It is difficult to go into detail about the number of dances in which the United States and China are immersed at the moment. The technological war is the umbrella under which the trade warthe attempt of military conflict in the South China Seathe robot racethat of the energy and that of artificial intelligence. Everything is related to each other and, although the United States has exercised an aggressive technological blockadeStanford University is clear that it has not been of much use.
And they are clear that the AI gap between the powers is “practically closed.”
The report. When a new model or version of AI is presented, those responsible show graphs and tables in which they comment on how good their product is. It is something that always has to be taken with a grain of salt because the idea is to make your product look good – it would be necessary to do more – and, for this reason, an external analysis is needed to show us the complete photo. In this sense, the annual report from Stanford University (in its ninth edition) is one of the best thermometers for taking temperature in the state of AI.
One of the conclusions of study is that the Chinese models are very close to the American ones. If at the beginning of the AI boom those from the US set the tone with an abysmal difference, at the beginning of 2025 the distance was greatly reduced to the point that DeepSeek-R1 equaled the best American models on several occasions. Since then, the absolute top model is that of Anthropic, but currently only with a 2.7% advantage over the best Chinese model.
Different approaches. In fact, in the graph with this specific analysis you can see how the distance between the two is closed as the performance of the Chinese models increases exponentially in a very short time.
And something that the study highlights is that, although the United States continues to lead the battle because it is the one that produces the most top-level AI models and with the most important patents, China leads in volume of both models and production of those patents. Also in other sectors, such as AI in robotics, for example.
sticks in the wheels. And the most notable thing is that China has achieved this evolution without having the best tools. As a result of the technological and trade war, it is known that the United States has done everything in its power to prevent cutting-edge technology will reach the hands of the Chinese industry. For years they prohibited American companies (which are the ones that control the AI leader like NVIDIA or AMD) sold their higher-end platforms to Chinese companies, but also They shorted the European ASML and the South Korean Samsung and SK Hynix.
Because the US has the aforementioned NVIDIA and Intel, but ASML is the one that manufactures the most advanced machines for making chips, Samsung is one of the world’s leading foundries and leader in high bandwidth memory along with SK Hynix and then there is the Taiwanese TSMC as the largest foundry on the market. Although the US has more recently shaken hands with China in this regard, there are still restrictions on Chinese companies accessing the latest technology.
Counterproductive. However, through innovation, government support and a little bit of smuggling, gray market and reverse engineering, companies like SMIC -the Chinese foundry- or Huawei They have managed to develop their advanced equipment and chips. The US has tried to put all the pieces in the wheels of the Chinese industry, but as some reputable voices in the chip sector have pointed out, this has only served for China to directly and advances its technological sovereignty program.
That is to say, the vetoes that had such a hard impact at the beginning have served to light the flame of technological development. Huawei is the best example of this, since was ostracized five years ago and recently showed that not only has he recovered, but he has returned in better shape than ever, even becoming one of the main drivers of AI for Chinese industry.
Approach. Something that the Stanford AI study also highlights is how the two countries are approaching this segment of AI. And we talk about money, of course. While private investment in AI in the United States reached almost $286 billion, in Europe The investment was almost 21,000 million and in China it was only 12,400 million.
This is tricky, since it involves private financing (and this year among just a handful of American companies 650,000 million dollars will be melted) and the state support from the Chinese government should not be underestimated, but beyond investment, American companies have focused on creating the most powerful models regardless of the price while the Chinese approach is to make a Cheap AI to be almost transparent to the user.
The goal in both cases is mass adoption, but here the cheaper the product and better integrated into everyday platformsbetter.
Taiwan. There are other adjacent topics. For example, China has the energy for the AI erabut The US has the data centers. According to the report, there are more than 5,400 data centers in the American country, which is more than ten times the amount that any other country has, but all this with a curious counterpoint: it is a Taiwanese company that manufactures almost all of its artificial intelligence chips: TSMC.
The company is expanding with foundries in the US, but although a conflict that will break those relationships is not in sight, it is evident that depending on a foreign country is not the best strategy for technological independence. That is why they are injecting a lot of money so that Intel be the great foundry, but the reality is that it is still very far from TSMC and, although the US is making attempts with native companies such as Applied Materials, the main partners continue to be from outside the country.
In Xataka | NVIDIA and AMD face an implacable enemy that they did not count on: the US bureaucracy
GIPHY App Key not set. Please check settings