In This Article:
China's technology industry is scrambling for advanced chips to support its artificial intelligence (AI) ambitions amid growing interest around ChatGPT, but the soaring demand, combined with US-led trade curbs, has limited supplies, according to Chinese media reports.
The country's AI sector is being "choked" by a shortage of graphics processing units (GPUs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs) and accelerator chips, said an opinion piece published on Tuesday by the Economic Daily, a newspaper founded by the State Council, China's cabinet, and supervised by the Chinese Communist Party's propaganda department.
GPUs, in which American developer Nvidia holds a dominant market position, are the most common type of semiconductor devices used for training large language models (LLMs), which support OpenAI's ChatGPT and similar services.
Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.
FPGAs and ASICs, in turn, allow trained LLMs to be deployed in the real world with lower power consumption.
"[Developing] industrial applications of LLMs is a key initiative to enhance national competitiveness," said the article, which called for more government and capital support for home-grown advanced semiconductors.
China's rush for chips comes as Big Tech companies and start-ups alike are chasing after advancement in generative AI - algorithms that can understand prompts from humans and generate sophisticated responses.
As of last month, the country had at least 79 LLMs with more than 1 billion parameters, according to the state-affiliated Institute of Scientific and Technical Information of China, each of which require hundreds if not thousands of advanced chips to train.
Sam Altman, CEO of ChatGPT maker OpenAI. Photo: Reuters alt=Sam Altman, CEO of ChatGPT maker OpenAI. Photo: Reuters>
Deep-pocketed tech giants are leading in spending. TikTok owner ByteDance, which is said to be testing an AI chatbot internally, has ordered more than US$1 billion worth of GPUs from Nvidia so far this year, according to Chinese media LatePost.
Baidu, whose chatbot is ranked No 1 domestically by a state-run think tank, procured more than 10,000 Nvidia GPUs during the same period, matching Google's demand, the report said.
For comparison, OpenAI's LLMs have been trained on a supercomputer, unveiled in 2020 by Microsoft, that was powered by 10,000 Nvidia GPUs. It costs OpenAI up to US$700,000 a day to run ChatGPT because of "expensive servers", according to a report by technology news site The Information in April.