Can China Address the Global GPU Shortage?

According to media reports, Huawei is working on GPUs that exhibit prowess in running LLMs like GPT-4 effectively
Listen to this story

Many enterprises today want their own GPT models trained on their enterprise data. This has led to a shortage of GPUs in the market. 

According to Liu Qingfeng, the founder of HKUST Xunfei, Huawei has developed an analog GPU that is comparable to the NVIDIA A100 GPU. NVIDIA, the company that dominates the Graphics Processing Units (GPU) space, only sells around 1.5-2 million GPUs a year, which it’s not enough to meet the demand. Reportedly, despite increasing production, NVIDIA’s GPUs are, in fact, sold out for 2024. 

This is where Huawei, or other companies in the Chinese AI hardware ecosystem could step in. Although specific details about Huawei’s GPU are undisclosed, it exhibits prowess in running LLMs like GPT-4 effectively. Chinese media reports suggest that Huawei, which has introduced its LLM model named Pangu as a competitor to GPT models, aims to assist clients in constructing and training AI models using its proprietary Ascend AI processors and the MindSpore AI framework, which underpins the technology behind Pangu.

Chinese firms see an opportunity 

Similar to Silicon Valley, China too, is facing a hardware crisis. The country is also wary that the hardware crisis could further be exacerbated by additional US sanctions. Reports suggest the Biden administration is contemplating fresh export controls which could prevent NVIDIA from selling its chips directly to China. Hence, in their quest to become self-reliant in semiconductors, China too is aiming to become independent in AI hardware. 

While additional restrictions are likely to come further constraining China, domestic companies, however, see this as an opportunity. Since the start of the US-China trade war, numerous AI hardware startups have popped up in the country. A handful of them, notably, have been successful in building AI hardware. Last year, Shanghai-based Vastai Technologies launched its 7nm GPU for cloud AI applications claiming it offers industry-leading graphics rendering performance and world-leading encoding capabilities of ultra-high throughput, ultra-high quality and low latency.

Another Chinese startup named Moore Thread, established in 2020, initially created GPUs for gaming but is now redirecting its efforts towards crafting GPUs for data centres. Interestingly, the startup was founded by former global VP and China GM of NVIDIA, Zhang Jianzhong with fundings from Shenzhen Capital Group, Sequoia Capital China, ByteDance, and Tencent. Besides, companies like ILuvatar CoreX and Biren Tech are actively partnering with indigenous cloud computing providers, such as Baidu, to implement their LLM services.

A global opportunity

While a handful of companies across the globe are working towards breaking NVIDIA’s almost monopolistic hold in the GPU space, China could potentially emerge as a contender. China made its name by making cheaper  ‘copycat’ alternatives to electronic gadgets in the last two decades. This culture became deeply ingrained, driven by factors such as a lack of intellectual property enforcement, a large pool of skilled labour, and a strong manufacturing base. This ‘copycat’ culture has been instrumental in China’s ability to produce cheap electronics and other goods. 

By reverse-engineering established products, Chinese manufacturers can replicate them at lower costs, omitting the research and development phases. While making AI hardware is a different ball game altogether, China, nonetheless, already has the manufacturing base, and by tapping into the country’s ability to produce cheaper alternatives, it could emerge as a potential player to solve the GPU crisis. 

While China’s primary goal is to address its domestic demand, it’s highly conceivable that the country will expand its AI products onto the international stage. Presently, China is engaged in a fierce competition with the US in its pursuit to lead the AI field. This rivalry extends beyond AI hardware; there are indications that Huawei is developing a LLM that might vie with GPT-4, the current pinnacle of LLM technology. Additionally, prominent Chinese companies like Baidu and Alibaba have already introduced their own LLMs, underlining China’s comprehensive efforts in the AI space.

China’s questionable reputation

Nevertheless, due to its autocratic governance, China often faces negative perceptions from other nations. Moreover, its economic strategies, including state subsidies, intellectual property infringement, and unjust trade practices, have triggered apprehensions among global counterparts. In a parallel context, India has taken measures to ban numerous Chinese mobile applications on the grounds of potential national security threats.

The US has banned Huawei, raising concerns that its telecommunications equipment could be used for espionage by the Chinese government. So far, the US government has implemented multiple measures to curtail Huawei’s operations within its borders. Consequently, for China’s AI offerings to gain acceptance on a global scale, the nation needs to address and rectify its questionable reputation. 

Download our Mobile App

Pritam Bordoloi
I have a keen interest in creative writing and artificial intelligence. As a journalist, I deep dive into the world of technology and analyse how it’s restructuring business models and reshaping society.

Subscribe to our newsletter

Join our editors every weekday evening as they steer you through the most significant news of the day.
Your newsletter subscriptions are subject to AIM Privacy Policy and Terms and Conditions.

Our Recent Stories

Our Upcoming Events

3 Ways to Join our Community

Telegram group

Discover special offers, top stories, upcoming events, and more.

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Subscribe to our Daily newsletter

Get our daily awesome stories & videos in your inbox

6 IDEs Built for Rust

Rust IDEs aid efficient code development by offering features like code completion, syntax highlighting, linting, debugging tools, and code refactoring