Chinese companies are adapting their strategies in response to recent export controls implemented by the United States. Tech giants have proactively stockpiled AI chips, anticipating a potential tech conflict, whereas smaller AI players are grappling to keep up with these developments.
(Photo : JADE GAO/AFP via Getty Images)
This photo taken on July 13, 2022 shows a night view of the city skyline in Shenzhen, China's southern Guangdong province.
Anticipating an Escalating Tech Conflict
Prior to the U.S. export restrictions on high-performance graphic processing units from Nvidia to China, the country's tech giants had been stockpiling these components, anticipating an escalating tech conflict between the two nations.
As revealed by the company's CEO Robin Li during an earnings call this week, Baidu has successfully secured a substantial quantity of AI chips, ensuring the continuous training of its ChatGPT counterpart, Ernie Bot, for at least the "next year or two.
Li also emphasized the company's confidence in using less powerful chips for inference, supported by their chip reserves and exploration of alternative sources, enabling them to sustain various AI-native applications for end-users.
In response to U.S. export controls, other financially robust Chinese tech companies, including Baidu, ByteDance, Tencent, and Alibaba, took proactive measures. Financial Times reported that they collectively placed orders for about 100,000 units of Nvidia's A800 processors, amounting to a substantial investment of up to $4 billion.
Also Read: Biden Administration to Target Export Controls to China Despite Revenue Loss for Some US Firms
Additionally, these companies acquired $1 billion worth of GPUs scheduled for delivery in 2024. These strategic moves underscore the industry's commitment to navigating challenges and seeking alternatives to ensure the uninterrupted progression of AI development in China.
The substantial initial investments required to compete in the Large Language Model (LLM) race may act as a deterrent for many startups. However, exceptions arise for those able to secure significant investments promptly.
An example is 01.AI, founded in late March by prominent investor Kai-Fu Lee, which managed to obtain a substantial number of high-performance inference chips through loans. The company successfully repaid its debt after raising capital resulting in a valuation of $1 billion.
In a strategic move, Baidu, leveraging its reserve of GPUs, recently introduced the Ernie Bot 4. CEO Robin Li asserted that this latest creation is not inferior in any respect to GPT-4.
The assessment of LLMs poses challenges due to the intricate nature of these AI models. Many Chinese AI firms resort to ranking boosting by meticulously meeting the criteria set by LLM charts. However, the true effectiveness of these models in real-life applications is still awaiting comprehensive judgment.
Struggling to Keep Up
Meanwhile, TechCrunch reported that smaller players in the AI landscape, constrained by limited financial resources to accumulate chip reserves, will likely opt for less potent processors not subject to U.S. export controls.
Another avenue available to them is to patiently await potential acquisition opportunities. Li envisions that a convergence of factors, encompassing the scarcity of advanced chips, soaring demand for data and AI expertise, and the significant upfront investments required, will propel the industry toward a forthcoming "consolidation stage."
Related Article: US Broadens Export Controls Targeting Chip Transfers to China
