The chip war

Chips are currently at the center of the geopolitical battle between Washington and Beijing. The latest chapter in this saga—on which much more depends than we imagine—has shown how far the US government is willing to go. But where does this dispute come from?

Semiconductors, or chips, have become one of humanity's most precious assets: they're everywhere. They're essential for everyday tasks, like what we do with our phones, but they're also key in security and defense. Just look at the wars in the Middle East and Ukraine, where access to chips is the main bottleneck for producing more and better drones. It's no coincidence that the world's largest company is Nvidia, with a market capitalization of $4.4 trillion. It began designing graphics cards for video games (GPUs), but over time, this design positioned itself at the center of the AI revolution. Today, data centers in the US, Asia, and Europe use thousands of these chips.

These data centers, large language models (LLM), and predictive algorithms aren't just used by university students to cheat: they're also used in government, security, and defense. GPUs are strategic because their architecture, consisting of thousands of cores, allows for the parallel processing of huge volumes of data, essential for AI, scientific modeling, military simulations, massive data analysis, and cryptography. Their use grew first in graphics and video games, then in cryptocurrency mining, and, above all, in training AI models since 2012, when they proved to be much faster and more efficient than CPUs.

Today, technological dominance and access to cutting-edge GPUs—controlled by a few manufacturers like Nvidia—are critical factors of economic and military power: they determine a country or corporation's ability to develop advanced AI, strengthen its defenses, innovate in biotechnology and energy, and compete at the technological frontier. Their supply is now a geopolitical asset comparable to oil in the 20th century.

Therefore, the US, under the Biden administration, has sought to limit China's access to next-generation GPUs. Since 2022, it has imposed increasing restrictions due to their strategic value in AI, defense, and supercomputing: in October of that year, it included chips such as the Nvidia A100/H100 and AMD MI250/MI300 on the Commerce Control List, demanding almost impossible licensing requirements; in 2023, it expanded the Entity List to block companies such as Inspur, Loongson, Biren, and Moore Threads; and in 2025, the Trump administration banned the sale of models adapted to evade restrictions, such as the Nvidia H20—specifically designed with limited capabilities—and the AMD MI308. It then proposed allowing their export by charging 25% of revenue for H20 and 15% for other chips, in addition to covert measures such as inserting trackers into servers with GPUs to detect misuse.

China, for its part, is trying to get its tech giants to reduce their hydrogen purchases and is trying to boost the development of local industry. However, while it has extensive experience replicating technologies, Nvidia's designs are practically unique, and replicating them will require time and significant investment. This is just one more front—perhaps one of the most critical—in the inevitable economic confrontation between the US and China. Everything indicates that the war over chips is just beginning. And as with oil in the 20th century, control over chips will determine winners and losers.