Nvidia is developing a new artificial-intelligence processor for the Chinese market, according to people briefed on the matter. The chip, internally code-named B30A, is based on the company’s latest Blackwell architecture and is designed to satisfy U.S. export-control rules while offering Chinese customers higher performance than the H20 accelerator that regulators recently cleared for sale. The B30A uses a single-die design that is expected to deliver roughly half the raw computing power of Nvidia’s flagship dual-die B300 card but still exceed H20 capabilities. It will feature high-bandwidth memory and the company’s NVLink interconnect, specifications aimed at data-center AI workloads. Nvidia hopes to provide sample units to Chinese clients as early as September, the people said. Whether the product gains final approval from the U.S. Department of Commerce remains uncertain. President Donald Trump signaled last week he could authorize sales of scaled-down Blackwell chips and has imposed a 15% levy on revenue from certain advanced-chip transactions in China, but lawmakers in both parties continue to question broad technology transfers. China accounted for about 13% of Nvidia’s revenue last fiscal year, making the market strategically important as domestic competitors such as Huawei advance their own AI silicon. Alongside the B30A, Nvidia is also readying the lower-spec RTX6000D for inference tasks, underscoring its effort to preserve market share while navigating an increasingly complex regulatory landscape.