Why Byte Dance is Moving Away from Nvidia and Choosing Huawei Chips for AI Development
 |
| Image Source: Google | Image by: Huawei |
- ByteDance is developing an AI model using Huawei's Ascend 910B chips due to restrictions on Nvidia.
- The company received fewer than 30,000 of its 100,000 ordered chips, causing potential delays.
- ByteDance’s new AI model may be less powerful than its current model, Doubao.
ByteDance, the parent company of TikTok, is preparing to release a new AI model using Huawei chips, according to sources reported by Reuters. As tensions rise between the US and China, Chinese tech giants are shifting to domestic chip suppliers like Huawei, particularly since the US government imposed restrictions on advanced AI processors like Nvidia’s in 2022.
The latest move by ByteDance signals a strategic shift in response to these restrictions, but the company may face challenges due to the lower processing power of Huawei's chips compared to Nvidia's. Let’s explore what this means for ByteDance, the future of AI development in China, and the role of Huawei in this growing field.
Why ByteDance is Turning to Huawei Chips
The AI industry requires immense computing power to train large language models (LLMs), analyze massive amounts of data, and power applications like virtual assistants, recommendation algorithms, and pattern recognition systems. Nvidia, one of the world’s leading producers of high-performance GPUs, has been the go-to supplier for companies developing these AI models.
However, the US government banned the sale of Nvidia’s top-tier AI chips to China in 2022, aiming to limit Beijing’s access to critical technology. This has forced many Chinese tech companies to look for alternatives, with Huawei emerging as a key player in the domestic chip market.
Huawei’s Ascend 910B chips have been seen as a local alternative, though they are not as powerful as Nvidia’s GPUs. ByteDance, following Beijing’s urging to reduce reliance on foreign technology, has decided to use these chips for its new AI projects.
ByteDance’s Current Use of AI
ByteDance is already a significant player in the AI space. In August 2023, the company launched a large language model called Doubao, which powers various applications like Jimeng, a text-to-video tool. Doubao has proven to be a major success, with ByteDance unveiling new video-focused versions of the model to compete with giants like OpenAI.
AI plays a crucial role in ByteDance’s operations, from its algorithm-driven content recommendations on TikTok to advanced tools like chatbots and video editors. The Doubao chatbot alone has grown to over 10 million monthly active users, making it one of the most popular AI applications in China.
As ByteDance continues to expand its AI offerings, it has become one of Huawei’s biggest customers for AI chips, further cementing its reliance on domestic technology.
Challenges for ByteDance’s New AI Model
While ByteDance is making moves to develop a new AI model using Huawei’s Ascend chips, there are several challenges ahead. Sources indicate that the new model may not be as powerful or complex as the existing Doubao model, which was built using more advanced AI chips.
One major issue is that ByteDance had ordered 100,000 Ascend 910B chips to support its AI projects but received fewer than 30,000 by July 2024. This shortage of chips is likely to delay the development of ByteDance’s new AI model and limit its overall capabilities.
The Ascend 910B chips, while advanced by Chinese standards, do not match the performance of Nvidia’s GPUs, which are widely regarded as the best in the industry for AI development. This gap in computing power may leave ByteDance’s new AI model less competitive on a global scale, especially as companies in other countries continue to use Nvidia’s cutting-edge technology.
Huawei’s role and China’s AI ambitions
Huawei has become a vital part of China’s effort to reduce its dependence on foreign technology, particularly in the AI and semiconductor industries. However, Huawei’s chips still lag behind their Western counterparts in terms of performance, and the company faces several production challenges that could slow its ability to close the gap.
Huawei’s Ascend chips are generally seen as a solid step toward building a more self-sufficient tech industry in China. However, without access to the most advanced manufacturing equipment, which is restricted by US sanctions, Huawei’s ability to produce cutting-edge chips remains limited.
Despite these challenges, the demand for AI chips in China continues to grow rapidly. Companies like ByteDance, Baidu, and Tencent are investing heavily in AI research and development, hoping to remain competitive in the global tech landscape.
In fact, Nvidia itself is reportedly working on a custom AI chip designed for the Chinese market, which would comply with US regulations while still offering some level of advanced AI computing power.
What’s Next for ByteDance?
For now, ByteDance is moving ahead with its plans to develop a new AI model using Huawei’s chips, though the timeline may be delayed due to the limited supply of Ascend 910B processors. While the new model is expected to be less powerful than Doubao, ByteDance remains committed to growing its AI capabilities.
As US trade sanctions continue to impact Chinese tech companies, we can expect to see more reliance on domestic solutions like Huawei’s chips. However, it’s unclear whether these solutions will be enough to keep up with the rapid advancements happening in AI development globally.
In the coming months, it will be important to watch how ByteDance navigates these challenges and whether it can continue to innovate in the face of limited access to the most advanced AI technology.
Conclusion
ByteDance’s decision to use Huawei chips for its new AI model highlights the growing tension between US and Chinese tech industries. While ByteDance has made impressive strides in AI development with its Doubao model, the company’s future progress may be slowed by the limitations of domestic chip suppliers like Huawei. Nevertheless, China’s determination to reduce reliance on foreign technology could lead to innovative solutions and further advancements in the AI field.
0 Comments