China Trains AI Model with 1 Trillion Parameters Using Domestic Chips

China Telecom has achieved a major breakthrough in artificial intelligence (AI) by successfully training large language models (LLMs) with 1 trillion parameters, using only domestically-produced chips. This accomplishment is part of China’s broader strategy to become self-reliant in key technologies, especially in light of U.S. restrictions on access to advanced semiconductors.

Get your Official Egonormic Chair For Healthy Lifestyle here

A Milestone in China’s AI Efforts

China’s ambition to establish independence in AI is evident in its state-owned enterprises, like China Telecom, taking center stage in developing AI technologies. According to China Telecom’s Institute of AI, the company developed two LLMs, one named TeleChat2-115B and another unnamed model. Both models were trained exclusively using tens of thousands of domestically-made chips, demonstrating China’s capacity to develop competitive AI solutions without relying on foreign semiconductors.

This achievement comes as the U.S. government tightens export controls, preventing Chinese companies from accessing advanced AI chips, such as Nvidia’s A100 and H100. As a result, Chinese tech firms have accelerated the development of their own processors, with China Telecom’s AI models emerging as a significant success in this context.

The Scale of China’s AI Models

One of the most remarkable aspects of this accomplishment is the scale of the models China Telecom has trained. The unnamed LLM boasts 1 trillion parameters, a milestone that significantly enhances the model’s capabilities and complexity. In AI, parameters are the variables that allow models to learn from data, and the more parameters a model has, the more sophisticated its performance.

Sign Up Free Trial To Create all Your ADS Creatives With Just a Click!

While many global AI giants are racing to develop larger and more powerful models, China’s domestic AI initiatives have made impressive strides. For instance, the TeleChat2-115B model, with 115 billion parameters, is designed to compete with mainstream AI platforms globally.

Overcoming U.S. Sanctions: China’s Alternative Chips

The AI models developed by China Telecom signal a shift towards homegrown AI solutions, especially in response to Washington’s restrictions on Nvidia’s cutting-edge AI chips. To compensate for these sanctions, China has begun relying on domestically-produced alternatives.

While China Telecom hasn’t revealed the specific manufacturers of the chips used in training its LLMs, Huawei’s Ascend chips are believed to be central to China’s AI ambitions. Huawei, which has been heavily sanctioned by the U.S., has ramped up efforts in AI and recently tested its Ascend 910C processor with local companies.

Other Chinese firms, such as Cambricon, a start-up specializing in AI processors, are also collaborating with major players like China Telecom. These partnerships reflect a broader trend in China’s tech sector to build an indigenous AI ecosystem that reduces reliance on foreign technology.

A Self-Sufficient AI Future

China’s growing expertise in AI and semiconductor technology indicates a broader push toward self-sufficiency in an increasingly competitive global market. By developing its own chips and AI models, the country is steadily reducing its dependence on foreign tech giants like Nvidia.

While U.S. sanctions make it challenging for Chinese companies to access advanced Nvidia hardware, a black market for such chips has emerged. However, many firms in China prefer to avoid operating in this grey area and instead opt for domestically-produced chips, which offer official support and services.

China Telecom’s success in training a trillion-parameter AI model is a testament to the nation’s commitment to developing its technological infrastructure and positioning itself as a major player in the AI race.

Sign Up Free For Honeygain to Start Getting Paid Daily Doing Almost Nothing

The Broader Implications

China Telecom’s achievement underscores the broader changes in the global AI landscape. As countries like China seek to bolster their domestic industries, competition in AI development is intensifying, and technological self-sufficiency is becoming more critical in light of trade restrictions and geopolitical tensions.

With companies like Huawei and Cambricon leading the way, China is gradually building an ecosystem that could rival global AI giants, while also protecting its economy from the potential impacts of foreign sanctions.

As China continues to invest heavily in AI, this breakthrough represents just the beginning of a new era of innovation and technological resilience.

For more updates on AI and other tech trends, check out the latest events like the AI & Big Data Expo taking place in Amsterdam, California, and London.

Helpful Tools You can Try Out Today!

AI Discoveries