China's State Administration for Market Regulation (SAMR) announced on September 15 that it has launched a further investigation into NVIDIA for violating the country's Anti-Monopoly Law. The agency said NVIDIA breached conditions set in its earlier approval of the company's acquisition of Mellanox Technologies and would now face additional regulatory scrutiny.

The development comes as NVIDIA is rumored to be preparing a new China-specific AI chip, the B30A, following restrictions on its H20 accelerator. Market reports suggest the B30A will be built on TSMC's N4P process, feature next-generation high-bandwidth memory, and deliver about half the computing power of the flagship B300. The chip could debut as early as the fourth quarter, though its approval remains uncertain.

NVIDIA CEO Jensen Huang, during a brief visit to Taiwan in late August, said it was “too early” to confirm whether new B-series chips would be launched for mainland China, adding that any such decision would ultimately depend on U.S. government approval.

Industry analysts believe the B30A could help NVIDIA maintain its dominant position in China's AI data center market, where demand remains strong despite export restrictions. The company previously indicated that revenue from H20 shipments to mainland China could reach up to $5 billion in the third quarter alone.

Still, experts caution that China's long-term strategy aims to reduce reliance on foreign suppliers. “Chinese tech companies want NVIDIA chips, but the government is committed to semiconductor self-sufficiency. The challenge is that domestic production cannot yet fill the gap,” said Chip War author Chris Miller.

Meanwhile, U.S. policy under President Trump continues to limit exports of advanced AI chips, requiring performance to be 30%–50% below top-tier Blackwell products before they can be shipped to China. Whether the B30A will meet regulatory thresholds—and secure approval for volume sales—remains to be seen.