Biren Revenue Jumps 2,000% as U.S. Controls Fuel Chinese AI Chip Surge
Biren posted 1,998% revenue growth in H1 2026 after Nvidia's exit from China amid U.S. export controls. We examine the drivers and challenges.
Chinese AI accelerator vendor Biren Technology posted revenue of $183.9 million for the first half of 2026, up 1,998% year over year. According to reporting by Anton Shilov of Tom’s Hardware, the surge was directly triggered by U.S. export controls that effectively forced Nvidia and AMD to withdraw from the Chinese market. The figures are based on data compiled by research firm Jon Peddie Research and show that demand for domestically produced AI processors rose sharply from the second half of 2025. However, the growth comes from an extremely small base, and in absolute terms the company remains a small supplier.
Biren reported first-half revenue of $183.9 million, up 1,998% year-over-year from around $8.665 million in the first half of 2025.
The company reported gross profit of $78.552 million for a gross margin of 42.7%, while posting a net loss of $56.2 million. The loss is attributed mainly to continued investment in next-generation AI accelerators, rack-scale solutions using optical interconnects, and software.
Market Gap Created by U.S.
Export Controls and Its Impact
The turning point in U.S. export controls on China was the measures imposed in May 2025 under the Trump administration. According to reporting by Anton Shilov of Tom’s Hardware, the halt in supplies of AI GPUs for China by U.S. companies including Nvidia left a large void in the domestic market. Before the tightening, even Nvidia’s performance-limited H20 outperformed Chinese alternatives in performance and software environment, so demand for domestic accelerators was limited.
According to TrendForce, Nvidia supplied about 2.2 million AI accelerators to China in the first half of 2025 alone, before the controls took effect. The bulk was the H20, sold at $12,000 to $15,000 per unit, dominating the market in both volume and value. By contrast, Biren’s shipments for the full year 2025 are estimated at only several thousand to tens of thousands of units. After the controls, this supply structure changed completely, with domestic vendors clearly capturing replacement demand.
Biren’s sales began to grow in the second half of 2025. Full-year revenue for 2025 was $154.17 million, representing less than 3% of the domestic AI accelerator market. The sharp increase in the first half of 2026 reflects a rebound from that low level combined with a shift by customers forced to seek alternative procurement. The growth has been driven by a substitution of demand that would not have occurred without the controls.
Unpacking the 1,998% Increase and Profit Structure
While the 1,998% growth rate appears to show massive business expansion at first glance, it should be noted that first-half 2025 revenue, the base for comparison, was extremely small at about $8.665 million. The $183.9 million recorded in the first half of 2026 is about 21 times that figure. The increase is about $175 million in absolute terms, still small compared with the sales Nvidia had generated in China.
In terms of profit structure, a gross margin of 42.7% shows the company has secured a reasonable level for a semiconductor product. Against gross profit of $78.552 million, R&D and commercialization costs weighed heavily, resulting in a net loss of $56.2 million. According to Jon Peddie Research, the loss stems less from one-off expenses than from upfront investment in next-generation accelerators, rack-scale infrastructure, and the software stack.
This investment stance can be interpreted as a strategy prioritizing market share over short-term profitability. With some viewing the gap created by export controls as not permanent, failure to put products and an ecosystem in place within a limited time could expose the company to renewed competition if controls are eased or alternative supply chains are rebuilt. The decision to continue investing despite losses appears to reflect that time constraint.
Product Lineup and Software Strategy in Full
Biren has developed at least three high-performance GPUs for AI: the BR106, BR110, and BR166. According to Jon Peddie Research, the company is now developing next-generation accelerators including the BR20X, BR30X, and BR31X. On paper, their specifications are said to be competitive with products from AMD and Nvidia, but their actual performance, supply stability, and maturity in operating environments still require verification.
The company’s proprietary Birensupa software stack is intended to rival Nvidia’s CUDA. In the Chinese market, even performance-limited Nvidia products had been chosen because of the advantage of the CUDA-centered development environment. Biren’s move to offer hardware and software together, combined with rack-scale solutions, aims to break down this dependence.
Software maturity takes more time than shipment volumes. Compatibility and migration costs when moving existing trained models and operating infrastructure to a domestic stack are barriers. The inclusion of software among the areas where Biren continues to invest appears to reflect the reality that accelerator performance alone cannot drive customers to switch. Even if domestic customers adopt its products temporarily in response to the controls, improving the developer experience will be essential to secure continued use.
Competitive Landscape and Challenges in
China’s AI Accelerator Market
China’s AI accelerator market was structured around overwhelming supply volumes from Nvidia through the first half of 2025. Compared with the 2.2 million units indicated by TrendForce figures, shipments by domestic vendors including Biren are orders of magnitude smaller. Biren’s shipments in the first half of 2026 also fall far short of the volumes Nvidia supplied in 2025. Viewed across the market as a whole, substitution with domestic products is still in its early stages.
The competitive environment is not limited to Biren. Domestic semiconductor companies including Huawei are also developing AI accelerators, and adoption of domestic products could advance centered on government procurement and state-owned enterprises. On the other hand, private companies and research institutions tend to place strong emphasis on performance and software maturity, and some point out that use of Nvidia products through unofficial distribution channels or inventories has not completely disappeared even under the controls. Reporting by Anton Shilov of Tom’s Hardware describes competition as having officially disappeared, while suggesting the actual situation remains opaque.
Supply-side constraints are also a challenge. Manufacturing of advanced semiconductors is highly dependent on external foundries, with manufacturing capacity, yields, and component procurement setting the ceiling for business expansion. Biren’s investment in rack-scale solutions using optical interconnects is seen as reflecting an intent to differentiate not only through single-chip performance competition but also through interconnect efficiency and operational efficiency at data-center scale. Rather than products alone, rack-level offerings could determine future order competition.
Future Focus and the Outlook for Sustainability
The key question for Biren’s future is whether the surge in sales will remain a temporary capture of demand or can be converted into a sustainable business foundation. Growth in the first half of 2026 originated from the supply disruption caused by the controls and will be heavily influenced by the regulatory environment and trends in alternative supply. If the controls continue, conditions will remain favorable for domestic vendors, but customers will demand long-term supply stability and a technology roadmap.
Whether the BR20X, BR30X, and BR31X under development can move to mass production as scheduled will also be an important milestone. Gaps can easily arise between prototype specifications and the performance, power efficiency, and supply volumes of mass-produced products. In addition, the extent to which the software stack, including Birensupa, can ensure compatibility with existing AI frameworks and models will determine whether adoption expands. Alongside increased hardware shipments, documentation, tools, and migration support for developers will need to be put in place.
On the financial front, the question is whether the company can narrow its losses while maintaining its 42.7% gross margin. The $56.2 million loss is the flip side of growth investment, but continued reliance on external financing or government support will make it difficult to establish an assessment of the business’s independence. How far it can raise its market share from less than 3% for full-year 2025 will be a prerequisite for recouping investment. Quarterly trends from the second half of 2026 onward are expected to provide material for judging the sustainability of growth.
Editorial Opinion
In the short term, we see the tailwind that U.S. export controls provide to domestic vendors continuing through the second half of 2026. With official supplies from Nvidia and AMD cut off, companies like Biren can more easily accumulate orders for replacement procurement from government-related entities and major cloud operators. However, because customers migrate in stages after performance verification, we assess that order growth will proceed gradually. Moves to fill the regulatory void could intensify order competition among domestic vendors.
In the long term, we believe maturity of the software ecosystem, rather than hardware performance competition, will determine market structure. Migration costs for CUDA-dependent development assets are high, and unless the domestic stack demonstrates compatibility and an operational track record, there is a concern that substitution will remain temporary. Biren’s investment in rack-scale offerings and optical interconnects can be evaluated as looking ahead to a stage when products are chosen for operational efficiency rather than single-chip performance itself. How manufacturing supply-chain constraints are overcome will also determine sustainability over the next one to three years.
The remaining question is whether this growth depends on the external factor of controls or results from self-sustaining technological competitiveness.
References
- “ China’s AI accelerator supplier Biren posts 2,000% year-over-year revenue growth — US export controls benefit homegrown chips as Nvidia and AMD exit market “, by Anton Shilov — Tom’s Hardware, 2026-09-10T12:40:00.000Z (ARR)
- Source URL: https://www.tomshardware.com/tech-industry/artificial-intelligence/chinas-ai-accelerator-supplier-biren-posts-2-000-percent-year-over-year-revenue-growth-export-controls-benefit-homegrown-chips-as-nvidia-and-amd-exit-market
Frequently Asked Questions
- What were Biren Technology's results for the first half of 2026?
- Revenue was $183.9 million, up 1,998% year over year, with gross profit of $78.552 million and a gross margin of 42.7%. Meanwhile, the company posted a net loss of $56.2 million due to investment in next-generation products and software. Because first-half 2025 revenue was small at about $8.665 million, the growth rate reflects the small base.
- Why did Biren's sales surge?
- The cause was the U.S. export controls on China imposed in May 2025, after which Nvidia and AMD halted supplies of AI GPUs for China, leaving a void in the market. Replacement demand for the scale Nvidia had supplied — about 2.2 million units in the first half of 2025 — flowed to domestic vendors from the second half of 2025, expanding Biren's orders.
- What are Biren's products and future development plans?
- It offers three high-performance GPUs for AI — the BR106, BR110, and BR166 — and is now developing the BR20X, BR30X, and BR31X. It also continues to invest in its proprietary Birensupa software stack to rival Nvidia's CUDA and in rack-scale solutions using optical interconnects.
Comments