NVIDIA is set to launch a new AI chip for China at a much lower price than its restricted H20 model, with mass production starting in June, sources familiar with the matter told Reuters.
The sources said the GPUs will be part of NVIDIA’s most recent generation Blackwell architecture AI processors, priced between $6500 and $8000, much below the H20 model, which sold between $10,000 and $12,000.
The lower price reflects its weaker specifications and simpler manufacturing requirements, which can reduce the company’s costs through simplified designs.
According to the two sources, the upcoming AI chip will be built on NVIDIA’s RTX Pro 6000D, a server-grade graphics processor. It will use standard GDDR7 memory rather than the more sophisticated high-bandwidth memory (HBM).
They further mentioned that it would not incorporate the advanced Chip-on-Wafer-on-Substrate (CoWoS) packaging technology from Taiwan Semiconductor Manufacturing Co (2330.TW).
According to a spokesperson from NVIDIA, the company is still assessing its “limited” options. “Until we settle on a new product design and receive approval from the US government, we are effectively foreclosed from China’s $50 billion data centre market,” Reuters quoted the source.
Reports show that NVIDIA’s position in the Chinese market stood at 95% four years ago, but it has significantly dropped to 50% due to the US government’s restricted regulations on exporting the company’s AI chips to China. This new AI chip leverages NVIDIA’s still-existing ground presence in the Chinese AI sector.
Export restrictions have resulted in financial setbacks for NVIDIA, including a $5.5 billion write-off and around $15 billion in lost sales. Hence, this approach could be a strategic move to compensate for the losses.
Reuters also reported that NVIDIA investors will look for definitive answers on how much US chip curbs on China will cost when the company releases its results on Wednesday, even as a pullback in other regulations is expected to open up new markets.
