Beijing, we have a problem. It is a lot that the Asian giant has achieved in recent years to compete from you to you with the US in the field of AI, but is losing a crucial battle: that of investment. There his rival was already overcoming him before. Now he is crushing him.
Investment gap. Winning in AI means doing it in many areas. China has managed to overcome many obstacles and is competing with the US in the ability of its models. He is even starting to have really promising chips that can put things to Nvidia. However, China has a big problem in the field of investment, because its companies do not invest as much as the North Americans do.
The US capex is astronomical. In the last five years Google, Microsoft, Meta and Amazon have accumulated a capex of 5.36 billion yuan. Meanwhile, the seven large Chinese technology companies (Tencent, Alibaba, Baidu, JD.com, Kuaishou – one, do not include Bytedance -, meituan, and netease) invested a total of 630,000 million yuan.
The difference is spectacular, but it has been increasing over time. According to a Jinduan Research Institute report, in 2020 the US capex ratio with China’s was 1: 6, but at present that ratio is approaching at 1:10. It is not so much that China does not invest: it is that the US invests much more.
Fuente: Jinduan Research Institute
Data centers send. Although not all the capex of these technological ones is intended for AI, the majority of that capital expenditure is certainly focused on this area. In fact, we have already seen how the great US technology have announced multimillion -dollar investments in data centers. From the 100,000 million dollars that Amazon intends to invest at 65,000 million dollars, the figures are absolutely dizzy.
RED EFFECT. To that problem is added that of the network effect that is causing US investment. This effect occurs when a product or service becomes more valuable as more people use it. The US investment allows better models to develop and attract more users that generate more data. These data “feedback” AI and improve models, which in turn makes more users use it, and thus in an infinite loop.
The quality of the Chinese models in front of those of the US is remarkable, at least according to some of the most popular benchmarks in the market. Source: Artificial Analysis.
Adoption rate. The problem, they indicate in the Jinduan report, is in the adoption of the AI by the US, which is also supposedly far superior to that of China. According to its data, the AI adoption rate by companies in the US is 78%, while in China that figure does not go from 15%.
The first data comes from a study by McKinsey consultant, while it is not clear where China comes from. The same goes for the number of active weekly users of chatbots of AI. In the US, 1,000 million are exceeded – only Chatgpt already has about 700 million – but in China that figure seems to be only 70 million according to the study, a figure that a priori seems doubtful. The Questmobile consultant revealed that last November the number of active APPS monthly users in China exceeded 100 million users.
China (probably) would like to spend more, but can’t. Although it is not clear if the adoption rate causes that minor capex or is the other way around, but what is certain is that Chinese companies would probably want to increase their capex to bet even stronger for AI. The problem is that they cannot due to the export controls imposed by the commercial war between the US and China. If the United States vetoes the export of advanced chips and components from AI to China, those companies simply cannot dedicate more money to Buy them.
Dividends. The authors of the Jinduan report point to another reason for that difference in Chinese Capex-Eeu. Chinese companies are using the benefits obtained to repurchase actions and offer dividends instead of dedicating them to CAPEX. According to this report “in 2024, the total net amount of the repurchases of shares, dividends and debt amortizations of Tencent reached 1.68 billion RMB, more than double its capital expenditure for that year.”
Thus, the restrictions imposed by the US are only part of the problem. The “deflation of AI” China seems to be due to a certain inaction by these companies, the report points out, and that can end up causing a big problem, especially in the long term.
Imagen Karolina Grabowska
In WorldOfSoftware | The infrastructure boom for AI begins to show cracks: China accumulates unreasonable data centers, and is not the only one
