China is helping its artificial intelligence (AI) start-ups facing challenges due to a shortage of computer chips. Large tech companies in China are using most of the available AI-training computing resources, leaving smaller start-ups struggling.
Seventeen city governments, including Shanghai, have promised to offer “computing vouchers” to assist these AI start-ups. These vouchers, valued between $140,000 and $280,000, can be used to cover the costs of using AI data centers for training and running large language models (LLMs) that understand and generate natural language.
This initiative follows a situation where internet companies with cloud computing services canceled contracts due to stricter US controls. This led them to prioritize their own use of graphics processing units (GPUs), causing difficulties for AI start-ups.
Major tech companies like Alibaba, Tencent, and ByteDance have limited the rental of Nvidia’s GPUs, reserving them for internal use and important clients. Over the past two years, US restrictions on AI chips have made it challenging for Chinese companies, forcing them to stockpile, repurpose gaming chips, or resort to the black market.
To address the issue, at least 17 city governments are providing subsidies through computing vouchers. However, analysts believe this won’t completely solve the problem, as the scarcity of resources remains a challenge.
Beijing is planning to introduce a subsidy program for AI groups using domestic chips, aiming to reduce reliance on foreign components. Additionally, China is building an alternative to Big Tech’s data centers, creating a network of state-run data centers and online platforms where AI companies can rent computing power.
While the vouchers will reduce computing costs for start-ups by 40 to 50 percent in government-run data centers, there are still strict requirements for applicants. This includes a minimum revenue threshold or involvement in a government-sponsored research project.
China is actively promoting the adoption of AI technology, centralizing the distribution of computing power through state-run trading platforms. The country aims to enhance resource allocation efficiency and lower energy consumption while closely overseeing AI applications.
Despite these efforts, there are traces of an underground chip trade on brokering sites partnering with companies like AliCloud and Tencent Cloud. This reveals the ongoing challenges and strategies in China’s AI landscape.
Source: Ft
