TellWorldAI World Archive

Home›Companies

OrganizationCompany

Biren Technology

Chinese chip designer founded in 2019 that builds general-purpose GPU (GPGPU) chips and intelligent computing solutions on its self-developed BILI series hardware and BIRENSUPA software, for AI training and inference from cloud to edge.

Founded 2019Website birentech.com Era Large models

What Biren does

Biren Technology develops general-purpose computing on graphics processing units — GPGPU — chips, along with intelligent computing solutions built on top of them. The hardware side is its self-developed BILI series of chips; the software side is a platform called BIRENSUPA[1].

GPGPU is a way of describing what the chip is for. A graphics processor is originally designed to draw images, and its architecture is good at doing many simple calculations at the same time. Point that architecture at general numerical work instead of pictures, and you have a GPGPU. Most of the arithmetic inside a neural network is matrix multiplication, which is exactly the kind of work that suits many small calculations running in parallel. A software platform like BIRENSUPA sits between the chip and the programmer: it is what lets developers put the hardware to work without writing every routine from the ground up.

Origins

Biren was founded in 2019 and focuses on GPGPU chips and intelligent computing solutions for AI training and inference, from cloud to edge[2]. Training is the process of adjusting a model's parameters against data; inference is using the finished model to handle real requests. "Cloud to edge" describes where the work happens — in data-centre servers, and in equipment closer to where data is produced.

Its BR100 chip was announced in August 2022, roughly three years after the company was founded[3].

Who it sells to

Biren names internet companies, AI data centres, foundation model developers, telecoms, energy and utilities, and financial technology as its key customer segments[1]. These are the buyers that run large AI workloads or need computing capacity on their own premises — internet platforms and model developers for training, and sectors such as telecoms, utilities and finance for inference on their own data.

According to its prospectus, the company's intelligent computing solutions began generating revenue in 2023. It had 14 customers in 2024 and 12 in the first half of 2025, contributing 337 million yuan and 58.9 million yuan respectively[2].

The same prospectus gives a picture of the order book. As of the latest practicable date before its listing, Biren had 24 outstanding binding orders for specialised technology products, worth about 822 million yuan[2]. It had also signed five framework sales agreements and 24 sales contracts, worth about 1.241 billion yuan in total[2]. The distinction matters for reading the numbers: framework agreements set out terms for future purchases, while binding orders are commitments already placed.

Research and people

Biren's investor relations page says the company had filed more than 1,900 patent applications and been granted more than 1,000 patents[1]. The same page says more than 80% of its staff work in research and development[1]. Chip design is the kind of work where that ratio is not unusual: the product is the design, and turning a design into a working chip for AI workloads takes long cycles of architecture work, verification and software support.

Biren has won the SAIL Award at the World Artificial Intelligence Conference twice, according to its investor relations page[1].

Where it sits

The numbers describe a company that has moved from chip development into early commercial sales. Revenue begins in 2023, and the customer count — 14 in 2024, 12 in the first half of 2025 — shows sales concentrated among a modest number of buyers, which is typical when the product is expensive computing hardware sold to organisations that run large workloads. The order book is the figure to watch: binding orders of about 822 million yuan and contract value of about 1.241 billion yuan are what stand behind future revenue[2].

Biren is part of a group of companies building general-purpose GPU accelerators aimed squarely at AI computing, rather than at graphics. That market is shaped by two things at once: how fast the software ecosystem around a chip matures, and whether customers can get enough of the hardware to plan around it. Biren's approach addresses both ends — BILI chips as the hardware, BIRENSUPA as the layer that makes them usable[1] — with the company's own stated split of effort, more than 80% of staff in R&D, indicating where it is still investing[1].

TechniquesDeep learning

ApplicationsEnterprise

Ecosystem & regionChinaHardware & compute

20261

  1. MilestoneBiren lists on the Hong Kong Stock Exchange

    Biren Technology listed on the Hong Kong Stock Exchange's main board on 2 January 2026, drawing 23 cornerstone investors to its share offering, with its market value briefly passing HK$100 billion on the first trading day.

20231

  1. PolicyUS tightens advanced chip export controls on China

    On 17 October 2023 the US Bureau of Industry and Security issued three rules updating its October 2022 chip export controls, changing the chip parameters, adding licence exceptions and requirements, and expanding the Entity List.

Models and products over time

1
  1. 2022
    • Aug 2022 Biren BR100 Chip

      Biren's general-purpose GPU, released in August 2022, packages two dies on a 7 nm process with 77 billion transistors; the company claimed PFLOPS-level single-chip compute, a first for a Chinese firm.

Notes and sources

  1. Investor Relations - Biren Technologyhttps://ir.birentech.com/
  2. "国产GPU四小龙"之一的壁仞科技启动招股https://news.chengdu.cn/2025/1224/694ba1c381c58e5630940cbd.shtml
  3. 详解壁仞刚刚发布的GPU:单芯片PFLOPS算力是怎样炼成的?https://ex.chinadaily.com.cn/exchange/partners/82/rss/channel/cn/columns/snl9a7/stories/WS62fb45eca3101c3ee7ae3ff6.html

Cite this page: Biren Technology. TellWorld, https://www.tellworld.net/companies/biren. Updated 2026-11-08