The first newly accepted A-share IPO application of 2026 has landed in China’s homegrown AI chip sector.

Just now, information from the Shanghai Stock Exchange showed that Shanghai Enflame Technology Co., Ltd.’s STAR Market IPO application has been accepted, with the company seeking to raise 6 billion yuan.

Enflame was valued at nearly 20 billion yuan in its latest funding round, according to available information. Tencent Technology and its concert party, Suzhou Paiyi, together hold 20.258% of the company, making them its largest shareholder.

As of now, Enflame’s STAR Market IPO application has been accepted.

Among China’s so-called “four little dragons” of domestic GPUs, Moore Threads and MetaX have already listed on the STAR Market, while Biren Technology chose Hong Kong.

Enflame’s IPO Push

Enflame plans to raise 6 billion yuan in this offering, with the proceeds highly concentrated and almost entirely earmarked for R&D-related projects.

Specifically, the funds will mainly go to three areas:

1.503 billion yuan for the R&D and industrialization of its fifth-generation AI chip product line

1.196 billion yuan for the R&D and industrialization of its sixth-generation AI chip product line

3.3 billion yuan for an advanced AI software-hardware collaborative innovation project

On the revenue side, the prospectus shows Enflame generated more than 540 million yuan in operating revenue from January to September 2025.

Looking back from 2022 to 2024, Enflame’s main business revenue rose from about 90 million yuan in 2022 to more than 720 million yuan in 2024, implying a two-year compound growth rate of 183.15%.

It was during this period that Enflame’s revenue mix began to shift noticeably.

Since 2023, as its technology iterated and its products saw large-scale adoption, Enflame’s revenue center of gravity has changed materially:

AI accelerator cards and modules, along with AI computing systems and clusters, have become the company’s main revenue sources, accounting for more than 50% of revenue in 2023.

By January to September 2025, Enflame’s third-generation AI accelerator cards had ramped across multiple AI business scenarios at internet customers. Revenue from AI accelerator cards and modules alone during that period had already exceeded the full-year level in 2024.

Notably, Tencent was Enflame’s largest customer from January to September 2025, with related sales accounting for 57.28% of revenue in the period.

On profitability, however, Enflame remains in an investment phase.

Because cloud AI chip R&D requires heavy spending and customer validation cycles are long, the company is still loss-making.

January to September 2025: net loss of about 887 million yuan

2024: net loss of about 1.51 billion yuan

2023: net loss of about 1.66 billion yuan

2022: net loss of about 1.12 billion yuan

Overall, the scale of losses has been gradually narrowing.

According to the prospectus, based on orders on hand, R&D planning, and the cadence of product iteration and delivery, Enflame could reach breakeven as early as 2026.

Looking back at Enflame’s listing process, the company first filed for IPO tutoring in August 2024, with CICC as its tutoring institution.

It then submitted another listing tutoring filing on November 1, 2025, changing its sponsor to CITIC Securities.

By January 1, 2026, the company had completed listing tutoring, and on January 22, 2026, its STAR Market IPO application was reportedly accepted.

If the offering succeeds, Enflame will become the third GPU company listed on the A-share market.

About Enflame Technology

Enflame Technology was founded in March 2018 by Zhao Lidong and Zhang Yalin, two former AMD colleagues.

The company focuses on cloud AI chips, with core businesses spanning three layers: chips and hardware, software and programming platforms, and computing power cluster solutions.

Over the eight years since its founding, Enflame has completed in-house iterations across four generations of architecture and five cloud AI chips. Around its self-developed chips, it has built a full product system covering AI chips, the Suisi series; AI accelerator cards and modules, the Cloudblazer series; AI computing systems and clusters; and AI computing and programming software platforms.

At the hardware level, Enflame designed an original computing architecture based on its own instruction set. Benchmarking NVIDIA’s TensorCore and NVLink systems, it developed its in-house GCU-CARE accelerated computing unit and GCU-LARE high-speed inter-chip interconnect technology.

The architecture targets large-model training and inference scenarios, emphasizing performance release under highly parallel computing while retaining a high degree of programming flexibility.

At the software level, Enflame did not follow the CUDA ecosystem dominated by NVIDIA. Instead, it developed TopsRider, a full-stack AI computing and programming platform covering drivers, programming languages and compilers, operator libraries, and toolchains.

The platform is mainly used to connect Enflame’s in-house hardware with mainstream AI models, lowering model migration and development costs so hardware performance can be more fully realized in real-world applications.

On that foundation, the company has extended its software and hardware capabilities upward into AI computing systems and clusters. During the reporting period, it generated revenue from AI data center projects at the thousand-card and ten-thousand-card scale, and began working with customers on larger-scale, commercially valuable high-speed interconnect cluster solutions.

Domestic GPUs Take Turns Entering the Capital Markets

Overall, China’s cloud AI chip industry remains in an early stage of development.

NVIDIA still dominates the China market, but local vendors are gradually breaking through technical and scale barriers, and the industry structure is beginning to loosen.

By technical route, Chinese cloud AI chip vendors can be roughly divided into two groups:

One group consists of non-GPGPU architecture companies represented by Huawei HiSilicon, Cambricon, and Enflame Technology. The other consists of GPGPU architecture companies represented by Moore Threads, MetaX, Iluvatar CoreX, and Biren Technology.

According to IDC data, China’s overall shipments of AI accelerator cards exceeded 2.7 million units in 2024, of which NVIDIA shipped about 1.9 million, giving it a market share of roughly 70%.

In the same period, Enflame sold about 38,800 AI accelerator cards and modules, corresponding to around 1.4% of the domestic market and placing it among the leading domestic AI chip vendors.

Against this backdrop, Moore Threads, MetaX, Biren Technology, and Enflame Technology are often referred to together as China’s “four little dragons” of domestic GPUs.

On the timeline, Moore Threads was founded in June 2020 and listed on the STAR Market on December 5, 2025, becoming the first domestic GPU stock in China;

MetaX was founded in September 2020 and listed on the STAR Market on December 17, 2025;

Biren Technology was founded in September 2019 and listed on the Hong Kong Stock Exchange on January 2, 2026, becoming the first domestic GPU stock in Hong Kong.

As the earliest-founded of the four, Enflame has finally seen its A-share STAR Market IPO accepted at the start of 2026.

Although all four are described as GPU companies, their positioning differs, with each focused on a different emphasis.

Moore Threads focuses on full-function GPUs, benchmarking NVIDIA’s strategy of using one architecture to serve multiple scenarios. A single chip integrates capabilities including AI computing, graphics rendering, physics simulation, scientific computing, and ultra-high-definition encoding and decoding.

Biren Technology focuses on high-end cloud GPGPUs with large-scale computing power, mainly serving core scenarios such as cloud data centers, telecom operators, and AI data centers, supporting general-purpose computing workloads including large-model training, AI inference, and high-performance scientific computing.

MetaX benchmarks AMD, with AI training and inference and general-purpose computing at its core, while also covering graphics rendering. Enflame Technology focuses on cloud AI training and inference, with particular emphasis on the AI inference market.

Looking back over just the past few months, domestic GPU companies have been rushing into the capital markets: Moore Threads and MetaX listed on A-shares one after another, while Biren listed in Hong Kong, and all saw their share prices strengthen after listing.

At the same time, domestic chips backed by major tech companies, including Baidu’s Kunlunxin and Alibaba’s T-Head, have repeatedly been rumored to be advancing IPO plans.

At its core, this wave of capital-market activity among domestic GPU companies is the result of overlapping forces: technological progress, the policy environment, and market demand.

The listing wave gives companies more ammunition to keep investing in R&D and ecosystem building, while also signaling that the industry is moving from early exploration toward a more standardized and sustainable stage of development.

Next, as leading vendors and major tech-backed players move forward in tandem, the competitive focus for domestic chips will gradually narrow to architecture capability and software ecosystems, laying a firmer foundation for independent AI computing power.