Full speed ahead: China’s race to implement AI

China's tech sector is experiencing a profound transformation driven by aggressive investments in artificial intelligence (AI). Notably, the regulatory landscape in China has shifted to become increasingly supportive of large language model (LLM) development, marking a significant change from previous policy.

Currently, there are more than 300 foundational models in development, although some are encountering challenges and discontinuation. A subset of these models is on a trajectory to advance from generative pre-trained transformer (GPT) 3.0 capabilities to 3.5, with ambitious plans to reach 4.0 levels within the next year. However, this bold timeline may face hurdles due to hardware and knowledge constraints. The current scenario involves a tight supply of graphics processing units (GPUs) with lengthy waiting times. This GPU scarcity could pose a distinct challenge for Chinese firms compared to their Western counterparts. Additionally, further sanctions could exacerbate this situation. Nevertheless, in the United States, there is a shift away from pursuing larger models, with a focus on refining data and fine-tuning techniques instead of merely increasing model parameters.

The likely beneficiaries in China's LLM landscape are expected to be domestic companies, given that most foreign LLMs may not secure authorization to operate within China. Concerns also loom over foreign open-source models used in China, particularly regarding the data on which these Western models are trained, raising uncertainty about their continued use in the country.

In the evolving LLM space, only a select few models are anticipated to thrive, with scale emerging as a critical advantage. Leading Chinese tech giants are at the forefront of LLM development, and there's a growing consensus on the importance of "neutral" models not affiliated with large internet companies. Intriguingly, the market does not seem to assign substantial value to LLMs for these publicly traded Chinese tech giants. Presently, numerous Chinese tech firms are demonstrating robust growth, expanding margins, and returning value to shareholders. Shareholder return has become a top priority, reflecting the transition from hypergrowth to slower yet robust growth with increased free cash flows.

In terms of infrastructure, prominent tech companies dominate the hyperscale data center sector, partly due to GPU scarcity in China. However, balancing internal GPU demands with external revenue opportunities presents sustainability challenges.

In the application layer, Chinese internet and software firms are enthusiastic about the potential of generative AI. Higher productivity is likely to be the most critical use case for LLMs in the near future. Leading internet companies equipped with AI consumption scenarios and proprietary data, along with early tech adopters, are well positioned to harness generative AI tools for productivity gains. Caution is warranted, though, as stocks associated with the "AI concept" in the A-share market appear overvalued. Monetizing LLMs for enterprises is expected to be a gradual process and fraught with difficulties. Currently, most companies are flexible in using multiple LLMs, favoring cost-effective solutions. Traditional industries may take longer to recognize the value of LLMs.

In summary, China's AI landscape is dynamic and evolving rapidly. While challenges and uncertainties persist, China's commitment to AI innovation, coupled with its supportive regulatory environment, positions itself as a formidable player in the global AI arena. Chinese tech giants, armed with their scale and resources, are poised to lead in the development and application of LLMs, potentially reshaping industries and driving innovation in the years ahead.