AI and Geopolitics

By Treston Wheat, Adjunct Professor, Georgetown University

Technology corporations focused on artificial intelligence (or other businesses wanting to utilize AI) will have to contend with severe geopolitical issues around developing the requisite computing power. While coding obviously matters for AI development, corporations still need data centers and semiconductors to be effective, but those areas have significant problems. Both data centers and the semiconductor industry have major geopolitical issues. Therefore, businesses need to understand them to operate more effectively.

Data Centers

The two biggest geopolitical issues surrounding data centers are storage and natural resources. To start, where will a country require corporations to store their data? That’s a critical question for the development and future of AI, and countries will compete with each other based on data. Data localization is one of the most contentious issues, which matters for AI because data is one of three fundamental parts of developing programs. As a Washington Post article put it, “Chinese companies have been investing in AI for years…and have long been at the cutting edge of surveillance technology.” Their surveillance gives them a strong data advantage.

Different countries and regions approach the issue differently, though, such as the EU wanting to protect the “privacy” of citizens while the CCP wants to have access to spying. For example, China’s cybersecurity law (2017) and data security law (2021) require that companies operating in the country keep the data there where it can be accessed by the CCP. The physical location of data centers (and access to them) will be an important part of AI competition, and the US has attempted to limit China’s access to data and computational power by restricting access to cloud services.

AI is dependent upon computational power, which means AI is dependent upon semiconductors (colloquially referred to as “chips”).

Another aspect of data centers connected to geopolitics will be the water requirements to keep the facilities running. Water is critical for the data centers used to compute AI training because of the cooling process needed to maintain the computers, using up to 5 million gallons of water per day. Natural resources can be a zero-sum game, particularly when competing powers are unwilling to negotiate or compromise. Even when there is no direct competition, resources can be problematic to acquire, and countries might face limitations when they cannot acquire those resources. Globally, almost one-third of data centers are in the United States, and currently, data centers account for about 2% of energy consumption in the country. Within the United States, approximately one-fifth of data centers are located in water-stressed regions because those areas have better access to solar and wind energy, improving efficiency. The US’s failure to work on desalinization has exacerbated the problems in the western part of the country. In the US, there are already political risks from activists, conservationists, and local governments because of a strain on water access.

The Middle East and Africa are also going to have significant problems when it comes to data centers and water. According to the 2022 report “Making AI Less ‘Thirsty,’” OpenAI’s ChatGPT requires approximately 1.3 gallons of water for every 200-500 questions and answers. More advanced models will require even more water. The report also found that “training GPT-3 in Microsoft’s state-of-the-art U.S. data centers can directly consume 700,000 liters of clean freshwater… and the water consumption would have been tripled if training were done in Microsoft’s Asian data centers…” Water consumption for data centers will also impact the future economics of the region. World Bank estimates find that climate-related water scarcity could reduce the GDP of the Middle East by 6-14% by 2050. All of this together will produce geopolitical problems that will need to be assessed and confronted when analyzing how AI will impact power dynamics.


AI is dependent upon computational power, which means AI is dependent upon semiconductors (colloquially referred to as “chips”). The design and manufacturing of chips are consistently threatened by geopolitical risks and highlight the deep physical-cyber connection that exists within AI. Producing semiconductors is severely limited in the number of countries that have the ability to design, develop, and manufacture chips. Even with the limited number of countries involved, producing chips is a complicated endeavor that involves tens of thousands of parts and resources. Resources like silicon and gallium arsenide are directly part of the production and, therefore, part of the geopolitical calculations.

While researchers in the United States lead the world in advanced chip design, countries like the Netherlands, South Korea, Japan, and Taiwan are critical for the manufacturing of them. For example, advanced chips use extreme ultra-violet light lithography for production, but the Netherlands’ ASML is the only maker of that equipment. Russia’s invasion of Ukraine threatened this process because of a neon gas shortage. Neon is needed for the lasers to carve patterns on the chips, and Ukraine produced approximately half the world’s neon before the invasion.

Other geopolitical or physical security problems include the trading routes, natural resources, manufacturing facilities, etc., necessary for semiconductors, and they are all threatened by great power politics and conflict. Also, TSMC is the most important manufacturer of advanced chips, and China’s threats to Taiwan could cause the global collapse of technology. Great power competition is already threatening the making of advanced semiconductors. In 2022, the Biden administration restricted the sale to China of advanced chips that are made with US equipment, even if they were produced outside America. This was to prevent the Chinese military from gaining technological advantages. China has responded by restricting raw materials, such as gallium and germanium products, in the process. Importantly, China controls about 13% of this market, East Asia controls a total of 57%, and the US controls only 11% of the market. Whoever controls the natural resources can influence the future of AI.

AI, Geopolitics, and Business

Understanding these critical issues will help technology companies navigate the geopolitical risks and plan for the long term. There are a myriad of geopolitical problems, but the specific issues around data centers and semiconductors will determine what AI capabilities corporations will be able to develop and use over the coming decades.

Bio: Dr. Treston Wheat is an intelligence research specialist with a geopolitical risk consultancy, editor of a security journal, and adjunct professor at Georgetown University.