China Telecom Develops Large Language Models Using Domestic AI Chips, Bypassing US Sanctions

China Telecom's Breakthrough in AI Development

China Telecom, one of the world's largest mobile networks, has announced a significant advancement in artificial intelligence by developing two large language models (LLMs) trained entirely on domestically produced chips 1

1

. This move marks a crucial milestone in China's pursuit of technological self-sufficiency, especially in the face of US export restrictions on high-end GPUs.

The Models: TeleChat2-115B and a Trillion-Parameter LLM

The state-owned carrier has developed two notable models:

1. TeleChat2-115B: An open-source model with over 100 billion parameters.
2. An unnamed model reportedly boasting 1 trillion parameters.

These models were trained using tens of thousands of locally manufactured chips, demonstrating China's capability to develop advanced AI technologies without relying on Western hardware 2

2

.

Huawei's Role in China's AI Chip Industry

While China Telecom hasn't explicitly named its chip supplier, it's highly likely that Huawei provided the majority, if not all, of the chips used in training these LLMs 1

1

. Huawei has been positioning itself as a domestic alternative to Nvidia, with its Ascend line of AI processors gaining traction in the Chinese market.

Training Infrastructure and Capabilities

The GitHub page for TeleChat2-115B mentions compatibility with the "Ascend Atlas 800T A2 training server," a Huawei product that supports Kunpeng 920 processors based on the Arm 8.2 architecture 3

3

. While the parameter count of TeleChat2-115B (100 billion) is lower than some Western models, it still represents a significant achievement given the hardware limitations.

Implications for the US-China Tech Rivalry

This development has several implications for the ongoing technological competition between the US and China:

1. It demonstrates China's progress in developing domestic AI capabilities despite US sanctions.
2. It showcases the potential of Chinese companies like Huawei to fill the void left by restricted access to Nvidia's high-end GPUs.
3. It aligns with Beijing's push for technological self-reliance and reduced dependence on foreign semiconductors.

Future Outlook

As China continues to invest heavily in its semiconductor industry and AI research, we may see further advancements in domestically produced AI chips and models. This could potentially reshape the global AI landscape and intensify the technological rivalry between the US and China.