Biren Technology

Biren Technology is a Chinese fabless semiconductor company focused on general-purpose GPU compute for AI and high-performance computing. Founded in 2019 and headquartered in Shanghai, it was started by executives with backgrounds in firms such as SenseTime, Qualcomm, and Huawei. Biren became a publicly listed company in Hong Kong in early 2026.

‍

The company’s early flagship products include the BR100 and BR104 accelerator GPUs, introduced in 2022 for data center AI and compute workloads. Biren pairs its hardware with a proprietary software platform, BirenSUPA, to support AI model training and inference across deployments from cloud to edge.

‍

Biren has operated under tighter external constraints since October 2023, when the U.S. Commerce Department added Biren and related entities to the Entity List. On January 2, 2026, its H shares began trading on the Hong Kong Stock Exchange under stock code 06082, marking a shift from private funding toward public-market disclosure and capital access.

Biren Technology is a China-based, fabless chip designer focused on general-purpose GPU compute for AI and high-performance workloads. The company listed in Hong Kong under stock code 6082, with trading starting on January 2, 2026.

‍

How Biren makes money

‍

1. Integrated intelligent computing solutions
   Biren’s core commercial product is sold as a bundle that combines GPGPU-based hardware systems with BIRENSUPA, its computing software platform.
   ‍
2. Hardware-led deployments that scale into clusters
   The company sells GPU hardware into enterprise and data center projects, including large cluster deliveries, then supports deployment with localized, on-the-ground customer support.
   ‍
3. Framework agreements and contract-based deliveries
   Sales are driven by signed framework agreements and binding sales contracts, which convert into revenue when deliveries are realized. As of the latest practicable date in the prospectus, Biren disclosed 24 unfulfilled binding orders valued at about RMB 821.8 million, plus five framework sales agreements and 24 sales contracts totaling about RMB 1,240.7 million.
   ‍
4. Other revenue lines
   The prospectus also describes “agent fee” revenue from selling servers, other hardware devices, and software on an agent basis.
   ‍

Market position in China

‍

• Commercialization is recent, with a fast revenue ramp. Biren states it started generating revenue from intelligent computing solutions in 2023. Revenue rose from RMB 62.0 million (2023) to RMB 336.8 million (2024). In H1 2025, revenue was RMB 58.9 million.
  ‍
• Customer count is still small, with larger-ticket buyers in 2024. Biren reported 14 customers in 2024 and 12 customers in H1 2025 for its Specialist Technology Product, and it notes 2024 customers skewed toward “leading players.”
  ‍
• Losses remain large due to R&D intensity. The prospectus reports the loss for the period increased to RMB 1,600.5 million in H1 2025.
  ‍
• Geopolitics is a structural constraint. Biren-related entities were added to the U.S. Entity List in October 2023, tightening access to U.S.-origin technology and suppliers.
  ‍
• Post-IPO funding supports longer runway. Reuters reported Biren raised about HK$5.58 billion (about $717 million) in its IPO at an offer price of HK$19.60 per share.

China is Biren’s operating and sales base, with a go-to-market built around large domestic buyers that are scaling compute. In its listing prospectus, Biren names target verticals that include AI data centers, telecommunications, AI solutions, energy and utilities, financial technology, and internet companies, backed by local delivery and on-the-ground customer support.

‍

Commercial traction is still concentrated but measurable. Biren states it began generating revenue from its integrated intelligent computing solutions in 2023. In 2024 it recorded RMB 336.8 million of revenue from 14 customers, and in H1 2025 it recorded RMB 58.9 million from 12 customers. It also disclosed 24 unfulfilled binding orders worth about RMB 821.8 million, plus framework agreements and sales contracts totaling about RMB 1,240.7 million.

‍

Biren also says it has delivered solutions to nine Fortune China 500 companies, including five that are in the Fortune Global 500, with repeat purchases from key customers.

Biren’s growth path is built around China’s demand for domestic AI compute and its ability to ship a complete “intelligent computing solution” that combines GPGPU hardware with its BIRENSUPA software stack. The company raised HK$5.58 billion in its Hong Kong IPO, and shares closed up 76% on the first trading day (January 2, 2026).

‍

The prospectus frames IPO proceeds as fuel for a multi-year buildout. It estimates HK$4,350.6 million of net proceeds at the midpoint price and allocates 85% to R&D, split into 45% for hardware evolution and 40% for software platform development and upgrades, plus 5% for commercialization and 10% for working capital. Hardware spend includes next-generation GPGPU chips BR20X and BR30X, plus broader form-factor support across PCIe cards and OAM. Software spend targets drivers, compilers, libraries, tooling, compatibility with mainstream AI frameworks, and Biren’s inference acceleration framework suInfer.

‍

Key growth drivers include

‍

1. Domestic substitution demand and capital access
   Hong Kong’s renewed IPO window for China AI and chip names supports funding for long R&D cycles, while export restrictions keep domestic alternatives in focus for buyers.
   ‍
2. Scaling from chips to repeatable solution deployments
   Biren sells integrated systems and positions software as the adoption layer, which supports repeat purchases after initial deployments.
   ‍
3. Software platform maturity as the adoption gate
   The proceeds plan puts almost as much weight on software as on hardware, with explicit resourcing for model support, framework compatibility, and internal test and release infrastructure.
   ‍

Challenges ahead

‍

• Export controls and supply chain constraints tied to Biren’s Entity List designation (October 2023).
  ‍
• High burn from R&D intensity, with R&D running as the bulk of operating expenses across recent periods, alongside large reported losses.
  ‍
• Ecosystem gap versus CUDA-centered workflows, which raises switching costs for large clusters and slows standardization decisions.
  ‍

What to watch through 2026: conversion of signed orders into revenue, visible cluster wins with repeat buyers, and concrete progress on the BR20X/BR30X roadmap plus software releases that improve training and inference stability at scale.