Nvidia faces local competition with ‘China-focused’ GPUs

Nvidia has cut prices for the special GPUs it makes for the Chinese market in the face of local competition from Huawei.

The GPU giant developed new products last year for the vast and lucrative Chinese market after the United States imposed restrictions on the high-performance accelerators that could be exported to China.

but, Registry I got it. Earlier this year, Nvidia discovered that some Chinese customers were less interested in its lower-specced products and tended to favor locally made kit to speed up AI workloads.

according to ReutersThe company is now being forced to sell some of its products at lower prices than competing models made by Chinese tech giant Huawei, and weak demand has allegedly led to a glut of Nvidia chips on the market.

Nvidia declined to comment.

The company’s kit for China includes: H20, L20, L2 GPUsIn particular, the H20 appears to be being sold at a discount of more than 10% compared to Huawei’s Ascend 910B.

The Ascend 910B is said to have performance on par with Nvidia’s A100 GPU and is believed to be an in-house design manufactured by Chinese chipmaker SMIC using 7nm process technology. Ascend 910 product.

The development highlights the problems Nvidia faces in China, which has accounted for 20-25 percent of the company’s revenue in recent years.

That figure has now fallen to about 17%. Some reportsIf U.S. regulations continue to prevent Nvidia from competing effectively with local manufacturers, the stock price could fall further.

Meanwhile, Samsung has denied reports that it is having trouble getting its high-bandwidth memory (HBM) for Nvidia’s GPU products to pass required tests before they can be put on chips.

according to ReutersCiting anonymous sources, the report said the issue is related to heat and power consumption issues with Samsung’s HBM3 chips and may also affect next-generation HBM3E products.

• Nvidia has once again beat market expectations, but for how long can that last?
• AMD CFO Jiang Hu talks about CPUs, GPUs and what’s next
• Arm servers are on Nutanix’s long-term radar, but not yet on its to-do list
• NVIDIA’s future in scientific computing lies in the convergence of AI and HPC

(For more information about HBM, please visit our partner website. Blocks and Files)

Samsung was not immediately available to comment on the matter, but told Reuters that “HBM is a customized memory product that requires an optimization process tailored to customer needs,” and denied there were any issues, saying claims that heat and power consumption were causing the units to fail were untrue.

Nvidia declined to comment on the matter.

Demand for HBM parts could lead to supply shortages. Registry report The situation could get even worse if problems emerge with chips made by Samsung this week.

Although HBM chips are manufactured by other suppliers, SK Hynix issued a warning. Early this month The company said it has already sold out most of its production for this year and 2025, citing rising demand due to the AI ​​boom.