SINGAPORE/HONG KONG, Jan 2, 2026, 06:46 ET
- Shanghai Biren Technology closed up 76% in its Hong Kong debut after a heavily oversubscribed IPO.
- Baidu said its AI chip unit Kunlunxin confidentially filed for a Hong Kong listing on Jan. 1.
- Samsung said customers have praised its next-generation HBM4 memory as it seeks to supply Nvidia.
Shares of Chinese AI chip designer Shanghai Biren Technology closed up 76% in their Hong Kong debut on Friday, the market’s first listing of 2026. The stock opened at HK$35.70, hit HK$42.88 at an intraday high and ended at HK$34.46, versus an offer price of HK$19.60. Biren raised HK$5.58 billion, with institutional demand nearly 26 times the shares on offer and the retail tranche oversubscribed about 2,348 times, exchange filings showed. 1
The strong start underscores how quickly China’s AI hardware firms are tapping public markets as Washington tightens export restrictions on advanced chips. For Nvidia, the rush matters because graphics processing units, or GPUs — processors used to train and run AI models — sit at the center of data-center spending, and fresh capital can speed up domestic rivals.
Baidu said its AI chip unit Kunlunxin confidentially filed a listing application with the Hong Kong stock exchange on Jan. 1, paving the way for a spin-off and separate listing. Kunlunxin was valued at 21 billion yuan ($3 billion) in a fundraising round, and Baidu said the chip unit is expected to remain a subsidiary after any spin-off, with deal details not yet finalised. Hong Kong raised $36.5 billion from 114 new listings in 2025, LSEG data showed, as China’s push for domestic semiconductor alternatives gathers pace. 2
Samsung Electronics said customers have praised the competitiveness of its next-generation high-bandwidth memory chip, known as HBM4, as it tries to close the gap with market leader SK Hynix in memory used for AI data centers. In a New Year address, co-CEO and chip chief Jun Young-hyun said customers told Samsung that “Samsung is back,” and the company is in discussions to supply HBM4 to Nvidia. 3
High-bandwidth memory is stacked DRAM that sits close to an AI processor, helping move data faster than conventional memory. Tight HBM supply can cap how many AI systems can be built, regardless of how many processors are on order.
Export controls are also reshaping the manufacturing side of the chip industry. The U.S. government granted Taiwan Semiconductor Manufacturing Co an annual licence to import U.S. chipmaking equipment to its plant in Nanjing, China, after prior exemptions under a “validated end-user” status expired at year-end, Reuters reported. 4
Validated end-user status is a U.S. export rule that lets pre-approved firms receive certain controlled items without separate licences for each shipment. Shifting to annual licences adds friction and uncertainty, even for mature-node fabs that make older-generation chips.
China’s equity markets have already rewarded AI chip listings aggressively. MetaX Integrated Circuits jumped about 700% in its Shanghai debut in December after raising roughly $600 million, and larger rival Moore Threads gained about 400% after its own offering, as investors bet on Beijing’s drive to reduce reliance on foreign chipmakers. 5
The new IPO wave is giving domestic designers more cash for research and commercialization. It does not erase the technical and supply hurdles that come with building competitive chips, software tools and manufacturing partnerships at scale.
Nvidia has dominated data-center AI computing in part because its processors are tightly coupled with software that developers rely on to build and run models. Investors are now watching whether China’s newly listed chipmakers can build comparable ecosystems while winning customers beyond their home markets.
