Chinese AI Tech & Chips Soar As The U.S. Gov Reconsiders Its Policies

There’s no question that artificial intelligence (AI) and generative artificial intelligence (GenAI) technologies are revolutionizing industries all around the world.

Some would even say that it's dictating how and when sectors advance—a trait the technologies are expected to possess for the foreseeable future.

So, it’s only natural that businesses and governments are playing the AI game extremely cautiously.

The United States is home to the leading AI companies, a list that spans AI models, platforms, and hardware, and consists of names such as OpenAI, xAI, Alphabet (Google), Amazon, Meta, Anthropic, NVIDIA, Apple, Adobe, Palantir, and more.

Suffice it to say, the U.S. is leading the AI race.

However, its biggest competitor—China—is not too far behind. Despite numerous sanctions and trade restrictions being imposed on the country, China has shown great promise when it comes to AI technology—no one’s forgotten about the disruptive DeepSeek that was poised to overthrow ChatGPT, have they?

This list extends to include AI companies such as Manus AI, Alibaba, and Baidu, among others.

Of course, ChatGPT wasn’t knocked off its perch, but it did highlight what the U.S. was trying to avoid by imposing restrictions on China—it didn’t want to lose pole position in the AI race.

This even led to graphics processing unit (GPU) manufacturers to create specialized microchips dedicated for Chinese markets—ones that complied with the trade restrictions. Yet, those cutbacks didn’t hinder China’s progress, as manufacturers from other countries also joined the party—or Chinese companies found alternative routes to source the most advanced U.S.-created chips.

Either way, the U.S. is now reconsidering some of its policies in this regard.

The new Trump administration is looking to do away with the Biden-era rules that control the flow of advanced AI chips based on tiers outlining which and how many chips a country has access to, according to three sources familiar with the matter. The idea is to move to government-to-government agreements.

The sources also said the details haven’t been finalized, and the plan remains under discussion but could be subject to changes. However, the prevailing theory is that if the changes are implemented, enabling access to advanced chips could be a powerful bargaining chip in the ongoing tariff wars, especially with China.

The rules, which are called “The Framework for Artificial Intelligence Diffusion” and came into effect a week before Trump took over, divide the world’s countries into three tiers. The regulations require AI chip makers to comply with them from May 15, 2025. 

Tier 1 consists of 18 countries and includes Taiwan (a key GPU maker), and Tier 2 includes 120 other countries that are subject to strict limitations, and covers India, Israel, Singapore, Switzerland, Yemen, and others. Tier 3 spans “countries of concern” such as China, Russia, Iran, Syria, Venezuela, and Myanmar—countries that are barred from accessing advanced U.S. AI compute capabilities.


On the other side of the world, China is having a great time with its progress in the AI sector.

Leading AI company Alibaba just launched Qwen 3, an advanced and upgraded version of its flagship AI model. The new model comes with new hybrid reasoning capabilities, which combine traditional AI functionality with advanced dynamic reasoning.

This followed a dual release by Baidu, which saw the availability of Ernie 4.5 Turbo and the more reasoning-capable Ernie X1 Turbo models.

The previously banned DeepSeek made a return to South Korea. Users can now once again download DeepSeek, as it’s available on South Korean app markets, following a two-month hiatus for breaching data protection rules.

DeepSeek’s return to the headlines isn’t limited to returning to South Korea.

The ChatGPT challenger has been soaking in rumors about the release of a new model—DeepSeek-R2. The new model is said to have 1.2 trillion parameter, 78 billion active, hybrid MoE (Mixture of Experts, which makes training faster and more efficient), be 97.3% cheaper than OpenAI’s GPT-4o ($0.07/M in, $0.27/M out), and trained on 100% trained on Huawei's Ascend 910B GPU cluster with 82% utilization.

Rumors have it DeepSeek’s R2 is set for a May launch. 

Ahead of AI models, AI chips also saw good news.

China-based Huawei Technologies is gearing up to test its newest and most powerful AI chip—the Ascend 910D—which it hopes will challenge NVIDIA’s top-shelf GPUs, especially the NVIDIA H100. The company is also set to ship its advanced 910C AI chip to Chinese customers within a month.

All of this came as the Chinese government is being vocal about a big AI push—one that will offer major breakthroughs and help the country soar in the AI realm. President Xi Jinping even pledged "self-reliance and self-strengthening" to develop AI in China, as the country looks to take on the U.S.

"We must recognize the gaps and redouble our efforts to comprehensively advance technological innovation, industrial development, and AI-empowered applications," said Xi.

Do you think China could succeed without the help of U.S. AI technology? Do you think the U.S. should do away with its previous trade restrictions?

Let us know in the comments below!