Nvidia to launch lower-priced Blackwell chip for China amid US export curbs: Report

Nvidia is preparing to launch a new artificial intelligence (AI) chip for the Chinese market at a significantly reduced price compared to its recently restricted H20 model, according to a Reuters report citing sources familiar with the matter.

The upcoming GPU, part of Nvidia's latest Blackwell-architecture AI chips, is expected to cost between $6,500 and $8,000, making it substantially lower than the H20's price tag of $10,000 to $12,000. Two of the sources told Reuters the reduced pricing reflects weaker specifications and simpler manufacturing requirements.

The chip will reportedly be based on Nvidia’s RTX Pro 6000D, a server-grade GPU, and will use conventional GDDR7 memory instead of the more advanced high-bandwidth memory (HBM). It will also not utilise TSMC's advanced Chip-on-Wafer-on-Substrate (CoWoS) packaging technology, according to the sources. These technical details, along with pricing and production plans, have not been previously reported.

Mass production could begin as early as June, Reuters said, citing three unnamed sources who requested anonymity due to lack of authorisation to speak to media.

An Nvidia spokesperson told Reuters that the company is still exploring its “limited” options. “Until we settle on a new product design and receive approval from the U.S. government, we are effectively foreclosed from China’s $50 billion data center market,” the spokesperson said. TSMC declined to comment.

China remains a key market for Nvidia, accounting for 13% of its revenue in the past fiscal year. This marks the third time Nvidia has had to redesign a GPU to comply with tightening U.S. export restrictions aimed at curbing China’s technological advancements.

Despite being less powerful than the H20, the new GPU is seen as a strategy to maintain Nvidia’s competitiveness in the market, where it has lost considerable share due to the export curbs. Huawei, with its domestically produced Ascend 910B chip, has emerged as a major rival.

“Domestic Chinese technologies like Huawei are expected to catch up with the computing performance of downgraded versions within one to two years,” Nori Chiou, investment director at White Oak Capital Partners, told Reuters. He added that Nvidia’s advantage lies in its CUDA software platform, which supports the development of AI applications and has fostered a strong developer ecosystem.

According to two Reuters sources, Nvidia is also developing a second Blackwell-based chip tailored for China, with production possibly starting in September, though its specifications remain unclear.

The company had initially explored creating a downgraded version of the H20, Reuters reported, but abandoned the plan due to architectural limitations under the latest US regulations. Huang has confirmed that Nvidia’s older Hopper-based architecture, which underpins the H20, can no longer be modified to comply with current export rules.

The final name for the forthcoming chip is yet to be confirmed. However, Chinese brokerage GF Securities suggested in a recent note that it might be branded as the 6000D or B40, without disclosing sources or pricing details.