TellWorldAI World Archive

Home›Companies

OrganizationCompany

Hygon Information Technology

A Chinese chip designer founded in 2014 that builds x86-compatible processors and DCU accelerators, listed on the STAR Market in 2022 and positioning itself around a self-developed C86 plus GPGPU product line.

Founded 2014Website hygon.cn Era Deep learning

What Hygon is

Hygon Information Technology is a Chinese chip company founded in 2014 that develops high-end processors and accelerators. It describes its goal as becoming a world-class chip company [1]. Its processors are compatible with the mainstream x86 instruction set, which Hygon says gives them a mature application ecosystem [1]. The practical meaning of that compatibility is straightforward: software already written and compiled for x86 machines can run on these chips without being rewritten for a different instruction set, which matters to customers with years of code invested in that architecture.

Hygon describes itself as holding China's only self-developed C86 plus GPGPU product matrix in commercial use [3]. In plain terms, the company sells two kinds of silicon: general-purpose processors, and accelerators built for the kind of massively parallel arithmetic that graphics and AI workloads need.

Origins and listing

The company was founded in 2014 [1] and spent its early years developing processors and accelerators for commercial customers. On 12 August 2022 it listed on the STAR Market, Shanghai's board for science and technology companies, under the code 688041. The issue price was 36 yuan per share, and the offering raised 10.6 billion yuan [2].

Products and ecosystem

Hygon's business rests on the pairing of processors and accelerators: the processor runs the general-purpose workload, while the accelerator, sold under the DCU name, takes on the parallel part. Building both is unusual, and the company presents the combination as its distinguishing feature [3].

A chip line is only as useful as the hardware and software built around it, and Hygon has put effort into that layer. In 2025 it opened its CPU interconnect bus protocol, HSL, to the industry, so that partners working on xPU, IO, operating systems and original equipment manufacturing could connect to it [4]. An interconnect protocol becomes valuable only when other companies build parts that speak it, so opening it is a way of recruiting those companies. The same logic sits behind the Guanghe industry alliance, which Hygon founded to build a domestic hardware and software ecosystem and which had gathered more than 6,000 partners [4].

Adoption of the accelerators has spread across customer industries. By the end of 2025, Hygon said its DCU products had been deployed in more than 20 industries and over 300 application scenarios [4]. Demand was strong enough that the production-to-sales ratio for high-end processors reached 98.84% in 2025, up 23.49 percentage points year on year — a measure of how closely output kept pace with what was sold [5].

Research and money

Hygon is a research-heavy company. It spent 4.569 billion yuan on research and development in 2025, equal to 31.78% of revenue, and its 2,766 R&D staff made up 82.99% of employees [3]. As of 31 December 2025 it held 1,101 invention patents, 127 utility model patents and 358 integrated circuit layout design certificates [3].

The financial record has grown quickly. Revenue for 2025 was 14.377 billion yuan, up 56.92% year on year, with net profit attributable to shareholders of 2.545 billion yuan, up 31.79% [3][4]. Growth carried into the following reporting period: first-quarter 2026 revenue was 4.034 billion yuan, up 68.06%, and net profit was 687 million yuan, up 35.82% [3][4]. In both periods profit grew more slowly than revenue.

Where it sits

Hygon occupies an unusual position among chip designers. Most processor companies either build for the dominant x86 instruction set or pursue a separate architecture with its own software stack; Hygon does the former while presenting the result as a self-developed product line, C86, and pairing it with in-house accelerators [3]. For buyers, the appeal is continuity — existing software keeps working — combined with a supplier that also offers the accelerator half of a machine.

The ecosystem work is the part that takes time. Opening a bus protocol, running an alliance of thousands of partners and accumulating patents are all attempts to make sure that hardware built by others fits Hygon's chips, and that software runs on them without modification [4][3]. Whether that bet pays off depends on how many partners actually ship products around the protocol, and on whether the accelerator line keeps finding new application scenarios beyond the 300-odd it had reached by the end of 2025 [4]. What the numbers show so far is scale: tens of billions of yuan in annual revenue, a research budget near a third of it, and a product mix that few Chinese chip companies attempt.

TechniquesDeep learning

ApplicationsEnterprise

Ecosystem & regionChinaHardware & compute

Models and products over time

1
  1. 2025
    • 2025 Hygon DCU Shengsuan No. 3 Chip

      Hygon's Shengsuan No. 3 is a GPGPU-architecture deep computing processor that reached the market in 2025, built on a 7 nm chiplet design with 1.8 TB/s of HBM bandwidth and a CUDA-like software stack.

Notes and sources

  1. 关于海光https://www.hygon.cn/about
  2. 海光信息https://zh.wikipedia.com/zh-tw/%E6%B5%B7%E5%85%89%E4%BF%A1%E6%81%AF
  3. 沪市首份一季报出炉 海光信息营收增长68%https://www.cnstock.com/commonDetail/666467
  4. AI算力需求爆发 海光信息Q1营收同比增68% DCU全面适配主流大模型https://www.cls.cn/detail/2337024
  5. 海光信息(688041):DCU加速出货 "双芯"构筑差异化竞争优势https://m.gelonghui.com/news/5207809

Cite this page: Hygon Information Technology. TellWorld, https://www.tellworld.net/companies/hygon. Updated 2026-11-08