CIO Tech Outlook Team | Wednesday, 20 August 2025, 11:12 IST
Nvidia is reportedly designing a new AI chip, tentatively named the B30A, for the Chinese market, leveraging its advanced Blackwell architecture. This follows US President Donald Trump’s recent decision in August 2025 to allow more advanced Nvidia chips to be sold in China.
B30A will be a single-die design and combines all the major components onto a single chip of silicon, providing an approximate three-fold decrease in computing capability compared to the dual-die Nvidia B300 accelerator card. The chip will also be equipped with high-bandwidth memory and NVLink data transfer technology at Nvidia, which transfers data at high speeds between processors.
Nvidia plans to provide B30A samples to Chinese clients for evaluation as early as next month, though final specifications are still being refined. The company stated, “We evaluate a variety of products for our roadmap, so that we can be prepared to compete to the extent that governments allow.” Nvidia further emphasized, “Everything we offer is with the full approval of the applicable authorities and designed solely for beneficial commercial use.”
Also Read: TATA Projects Drives Digital Transformation with SAP Cloud Solutions
Additionally, Nvidia is developing another chip, the RTX6000D, tailored for AI inference tasks in China, also based on the Blackwell architecture. Priced lower than Nvidia’s H20 chip, the RTX6000D will use GDDR memory with a bandwidth of 1,398GB per second, just below the 1.4TB threshold set by US export restrictions.
Nvidia and AMD have agreed to contribute 15percent of revenue from certain advanced chip sales in China to the US government, aligning with new regulatory agreements. These developments highlight Nvidia’s strategic efforts to navigate evolving US export policies while maintaining its competitive edge in China’s AI market.