China tech - Throwing down the gauntlet to the US
DeepSeek’s advances are emblematic of broader progress in China’s tech sector. This is allowing some Chinese companies to challenge the dominance of their US rivals, and it holds out the promise of AI being deployed across other industries – such as manufacturing – to enhance efficiency and productivity.
The significance of DeepSeek is that it not only shows that China has the ability to develop Large language models (LLMs) that are equally, if not more, capable compared to their US peers, but that it also has done so with far fewer hardware resources, through innovation in algorithm. This means significantly lower cost of training and using LLMs, deep learning models that are pre-trained on vast amounts of data.
China could scale innovation across other industries, for example in the manufacturing sector with so-called “smart factories”.
Since the launch of DeepSeek, some Chinese models have gained market share from US leaders, especially outside the US.
The lower cost of LLMs by companies like DeepSeek opens the door for much faster development of AI applications and could also benefit China’s domestic chip makers as these models are less demanding for computing power and can utilise China’s self-made AI chips, which are still less powerful than those offered by some US leading companies.
Additionally, lower inference costs may accelerate AI application development, positively impacting internet and software companies.
Scaling innovation
Beyond the tech sector, China could scale innovation across other industries, for example in the manufacturing sector with so-called “smart factories”.
Chinese manufacturers will have strong incentives to adopt AI. The level of competition in Chinese manufacturing probably is the highest in the world, partly due to over-capacity in many sectors. Companies are constantly facing pressure to upgrade their productivity to stay in the game. As a result, China is already the world’s largest market for industrial robots. When it comes to AI, as long as the technology can prove to enhance efficiency and productivity, Chinese companies will have strong incentives to incorporate them.
And China has the world’s largest talent pool of engineers, which gives it a strong competitive edge for applying AI to various industrial areas.
The cost of AI models will likely keep dropping, and the profit margins of the AI model creators will likely decline as well, even without the participation of Chinese companies. This is because computing power will always get cheaper and cheaper (as has been the case in recent decades) due to Moore’s Law, which states that the number of transistors on a microchip doubles about every two years with a minimal cost increase. Chinese companies’ participation in this trend will likely just make the process much faster.
In the future, AI could become part of IT infrastructure, just like the internet today. The profit margin of internet service per se is very thin today. But based on this infrastructure, an almost unlimited number of new services, products and other innovations have appeared, which could be very profitable. AI could follow the same model. This is to say, the companies that can make big profits will likely be those who provide the new applications.
In the near term, China’s trade conflict with the US could add more headwinds to China’s tech sector due to the tightened tech restrictions by the US, even if the two economic giants are working to de-escalate the conflict. However, these restrictions will also likely accelerate the development of China’s domestic semiconductor industry, which is trying to catch up with its US peers. So, longer term, the restrictions may not be a bad thing for China.