China’s tech giants had already been stockpiling high-performance graphic processing units (GPUs) in anticipation of a growing tech war with the US before Washington imposed a ban on Nvidia’s exports to China. While the ban is causing difficulties for China’s AI startups, the giants have managed to secure enough AI chips to continue their development for the time being.
Key Takeaway
The US chip export ban is posing challenges for China’s AI startups, as they struggle to access advanced chips needed for their development. However, the country’s tech giants, such as Baidu, have managed to secure enough AI chips to continue their projects. Smaller AI players are left with fewer options and may have to settle for less powerful processors or seek acquisition opportunities.
Securing AI Chips
Robin Li, the CEO of Baidu, announced during an earnings call that the company has obtained enough AI chips to support the training of its ChatGPT equivalent, Ernie Bot, for the next year or two. Li also mentioned that inference, which requires less powerful chips, can still be supported by existing reserves and alternative options. However, the long-term impact of the ban on acquiring the most advanced chips is seen to hinder the pace of AI development in China, leading the tech giants to actively seek alternatives.
Other Chinese tech companies, including Baidu, ByteDance, Tencent, and Alibaba, have also taken proactive measures in response to US export controls. They collectively ordered around 100,000 units of A800 processors from Nvidia, costing them up to $4 billion, and purchased $1 billion worth of GPUs scheduled for delivery in 2024.
Challenges for Startups
The upfront investments required for securing GPUs could deter many startups from entering the AI race. However, exceptions exist if a startup manages to secure substantial investments quickly. For example, 01.AI, founded by prominent investor Kai-Fu Lee, acquired a significant number of high-performance inference chips through loans and paid off its debt after raising capital that valued the company at $1 billion.
Smaller AI players, lacking the financial resources to stockpile chips, will have to settle for less powerful processors that are not under US export controls or wait for potential acquisition opportunities. Li predicts that the combination of factors such as the scarcity of advanced chips, high demand for data and AI talent, and significant upfront investments will lead the industry into a consolidation stage.
