer cloud AI developers a full foundational software stack that integrates mainstream algorithm frameworks, computation and communication libraries, programming languages and development tools.

The MXMACA platform is designed to be compatible with international ecosystems, lowering migration costs for users, and has demonstrated stable performance in large-scale cluster training and high-concurrency inference, the company said.

Financially, Metax has reported explosive growth. According to its IPO filings, the company’s compound revenue growth rate over the past three years reached more than 4,000%. Revenue in the first half of 2025 climbed to 915 million yuan, already exceeding full-year revenue for 2024. As of the listing date, Metax had outstanding orders worth 1.43 billion yuan.

The company is pursuing what it calls a “1+6+X” strategy, combining core GPU products with an open software ecosystem to deliver computing power across multiple industries. Priority sectors include finance, healthcare, energy, scientific research, education, transportation and entertainment.

Metax’s GPUs have already been deployed at scale across national AI public computing platforms, telecom operator data centres and commercial AI hubs, and have been used in more than a dozen AI clusters. The company is among a small group of domestic GPU vendors that have achieved large-scale commercial deployment in clusters exceeding 1,000 cards.

As geopolitical tensions and export controls reshape the global semiconductor landscape, Metax’s debut underscores China’s push to cultivate domestic alternatives in high-performance computing, even as competition intensifies and technological hurdles remain formidable.

Please sign in and then enter your comment