Shenzhen: Chinese tech companies are introducing support for their country's artificial intelligence (AI) chips as Beijing suppresses its dependence on US chip maker Nvidia and uses local processors to seek self-sufficiency.
Tencent Cloud said on September 16 that the computing platform, which uses many different chips to provide AI computing power, is perfectly suited to support “mainstream domestic chips.”
“In the last 12 months, more and more domestically produced chips have become available in the market,” Dowson Tong, Senior Vice President of Tencent, told reporters on September 15.
The company is not alone. Alibaba and Baidu are said to have begun using chips they developed to train several AI models, according to a September Information report.
In August, Chinese AI startup Deepseek released an upgrade to its flagship, a massive language model, saying it has a format compatible with the next-generation domestic chips. And in July, another Chinese model manufacturer Stepfun announced the “Ecosystem Innovation Alliance” with a small number of local chip companies.
“Chinese AI industry is rapidly changing with greater intake of domestic hardware as it enters a transition period that could become a completely Chinese AI stack,” said Paul Triolo, partner at consulting DGA-Albright Stonebridge Group.
“But we're not there yet,” he told The Straits Times. “The process will take some time, but it appears likely to be accelerated by domestic innovation, Chinese government preferences and continued export control pressures from the US.”
The adoption of domestic chips by Chinese tech companies is “accelerating, but still partial,” said Charlie Dai, vice-chairman and principal analyst at advisory firm Forrester Research.
By guiding businesses to domestic alternatives, China aims to “accelerate the maturation of semiconductor ecosystems,” he added.
According to forecasts from equity research and brokerage company Bernstein, locally developed AI chips are projected to gain a 55% share of the Chinese market in 2027, from 17% in 2027.
“Most of these (Chinese tech companies) are currently in the testing stage,” said Dr. Lin Chingian, senior analyst at Bernstein, China's Semiconductor.
Companies are focusing on domestic processors as uncertainty over Nvidia's H20 chip supply “forced local cloud service providers to find alternatives.” Meanwhile, “Local chip performance is catching up to (The) H20.”
Still, the Chinese chips – The fact that China's chips currently being improved offers Lag Navizia products for now has hampered wider adoption, analysts say.
The Financial Times reported in August that the release of the new Deepseek R2 model was delayed due to failure to train on Huawei's chips. Alibaba and Baidu use Nvidia's chips to develop the most advanced AI models.
It's not just about raw chip performances. Nvidia's Edge owe much to the software ecosystem used to leverage the chips using the CUDA platform widely used by AI developers.
China also tracks behind the cutting edge of high-bandwidth memory technology. This allows large AI workloads to run more efficiently. Furthermore, the amount of advanced tips that Chinese foundries can fire is currently limited, but this is said to increase in 2026.
Semiconductor analyst Ray Wang thinks it's too early for Chinese tech companies to conclude that adoption of homemade processors is on a massive increase. “For now, domestic chip use is largely limited to state-supported agencies and government-related data centers,” he said.
“If Nvidia solutions are available in any way, we expect to dominate adoption given the benefits of superior performance and ecosystems despite improved domestic delivery,” added the director of research in semiconductors, supply chains and emerging technologies at the advisory firm The Futurum Group.
Over the past few months, Nvidia has been under pressure from Beijing. On September 15, authorities said the company had violated China's anti-monopoly law.
Chinese authorities reportedly told major tech companies to stop purchasing Nvidia's RTX Pro 6000D chips.
“The willingness to block China's Nvidia is a strong indication of faith in domestic capabilities,” wrote Andrew Ng, a British-American AI expert who worked for both Baidu and Google in a LinkedIn post.
From the tech giants to the startups, many Chinese chip design companies are firing to emulate Nvidia's success with AI processors.
Hisilicon, the Huawei chip arm, is considered a leader in this respect. Develops promotion ranges for AI chips. AI chips are not individually stronger than Nvidia, but can be clustered to improve performance. On September 18, the company announced a three-year roadmap for new releases of these chips, saying it will double its computing power with each annual release.
Tech giants like Alibaba and Baidu design their own chips via T-head and Kunlunxisin units, respectively. Both companies recently made headlines by attacking substantial chip deals linked to state-run Telcos China Unicom and China Mobile respectively.
However, Alibaba announced on September 24 that it is partnering with NVIDIA on systems that include “physical AI” or robots and self-driving vehicles by integrating US Giant's software stack with a cloud platform.
Later, there have been chip design startups in China like Cambricon and Hygon. They have seen a surge in stock prices in hopes of profiting from the demand repurposed from Nvidia. Cambricon, which saw its revenue rise 44 times in the first half of 2025, became China's most expensive stock in late August, past Kweichow Moutai.
Additionally, others like Chinese chip companies Moores Red, Metax and Enflame have been published to raise funds for research and development. Many have founders who previously worked for a US chip company. Started by a former Nvidia executive, Moores Red received regulator approval on September 26th.
More people are paying attention to the AI chip scene.
“Nvidia is our goal,” says Knight Jiang, product director for the Xi'an-based graphics processing unit manufacturer. His company's chips have not yet been developed for AI workloads, but they intend to flow into a vibrant field in the future.
“Now we're way too late and others may not think of many of us,” he told St. “But we have a goal and we can aim to chase it.” -The Straits Times/Ann
