Advantage of Chinese AI in large language models versus the US
Technology specialist Yin Ruizhi takes a look at why China has been able to lower the cost of its large language models compared with the US.
My previous articles looked into the disadvantages that Chinese artificial intelligence (AI) faces in the field of large language models (LLMs) when compared with the US. Today, we will discuss the oft-overlooked advantages that China has against the US.
Making profits even at low prices
In early May, an advanced LLM called DeepSeek-V2 emerged in China. It achieved excellent performance, nearing the forefront of international standards in core AI technologies.
What shocked the industry was the model’s cost-effectiveness, requiring only 1 RMB (US$0.14) per million input tokens and 2 RMB per million output tokens. This is astonishing compared with GPT-4, which costs 100 times more. Even the budget Chinese AI model Moonshot AI costs over 20 times more.
With such a huge price difference, a large number of Chinese AI product suppliers are turning to DeepSeek. Many media platforms initially speculated that DeepSeek might be adopting a subsidised strategy to compete for market share. But DeepSeek’s team said that the company is still able to make a profit at such a price point and the cost to process one million tokens is just 0.50 RMB — the rest is profit. Even more shocking is the fact that DeepSeek is an open-source coding model — anyone can build their own model based on it.
... major innovations in its framework and the significant optimisation of its model architecture has allowed costs to be reduced to incredible levels.
Many in the tech community are curious about DeepSeek’s background. In fact, the renowned Chinese quantitative hedge fund High-Flyer Capital Management is behind DeepSeek. According to an open access thesis, major innovations in its framework and the significant optimisation of its model architecture has allowed costs to be reduced to incredible levels.
Price war
On 21 May, Chinese internet giants entered the price war. The first to be affected was Alibaba Cloud, which announced a significant price cut on its LLMs. This move immediately triggered a positive response from several well-known enterprises in the industry, provoking a subsequent wave of price cuts.
However, amid this intense price war, Alibaba Cloud’s price reduction strategy is starkly different from other vendors’ focus on adjusting the price of small language models. The former lowered the price of nine of its LLMs in one fell swoop, with its flagship model Qwen-Long — comparable to the internationally renowned GPT-4 — exemplifying the enterprise’s boldness. The application programming interface (API) input price of Qwen-Long has been reduced by a whopping 97%, from 0.02 RMB to 0.0005 RMB per 1,000 tokens.
With this price reduction, 2 million tokens can be bought with just 1 RMB, equivalent to the amount of text in five copies of the Xinhua Dictionary. This price is around the cost of DeepSeek that was announced in early May.
Even as OpenAI’s GPT-4o as well as Google’s Gemini 1.5 Pro have lowered the cost for API usage, Alibaba Cloud’s flagship LLM still has a market advantage that cannot be ignored when it comes to cost performance ratio.
As for Qwen-Long’s performance, information from the official website showed that it is close to GPT-4 in terms of function, able to support long, contextualised interactive dialogue of up to 10 million tokens, and can easily handle documents of roughly 15 million words or 15,000 pages, thus leading the pack when compared with similar LLMs.
Even as OpenAI’s GPT-4o as well as Google’s Gemini 1.5 Pro have lowered the cost for API usage, Alibaba Cloud’s flagship LLM still has a market advantage that cannot be ignored when it comes to cost performance ratio.
Engineers in China the advantage
How are Chinese AI firms able to create a product similar to one offered by their American counterparts at a tenth of the cost, and how did this price advantage come about?
The key lies in the large number of engineers in China, and the unique circumstances for the industry. China’s top software engineers are all focused on price optimisation of its AI LLMs, and China pumps in nearly ten times the number of talents for this purpose as compared with the US.
Firstly, there are around eight million software engineers in China — twice that of the US’s 4 million. Software engineer graduates from the top 20 universities in China have occupied a top three spot in terms of salary since 2008. Hence, there has been a constant influx of talent into the industry.
... for the few major AI platforms in China, they have assembled around ten times the talent compared with the US in the same field to work on cost-effectiveness...
Due to employment structure and a dearth of talent, most software engineers in the US would be able to stay in their jobs until they retire. In China, as the talent pool is sufficient, software engineers are mainly between the ages of 25 to 35 — at the same time they are subjected to an intense, overtime work culture. Very often, Chinese software engineers have double the workload or even more compared with their US counterparts.
Secondly, as mentioned in previous articles in this series, there are far fewer LLM-relevant entrepreneurial opportunities in China compared with the US. Thus, in comparison with their US counterparts, the top Chinese software engineers do not flock to startups and instead stay at Alibaba or a select few large platforms with financial backing. With the chip industry being restricted, their top priority is to reduce the cost of China’s AI LLMs.
In sum, for the few major AI platforms in China, they have assembled around ten times the talent compared with the US in the same field to work on cost-effectiveness — the efforts of Chinese software engineers have greatly reduced the cost for AI LLMs.