Why Biren Technology’s Revenue Jumped Nearly 2,000% in H1 2026

Biren Technology’s nearly 2,000% revenue increase in the first half of 2026 was both a genuine commercial step forward and a reminder of how percentage growth can exaggerate a young company’s scale. The Shanghai GPU developer reported RMB1.236 billion in revenue for the six months ended June 30, up 1,997.6% year over year, as shipments of its Bili GPGPU solutions and AI-computing clusters accelerated. 23

The key context: its H1 2025 revenue was approximately RMB58.9 million. Biren has moved well beyond that small base, but the headline growth rate alone does not establish parity with Nvidia or prove that China’s domestic GPU industry has solved its longer-term technology and supply constraints. 811

What drove the surge

Intelligent-computing solutions accounted for roughly RMB1.17 billion of Biren’s H1 revenue, according to reported interim-results coverage. That sales mix was linked to large-scale deliveries of Bili-series GPGPU products and AI-computing clusters. 2

This matters because customers deploying AI infrastructure generally buy more than a processor. They need deployable systems for training and inference, including servers, networking, software, integration and support. Biren describes its offering as GPGPU-based intelligent-computing solutions that combine hardware with its BIRENSUPA computing software platform. 1

Demand for integrated AI-computing capacity has given domestic suppliers a clearer commercial opportunity as access to leading foreign accelerators has become uncertain. Biren’s reported growth therefore appears to reflect actual systems shipments, not merely a financial revaluation or a one-time accounting change. 23

The low base is central to interpreting 1,997.6% growth

A 1,997.6% year-over-year increase means revenue was about 21 times the prior-year level. But the starting point was exceptionally small: RMB58.9 million in H1 2025 versus RMB1.236 billion in H1 2026. 811

That does not diminish the achievement. Reaching more than RMB1.2 billion in six-month revenue signals that Biren has converted a portion of domestic AI-compute demand into sizable deliveries. It does mean, however, that the percentage should not be read as evidence that the company has reached the revenue scale, installed base or ecosystem depth of the global GPU leaders.

A more useful reading is that Biren has made the transition from a comparatively small revenue base to meaningful commercial deployment. The next question is whether it can repeat that deployment at attractive margins while expanding its customer and software base.

Margins improved, but Biren remains loss-making

Biren’s gross profit rose to about RMB527 million, while gross margin increased by 10.8 percentage points to 42.7% in H1 2026. Reports attributed the improvement primarily to growth in cloud-training products. 4512

Those numbers are encouraging for product economics. Higher-margin cloud-training revenue and greater sales volume can improve the economics of a systems business, and Biren’s reduced losses suggest that operating leverage is beginning to emerge.

Still, the company was not profitable. Biren reported a loss for the period of RMB377 million, down 76.4% year over year, and an adjusted loss of RMB337 million. R&D expenses were RMB804 million, up 40.7% from a year earlier. 345

That gap captures the central challenge for domestic AI-chip companies: selling hardware and systems at scale is necessary, but ongoing investment in chips, interconnects and software can remain larger than operating profit for a long time. A narrower loss is progress, not proof that the business has reached self-sustaining profitability.

Software is part of the product

The GPU market is not decided by chip specifications alone. AI developers and cloud operators need compilers, libraries, frameworks, deployment tools, cluster management and technical support. The cost and risk of moving existing workloads are often as important as the accelerator’s raw performance.

Biren’s reported cluster and cloud-training sales show traction for an integrated offering, while its BIRENSUPA platform indicates that software is a defined part of its product strategy. 12 However, the available evidence does not establish software parity with Nvidia’s CUDA ecosystem. That distinction is important: shipping systems can generate revenue quickly, while building a deeply adopted developer ecosystem is a longer process.

For customers, the practical test will be whether workloads can be deployed, optimized and maintained with acceptable performance and engineering effort. For Biren, software maturity will help determine whether early demand becomes recurring platform adoption.

Export controls create demand, but not a guaranteed captive market

Restrictions and uncertainty around access to advanced foreign AI chips have increased the strategic appeal of domestic alternatives for Chinese organizations. At the same time, the market is not simply closed to outside computing options. U.S. authorities have acted to close a potential route for exporting advanced chips to overseas subsidiaries of Chinese companies, illustrating that policy and enforcement conditions continue to evolve.

The supplied evidence does not quantify illicit physical imports of restricted Nvidia processors or show their scale relative to legal supply, overseas compute access or domestic alternatives. It would therefore be inaccurate to treat an illicit-import channel as a measured explanation for Biren’s revenue or as a reliable estimate of competitive supply.

The broader implication is straightforward: policy can enlarge the addressable market for local GPU suppliers, but policy conditions can also shift. A durable business cannot rely solely on restricted foreign supply; it must win on usable products, service and delivery.

The unresolved constraint: dependable production at scale

Biren’s H1 results demonstrate demand and stronger commercialization. They do not, by themselves, demonstrate assured access to all of the manufacturing inputs needed to scale high-end AI accelerators over time.

The relevant challenge spans fabrication, packaging, memory and system integration. These constraints can affect cost, delivery schedules and product design across China’s AI-chip sector. The evidence provided does not support a definitive company-by-company comparison of manufacturing capacity among Biren, Huawei, Cambricon and Kunlunxin, so claims that one has a clear supply advantage over the others should be treated cautiously.

What Biren’s results reveal about China’s AI-chip opportunity

Biren’s first-half performance shows that domestic AI-chip demand can translate into substantial revenue when a supplier can deliver usable GPGPU systems and clusters. Its margin expansion and sharply narrower loss suggest better commercial execution than the prior year. 34

But the figures also show why a revenue surge is only the beginning of the story. Biren still needs to turn improved gross margins into sustained profitability while continuing to fund R&D. It must deepen its software platform, serve customers reliably at scale and navigate an evolving supply and policy environment.

In short, H1 2026 is evidence of a meaningful market opening for China’s domestic AI-chip industry—not conclusive evidence that the industry has closed the technology, ecosystem or manufacturing gap with established global leaders.