How US sanctions are fueling China’s ambition for chip self-reliance

Semiconductors, the tiny chips that power everything from smartphones to cars, have become one of the most critical components driving modern technological advancements.

As industries worldwide expand, the demand for these chips continues to surge, driven by the growing diversity of tech products and the increasing needs of global industries.

At the heart of this expanding market lies the chip war between the US and China.

Since 2022, a series of US sanctions have targeted China’s access to advanced semiconductors and AI chips, notably from companies like Nvidia.

The US claims that these restrictions are aimed at preventing China from accessing cutting-edge AI capabilities in its nuclear and conventional weapons programs.

"Even though we’ve seen a recent increase in the number of countries working on chip technology, when we look at the top-quality semiconductors that meet these demands, it's clear that the US and China are the two key players." Ersin Cahmutoglu, a cybersecurity expert at ADEO Cyber Security explains to TRT World.

He points out that the sanctions have escalated the competition between them, transforming what was initially a technological race into a national security issue​.

‘Cloud loophole’

Following the US export bans, Chinese firms began to explore alternative routes to access critical semiconductor technologies.

One of the key workarounds was through cloud computing. Chinese entities turned to Amazon Web Services (AWS) and other cloud platforms to access Nvidia’s A100 and H100 chips, which are essential for training AI models​.

Although US sanctions blocked the physical import of advanced chips, Chinese companies leveraged cloud-based computing power to continue developing their AI capabilities​.

Paul Triolo, technology policy lead at the Albert Stonebridge group, highlights the strategies Chinese companies are using to navigate around US export controls.

“Chinese companies have preferred Nvidia GPUs to train models and stockpiled a sizable number before and after the October 2022 US export controls targeting GPUs (graphic processing units).”

“One major loophole in the US export control system for GPUs,” Triolo notes, “is that it does not include cloud services that provide access to hardware through ‘infrastructure as a service’ arrangements.”

He points out that cloud service providers remain legally allowed to provide services to Chinese firms in the AI sector, as “there are currently no laws preventing this”.

Public records reveal that Chinese institutions like Shenzhen University used cloud services provided by AWS to bypass restrictions.

“So currently, Chinese firms can use their stockpiled GPUs, access GPUs in the cloud outside China, or use domestic GPUs manufactured by Huawei and a range of startup firms,” global tech expert Triolo tells TRT World.

However, this loophole is now under scrutiny.

In January 2024, the US Commerce Department proposed regulations to tighten cloud access, potentially restricting Chinese companies' ability to utilise US cloud services​ but also adding further urgency to Beijing’s push for domestic chip production.

Home-grown production

As a result of US sanctions on targeting the export of high-end semiconductors and AI chips, China began to pour more resources into its own semiconductor industry.

Triolo points out that this action has made Chinese semiconductor tool manufacturers much more competitive domestically and internationally in some key areas needed for chip manufacturing.

Companies like Huawei, the Shenzhen-based tech giant, took centre stage in this effort by developing new AI chips to replace those from American firms.

In 2024, Huawei introduced the Ascend 910C chip, a domestic alternative to Nvidia’s A100 —which was barred from being sold to China under US export controls in 2022​​.

According to sources, Huawei has already begun sending samples of the Ascend 910C to large Chinese internet firms and server companies for testing​.

"The US sees Huawei as a security threat, while China views companies like Intel and Nvidia in the same way. Both countries have imposed sanctions because they see each other's chip technologies as a threat," Cagmutoglu says, discussing how their rivalry emphasises the growing geopolitical importance of semiconductor technology.

“This pressure extends to their allies, with the US urging European countries to ban Huawei products.”

In response to these mounting pressures, the Chinese government has issued guidance to its companies, urging them to avoid purchasing Nvidia’s H20 chips and instead focus on locally produced AI chips.

While the directive is not an outright ban, Chinese regulators are encouraging companies to support domestic vendors like Huawei and Cambricon.

Huawei’s efforts to advance its domestic chip technology fit within this larger national initiative.

In 2024, China imported a record $26 billion worth of chip-making equipment, reflecting its commitment to increasing local production capacity.

As Cahmutoglu observes, “China’s current strategy aims to become even stronger than the US in chip production. The country is not only developing alternatives but working on a sustainable ecosystem that will allow it to control its own semiconductor supply chain."

China Telecom, one of the country's largest state-owned telecom operators, recently announced the development of two advanced large language models (LLMs) that were trained entirely on Chinese-made chips, including Huawei’s AI processors.

This development marks a significant milestone in China’s pursuit for self-reliance, demonstrating that its companies are increasingly capable despite Washington's restrictions.

China Telecom states that the company used tens of thousands of domestically produced chips to train their TeleChat2-115B model and another, unnamed model. The TeleChat2-115B boasts over 100 billion parameters, showcasing the growing sophistication of AI model development in China.

According to a report by ITIF, China leads in the number of AI research publications globally and is closing the gap with the United States in generative AI models.

While Chinese research traditionally lagged in terms of impact and private sector involvement, initiatives from institutions like Tsinghua University have produced some of the country's leading AI companies, such as Zhipu AI and Baichuan AI.

The report reveals that China now produces more AI research than the US, with universities like Tsinghua University being “ the breeding ground for China's leading AI start-ups”.

Future implications

By actively collaborating with domestic firms like China Telecom and Cambricon, Huawei is laying the groundwork for a comprehensive ecosystem of AI development powered by local chips.

This collaboration aims to reduce China’s reliance on Nvidia and other US suppliers, making it less vulnerable to future sanctions​.

As China continues to advance its domestic semiconductor capabilities, the effects are being felt on a global scale.

One of the significant consequences of China’s drive for self-sufficiency has been the impact on US and allied companies.

Paul Triolo notes that “the biggest losers have been US and allied toolmakers, who have lost billions in revenue, in some cases been forced to lay off workers and close facilities, at the same time as losing market share in China to competitors.”

However, challenges remain.

While Huawei’s AI chips, like the Ascend 910C, are being positioned as alternatives to Nvidia’s A100, experts remain cautious about whether these chips can fully match the performance of their American counterparts.

“Nonetheless, no one can say that China won’t catch up in the future. They’ve started this process with Huawei and may expand it with other local companies in the future. The US is trying to prevent this through sanctions and political pressure, but whether it will succeed remains to be seen.” expert Cahmutoglu says.