China’s State Administration for Market Regulation is investigating U.S. technology giant Nvidia for possible violations of antitrust laws and the agreement surrounding Nvidia’s 2020 acquisition of Mellanox Technologies.
While the first argument only makes sense as retaliation for the latest round of US sanctions against China, the second argument could be valid. Nvidia disputes the allegations.
In any case, NVIDIA denied rumors that it plans to reduce sales to China, which accounted for 15% of revenue in the three months ending in October (NVIDIA’s fiscal third quarter). In fact, NVIDIA is expanding its presence in China, focusing on areas that are not subject to export restrictions, such as self-driving cars.
China has already weakened the company’s core data center business by banning the export of cutting-edge GPU processors to China, as US Commerce Secretary Gina Raimondo and the department’s Bureau of Industry and Security (BIS) have already weakened the company’s core data center business. It seems illogical to accuse NVIDIA of monopolistic behavior.
On the contrary, perhaps China has given Huawei and other Chinese IC design companies the opportunity and strong incentives to build their own AI processor businesses while the hands of their most formidable competitors are tied. We should be grateful to the government.
In 2022, BIS banned exports of Nvidia’s top-of-the-line A100 and H100 processors to China. In 2023, it banned exports of the A800, a scaled-down version of the A100 specifically designed to meet BIS requirements.
The A800 was a bestseller in China, so BIS lowered its standards, forcing Nvidia to design another chip with even lower performance, the H20. However, Chinese AI processors, led by Huawei’s 910B, have proven to be competitive with the H20, which has not sold well in China. And Huawei claims its successor, the 910C, can match the performance of Nvidia’s H100.
In any case, Baidu, Tencent and other Chinese customers are aware that U.S. sanctions are not based on strict technical standards related to national security, as U.S. officials claim, and can easily change at any time. When it became clear, they started switching to Huawei and other domestic chips. This is to punish China.
Exports of Nvidia’s new, significantly more powerful Blackwell B200 AI processors to China are prohibited under existing sanctions, but a reduced-featured version potentially called B20 is reportedly in the works. There is. But why was it more successful than H20?
Regarding Mellanox Technologies, the Chinese government requires that access to Mellanox interconnect technology remain open, that Nvidia supplies Mellanox interconnect products to Chinese customers on a nondiscriminatory basis, that it does not bundle them with its own GPUs, and that Nvidia The acquisition was approved on condition that GPU interoperability be guaranteed. Proprietary GPUs and other interconnect products.
Nikkei Asia said: “It is not clear which of these conditions NVIDIA is alleged to have violated.”
Mellanox is an Israeli company that designs and supplies InfiniBand and Ethernet interconnect adapters, switches, and other devices and software for use in high-performance computing, data centers, cloud storage, and financial services. Nvidia arranged to acquire 100% of the company in 2019.
Where did China get the idea for an antitrust investigation? Now, last July, France confirmed it was investigating Nvidia for alleged anticompetitive conduct, and in September, the US Department of Justice The company was subpoenaed for information regarding potentially restrictive marketing practices. Nvidia’s share of the AI processor market outside of China is estimated at around 90%.
US sanctions not only block the export of advanced AI processors made by Nvidia and its smaller competitor AMD to China, but also prevent China from purchasing EUV lithography systems made by ASML in the Netherlands.
For this reason, it is impossible to manufacture integrated circuits (ICs) with design rules smaller than 5nm in China, and 7nm is the limit at which sufficient efficiency can be achieved. Blackwell processors, on the other hand, are manufactured by TSMC using the 4nm process. As a result, Nvidia now has a significant lead over Chinese players in the race to develop more sophisticated AI processors.
However, chasing Nvidia isn’t the only way to advance AI. In recent months, Chinese research institutes have been building AI processors based on the RISC-V (pronounced “risk-five”) open architecture design standard and large-scale language models for military use based on open source models, including Meta’s Llama. It is reported that it was developed.
The Chinese aren’t the only ones using RISC-V. Last February, up-and-coming American NVIDIA competitor Tenstorrent announced an agreement to license its RISC-V CPU technology to Japan’s Advanced Semiconductor Technology Center, manufacturing and packaging in Hokkaido with new Japanese technology. We developed an AI processor consisting of standardized chiplets. Lapidus from IC Foundry.
China is using chiplets as a way to get around the lack of access to EUV lithography tools. As explained in MIT Technology Review:
In contrast to traditional chips, which integrate all components on a single piece of silicon, chiplets take a modular approach. Each chiplet has dedicated functions such as data processing and storage. They are connected into one system.
Each chiplet is smaller and more specialized, making it cheaper to manufacture and less likely to malfunction. At the same time, individual chiplets in the system can be replaced with newer, better versions to improve performance, while other functional components remain the same.
RISC-V is an open standard instruction set architecture based on reduced instruction set computer design principles. This is a free, non-proprietary platform for IC processor development.
RISC-V, an alternative to Arm, Intel, AMD, and Nvidia, has great potential not only in China but also in the EU and for small businesses and IC designers looking to establish a low-cost, independent presence in the semiconductor and computing markets. It’s attracting interest.
US sanctions are forcing China to accelerate the development of a national ecosystem for advanced ICs based on RISC-V.
The RISC concept was invented at the University of California, Berkeley in 2010. The RISC-V Foundation was established in 2015 to support and manage this technology, with the Chinese Academy of Sciences Institute of Computing Technology joining as one of its founders. Other Chinese members of the foundation include Huawei, ZTE, Tencent, and Alibaba.
In 2020, the foundation relocated from the United States and incorporated as RISC-V International Association in Switzerland to avoid potential interference by US President Donald Trump. It is estimated that China now accounts for about half of the world’s RISC-V core shipments, beyond the scope of US sanctions.
Meanwhile, DigiTimes reported that NVIDIA has hired “hundreds” of new employees in China to work on autonomous driving. BYD and about a dozen other Chinese car companies are using the company’s DRIVE Orin system-on-chip (SoC) in self-driving electric vehicles, generating more than $1 billion in annual revenue for NVIDIA.
Looking ahead to 2025, at least five Chinese EV manufacturers plan to use DRIVE Thor, the successor to DRIVE Orin: BYD, XPeng, GAC-Aion, Li Auto, and Zeekr. However, DRIVE Thor incorporates the generative AI capabilities of Nvidia’s cutting-edge Blackwell architecture. Will the US government try to stop this too?
The Chinese are hedging. BYD, Li Auto, ChangAn and many other Chinese automakers are also partnering with Horizon Robotics, China’s leading computing solution for advanced driver assistance systems (ADAS) and autonomous driving (AD) for consumer vehicles. We are cooperating. So is Volkswagen. In October, Horizon Robotics went public in Hong Kong in the year’s largest IPO.
Follow this writer on X: @ScottFo83517667
