Wed Apr 16

Trump’s trade war puts America’s AI ambitions at risk

Written by Albert Zomaya, Professor, School of Computer Science, University of Sydney

The global trade war triggered by US President Donald Trump earlier this month shows no signs of ending anytime soon. In recent days, China suspended exports ^[1] of a wide range of critical minerals that are vital ingredients in everything from electric cars and drones to the semiconductor chips that power artificial intelligence (AI) servers.

Around the same time, Trump also indicated he would soon impose new tariffs ^[2] on semiconductor chips.

All of this is happening at the same time the US is forging ahead ^[3] with a US$500 billion (A$784 billion) project known as “Stargate” to accelerate the development of AI in the country.

But the escalating trade war does not square with America’s ambitious AI plans. In fact, Trump’s tariffs (which, in the case of China, now total 145%^[4]) are set to undermine these plans by increasing the cost of AI development and disrupting supply chains for AI goods.

In turn, this will hinder the pace of AI innovation and adoption in the US – and potentially elsewhere.

Inflating the prices of essential components

AI development requires significant computational power and specialised hardware such as high-end graphic processing units ^[5] (GPUs), which are predominantly manufactured in Taiwan and South Korea and often assembled in China ^[6].

US tariffs will directly inflate the prices of these essential components. One analysis ^[7] estimates tariffs could increase the material costs of data centre building by around 20%, with IT hardware components potentially rising by 25%.

This is a major concern for AI industry leaders such as OpenAI, which operates ChatGPT. For example, the company’s chief executive, Sam Altman, recently said ^[8] his team is “working around-the-clock” to determine how the trade war would affect the cost of running their AI models.

But the increased cost on AI development caused by the trade war will also mean tech startups in the US will have higher barriers to entry and fewer opportunities to test AI capabilities. In turn, this will harm AI innovation.

In theory, tariffs might support the reshoring of chip production in the US through initiatives such as the CHIPS and Science Act ^[9], which promotes domestic US semiconductor production. But it would take years for such efforts to fully bear fruit. And Trump has also recently taken steps to walk away ^[10] from the CHIPS and Science Act.

Aggressive AI nationalism

The trade war also creates risks for the international development of AI.

For example, the cost increases that flow from tariffs could create a reluctance to invest in AI infrastructure – particularly data centres. Other tech companies might also cancel or delay plans ^[11] to build data centres in the US partly because of higher equipment prices.

In addition, tariffs could push countries into further fortifying their AI efforts, creating a kind of aggressive AI nationalism. They could also encourage domestic AI development to promote national interests ^[12]. This could lead to isolationism and put another nail in the coffin of the open-source culture that once fuelled AI innovation.

Tariffs are supposed to promote domestic industries. But high costs and a fracturing of the cooperation that is indispensable to the continuation of the AI landscape might well be the outcome.

A corridor with server Racks and cabinets full of hard drives

One analysis estimates US tariffs could increase the material costs of data centre building by around 20%. IM Imagery/Shutterstock ^[13]

Knock-on effects for Australia

Australia is not the direct target of most US tariffs. But the tariffs on advanced technologies and critical components pose risks to its ability to develop AI.

Although Australia aims to bolster its domestic AI capabilities, it currently relies heavily ^[14] on imported hardware for AI development. Tariffs will likely make it more expensive for Australian companies and research institutions to acquire the necessary infrastructure, such as semiconductors, GPUs, and cloud computing equipment. In turn, this will potentially hinder their technological progress.

As the US clamps down on trade and technologies, Australia may find itself locked out of international research projects, perhaps those involving US companies or technologies.

Such limits on data sharing, international cross-border AI talent, and cloud infrastructure risk slowing the rate of innovation.

To mitigate the above risks, Australia must invest more in developing domestic AI capacity and diversifying its technological partnerships ^[15].