China’s Biren Technology triples revenue as domestic AI chip pivot accelerates

Shanghai Biren Technology tripled annual revenue to 400 million yuan ($57 million) in 2025, confirming that China’s domestic AI chip market has crossed from policy aspiration to commercial reality.

The GPU maker’s January 2 Hong Kong IPO opened at HK$35.7 per share—82% above its HK$19.6 issue price—valuing the company at HK$85.54 billion ($10.9 billion), according to PR Newswire. The debut caps a period of explosive growth: Biren generated 336.8 million yuan in 2024 and 58.9 million yuan in the first half of 2025 alone, with second-half acceleration pushing the full-year total past 400 million yuan, per Tom’s Hardware.

More telling than historical revenue is forward demand: Biren reported 2.1 billion yuan ($298 million) in pending and contracted orders at the time of its IPO filing. That backlog, equivalent to more than five times 2025 revenue, reflects China’s state-mandated procurement shift requiring data centers to source over 50% of chips domestically by 2025.

U.S. export controls created the market

Biren’s rise is inseparable from Washington’s semiconductor restrictions. The U.S. placed advanced AI chips under Export Controls in October 2022, then added Biren to the Entity List in October 2023—barring it from accessing Nvidia’s supply chain or fabrication partners using U.S. equipment. Rather than halting China’s AI ambitions, the controls artificially manufactured demand for domestic alternatives by forcing Chinese tech giants to abandon Nvidia dependency.

China now targets 76% AI GPU self-sufficiency by 2030, up from 33% in 2024, according to Digitimes. Thirteen top Chinese semiconductor executives formalized an 80% overall chip self-sufficiency goal by 2030 in a March 27 joint statement, Nikkei Asia reported. The targets carry weight: Beijing backed the pledge with production capacity expansions at SMIC (doubling 7nm output) and three Huawei fabrication plants coming online in late 2025 and 2026.

Performance gaps narrow but remain material

Biren’s BR100 GPU delivers 2,000 teraflops of BF16 compute—trailing Nvidia’s H100 (3,000 teraflops FP16) but competitive with the older A100 generation. The company claims 2.6x speedup versus A100 on certain inference workloads, per ARC Advisory Group. That positions Biren as viable for inference tasks—the compute-intensive process of running trained AI models—but still behind the frontier for training cutting-edge large language models.

The performance deficit matters less than market dynamics suggest. Chinese enterprises care more about supply security and regulatory compliance than benchmark leadership. A ‘good-enough’ domestic chip guaranteed to clear export control hurdles beats a superior foreign chip subject to overnight restriction.

Nvidia and AMD together held 71% of China’s AI chip market in 2024. Huawei captured 23%, while Biren, Moore Threads, and MetaX—collectively dubbed China’s ‘four dragons’—split less than 1%. But Techstrong.ai cited analysts expecting Huawei alone to claim 50% market share by year-end 2026, with the remaining domestic players splitting another 20-30 points. That would reduce Nvidia to low-teens share in the world’s second-largest AI market within 24 months.

Capital flows validate the thesis

Biren secured $372.5 million from cornerstone investors including Qiming Venture Partners, Ping An Group, UBS, and Singapore’s Lion Global Investors ahead of its Hong Kong debut. The IPO priced at the top of its range despite Biren posting consecutive annual losses—investors underwrote the story, not the earnings.

Market reception post-IPO tells a more complex story. Biren’s stock traded at $4.26 by March 18, well below its opening pop, reflecting broader volatility in Chinese tech equities and uncertainty over execution timelines. But the retreat from initial euphoria doesn’t invalidate the structural demand thesis: 2.1 billion yuan in contracted orders remain on the books regardless of secondary market sentiment.

“Going forward, we will continue to increase R&D investment, accelerate the development of our fully independent and controllable product stack, and help build a robust, domestic computing ecosystem,” Zhang Wen, Biren’s founder and CEO, said in the IPO announcement.

Geopolitical overhang intensifies

Biren’s success poses a direct challenge to U.S. semiconductor policy. Washington’s export controls aimed to restrict China’s access to frontier AI compute, buying time for the U.S. to maintain a technological lead. Instead, the controls created a $100+ billion protected market for domestic Chinese alternatives, subsidised by state procurement mandates and insulated from foreign competition.

The timeline to competitive parity remains uncertain. Congressional Research Service analysis of export controls notes that China still lags in advanced lithography, memory bandwidth, and power efficiency—gaps that compound at scale. But China compensates with advantages the U.S. cannot easily replicate: lower engineering costs, patient state capital, and domestic electricity supply growing fast enough to support mass AI deployment. Elon Musk observed in January that “China’s growth in electricity is tremendous,” positioning the country to absorb chip output that would sit idle elsewhere.

What to watch

Biren’s ability to convert its 2.1 billion yuan backlog into recognised revenue over the next 12-18 months will validate whether demand is genuine or inflated by one-time state procurement cycles. Monitor Huawei’s market share velocity: if it reaches 50% by late 2026 as analysts expect, Nvidia’s China position becomes structurally untenable regardless of diplomatic thaw.

Second-order effects warrant attention. China tripling AI chip output by 2026 will create surplus capacity beyond domestic absorption. That excess production targets Southeast Asia, Middle East, and Latin American markets where U.S. export controls hold less sway and price sensitivity favors Chinese alternatives. Nvidia’s pricing power in non-Western geographies will compress as Biren, Huawei, and Moore Threads flood global markets with chips priced 40-60% below comparable Nvidia SKUs.

Finally, track U.S. policy response. If Chinese domestic chips reach performance parity with Nvidia’s prior-generation A100/H100 by 2027—a plausible timeline given current trajectories—Washington faces a binary choice: further tighten controls and accelerate decoupling, or accept that the semiconductor restrictions permanently forfeited a $50+ billion annual market to strategic competitors.