China Urges Shift to Domestic AI Chips, Challenging NVIDIA's Dominance

Chinese think tank advises data centers to stick with Nvidia AI GPUs -- homegrown Chinese GPUs suffer from 'high costs' and 'complex engineering'

On Sunday, the Chinese-government-backed research institute and think tank CAICT issued a detailed report regarding China's computational power. The report (vias SCMP''') warned data centers to keep on using Nvidia-based solutions. The institute cited 'high costs' and 'complex engineering' as one of the few hurdles that data centers would face if they were to transfer their models to Chinese chips. It should not be surprising that the US government has barred Nvidia from exporting its high-performance A100 and H100 GPUs to China. Another round of GPU bans came last year when the Biden Administration restricted exports of Nvidia's A800 and H800 alternatives to China for obvious reasons. Furthermore, it appears that Nvidia has also stopped taking orders for its currently permitted HGX H20 GPU. "As the saying goes, "Necessity is the mother of invention," and likewise, many domestic Chinese chip manufacturers have taken this issue into their own hands. However, despite the clear incentive any Chinese data center would have to opt for a local alternative, CAICT advises continuing with Nvidia. "The CAICT stated that "it involves complex engineering to transfer models trained on Nvidia GPUs to domestic solutions due to differences in hardware and software." "Adding to the mix, new data centers in China face numerous problems, including the lack of GPU solutions, supply mismanagement, and rising computing power fragmentation, which essentially refers to the inefficient use of computer resources like a GPU. Data centers must merge solutions from different vendors and play a "mix-and-match" game, hoping to get their systems up and running. A large portion of this inefficiency is caused by incompatible software and varying hardware. CAICT warned that "there are big discrepancies on hardware in IDCs, such as in GPUs, AI accelerators and network structures, which made it harder to manage and dispatch hardware resources to accommodate for differential computing needs of AI tasks, further impeding the use rate." While CAICT has suggested Chinese data centers avoid trouble by sticking with Nvidia, the large gap between supply and demand remains unaddressed. China's goal of self-sufficiency in the semiconductor market is still a "work in progress" as the country, despite investments worth tens of Billions of dollars, still imported $350 Billion of chips in 2023. Players like Intel and AMD are struggling to keep pace with Nvidia due to Team Green's strong integration between hardware and software in the AI market, so homegrown Chinese GPUs might, at best, achieve parity in a few years.

Tech war: China advocates use of local AI chips over those from US powerhouse Nvidia

NVIDIA's CEO Jensen Huang displays products on-stage during the annual Nvidia GTC Artificial Intelligence Conference at SAP Center in San Jose, Calif. March 18. AFP-Yonhap Chinese authorities have informally advised local companies to use domestic-made artificial intelligence (AI) chips over those from Nvidia, according to two sources familiar with the matter, as alternative suppliers on the mainland look to catch up with the U.S. tech giant. While there is no official ban on Nvidia's China-tailored H20 graphics processing unit (GPU), the sources said mainland AI chip users have been informed to prioritize deployment of local alternatives, including those developed by Huawei Technologies. A third source said H20 orders made by mainland enterprises had not been restricted as of August. Reports about China's tacit rules against the American chip design firm first emerged in May, when US tech media The Information reported that regulators had asked local companies to cut back purchases of Nvidia chips and buy more from domestic suppliers like Huawei. A Bloomberg report last week said that China has called on local buyers to stay away from Nvidia. The country's Ministry of Industry and Information Technology, which oversees the domestic semiconductor and AI sectors, has not made any public announcement regarding that matter. The agency did not reply to a request for comment. Nvidia did not also respond to a request for comment on the reported China restrictions. The lack of any official confirmation about the matter somehow reflects the highly sensitive nature of the world's AI chip market, which Taiwan Semiconductor Manufacturing Co chief executive Hsu Ming-chi has projected to grow annually at a much higher rate than the overall semiconductor industry. When asked about his views on U.S. trade restrictions on China, Nvidia chief executive Jensen Huang told American media in Washington on Sept. 28 that the U.S. government is "doing a wonderful job" in finding the right balance between controls and exporting American technology to the world. "It is really terrific that the world is built on American standards," Huang said, "Nvidia is an American company and our government and administration would love to see us succeed." Nvidia's H20 remains popular among users in China, despite its reduced capacity to comply with U.S. sanctions. Sales of the tailor-made chip have steadily picked up since its release earlier this year, as demand from large Chinese cloud services providers enabled them to access Nvidia's technical support and maintenance services. Nvidia is expected to deliver more than 1 million H20 GPUs in China this year, racking up $12 billion in sales. The company logo of China Telecom is displayed at a news conference in Hong Kong, China, Aug. 20, 2018. Reuters-Yonhap Still, adoption of alternative Chinese-made AI chips is growing. State-owned carrier China Telecom, for example, said in a statement on September 28 that it has developed two large language models -- the technology behind generative AI applications like ChatGPT -- that were trained entirely on locally produced AI chips. While it did not identify the local supplier, China Telecom has a track record of partnering with Huawei. A week earlier, Huawei had started offering samples of its Ascend 910C processor to large Chinese server companies for hardware testing and configuration, according to a South China Morning Post report. That put the company a step closer to the commercial roll-out of its upgraded version of the 910B, which is on par with Nvidia's popular A100 chips. Nvidia was initially barred from selling its A100 and H100 GPUs -- two of the most in-demand chips for training and running AI models -- to clients in China in August 2022. It later modified those chips to create the A800 and H800 to get around those restrictions, which Washington barred from being exported to China under tightened curbs announced last October. Nvidia has since developed the H20, L20 and L2 GPUs to maintain its market share in China. Despite several rounds of U.S. restrictions, Nvidia said China was its third-largest market in its financial year ended January 28. Read the full story at SCMP.

China pushes local companies to not buy AI chips from NVIDIA

A new report has cited two anonymous sources familiar with the matter that Chinese authorities have told local companies to avoid purchasing NVIDIA AI chips. The report comes from the South China Morning Post and states that Chinese authorities have told local companies they should be choosing Chinese AI chip solutions over NVIDIA's products, which would include AI chips developed by Huawei Technologies. Notably, China hasn't officially banned NVIDIA-made GPUs, but as a result of the demand for GPU processing power fueled by the emergence of artificial intelligence, an export ban was placed on high-end chips, causing NVIDIA to create a China-specific H20 GPU in order to be able to sell its product across the region. This isn't the first time we have heard of China asking local companies to opt for locally-made chip solutions, as it was back in May that The Information published a story that detailed Chinese regulators asking companies to reduce their purchasing number of NVIDIA GPUs. Additionally, it was only last week Bloomberg reported China asked local buyers to avoid NVIDIA. It's clear from this timeline that Chinese authorities want complete control over the AI chips within its country, seemingly meaning that NVIDIA, an American company, has a limited timeline of being able to trade in China. However, that won't likely happen until Chinese chip developers such as Huawei catch up to the performance NVIDIA products can provide.