NVIDIA prepares export-compliant B30A for China at roughly double H20 price

NVIDIA has readied an export-compliant single-die B30A accelerator for Chinese markets and companies, priced at roughly twice the H20, as demand from Chinese Artificial Intelligence firms remains strong despite regulatory warnings.

NVIDIA acquired an export license for its Hopper H20 GPU and immediately prepared a modified Blackwell B30A design for China, according to the article. sources close to Reuters report massive demand from Chinese companies for NVIDIA hardware despite Beijing regulators warning against the use of NVIDIA-supplied equipment in new data center buildouts. the article states H20 pricing in China is not provided and is listed as Not stated, and it says the B30A will be priced at roughly double the H20 amount with specific ranges also Not stated; the lower price tier may be reserved for larger volume shipments. Chinese firms reportedly view the B30A as a good deal because of NVIDIA´s performance and software compatibility.

The B30A is described as an export-compliant single-die design that uses the same base die as the dual-die B300 but embeds only one die in the package to meet export restrictions. based on Blackwell Ultra (B300) numbers, the article provides expected peak throughput for the single-die B30A at roughly 7.5 petaFLOPS FP4, 3.75 petaFLOPS FP6/FP8, 1.875 petaFLOPS FP16/BF16, and 0.94 petaFLOPS at TF32, which is about half the peak throughput of the dual-die B300. these figures are presented as derived expectations rather than confirmed benchmarks for the B30A.

The article highlights market implications, noting NVIDIA could generate substantial sales in China because rewriting infrastructure to adopt domestic accelerators will take time and could slow China´s Artificial Intelligence ambitions. while domestic accelerator designs and software mature, NVIDIA GPUs are said to remain in persistent demand. the timeline for B30A shipments is described as unknown in the article, but it expects the first Chinese Artificial Intelligence labs to receive these accelerators within the next few weeks.

70

Impact Score

Giga Computing expands NVIDIA RTX PRO server portfolio for Artificial Intelligence

Giga Computing, a subsidiary of GIGABYTE Group, announced the XL44-SX2-AAS1 server integrating NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs with BlueField-3 DPU and ConnectX-8 SuperNICs. The system is designed to accelerate next-generation enterprise Artificial Intelligence workloads by combining high-density compute and high-speed networking.

NVIDIA pledges artificial intelligence education funding for K-12 programs

NVIDIA announced new support for Artificial Intelligence education in K-12 classrooms at a White House event, pledging unspecified funding and partnerships to adapt NVIDIA Deep Learning Institute and NVIDIA Academy content for high school students. The company said the effort aligns with a White House executive order and aims to expand educator training and student access.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.