It is a 100,000 m2 large, energy-consuming, grey box, that stores innumerate computers. What is it? If your guess was a data centre, you are right. While they might appear unassuming, they are becoming the new geopolitical focal point around the world.
Firstly, to understand what a data centre does, we need to imagine a collection of servers organised in endless corridors, with cooling systems and backup generators. They host, process and distribute websites, facilitate cloud backups, train and run AI. In other words those who control them, control the epicentres of the global data network.
Understandably, with the breakthrough of AI, the data question became ever more pressing. This is mainly because no other general-purpose technology burst into common usage at such a speed before. It took electricity and the steam engine 50 or more years to be widely used, whereas AI completed this trajectory in a few years. AI is implemented in essentially all fields, like medicine, government, education and the list goes on. Despite the initial concerns with AI, it now seems unlikely that anything will get in the way of its ascent. The problem with this astronomic rise is that AI-generated content requires an enormous amount of energy, 10 times more than an average Google search. This is placing significant stress on global resources which could potentially trigger further conflicts between the Global North and South.
The challenge now is to build as many data centres as possible, in order for countries to follow the AI revolution. However this means finding the resources and the capital needed for infrastructure, both of which are scarce in today’s world. Winning the data center war is one of the key objectives of the upcoming decades.
Resources wanted
This gargantuan task requires natural resources that we currently cannot afford to have, such as oil or gas. According to Goldman Sachs, it is estimated that carbon emissions of data centers will double in 8 years time. Furthermore, the rise in energy demand will cause 40% of data centers to be overwhelmed by 2027. It is clear that the energy transition is unlikely to cope with this exponential demand. Consequently, the risk is likely that a fossil fuel renaissance could take place. The noble goals of the green transition may be thrown out the window, if control over data is at stake.
States with greater access to fossil fuels could gain the upper hand at supporting data centers’ energy consumption. While countries like China and India seem to be reluctant to adhere to climate goals, the incoming Trump administration has openly declared that it would start fracking on US territory. This is certainly bad news for the developing world, who bear the brunt of climate change and for the welfare of the globe.
Furthermore, while certain innovations have made AI chips more energy-efficient, silicon, the main ingredient to semiconductors, is reaching its technological limits. The intensity of AI queries is causing traditional silicon chips to overheat and to break down. This raises a second resource-related question, namely what kind of materials can facilitate the increasing rate of data demand. The success is dependent on two factors, being more heat-resistant and more energy-efficient. Currently, the race to develop the new, improved chips is still on, however graphene is showing promising results. Graphene coated silicon chips are said to tick all the boxes that silicon chips on their own cannot withstand.
The catch is however, that the reserves of graphene are limited to developing countries like China, Brazil and Mozambique. This could certainly pose supply issues for Western states, particularly with the looming trade wars foreshadowed by the Trump administration’s tariff proposals and the EU’s inclination towards more protectionism. Access to precious metals could unlock a country’s potential to gain leverage over the semiconductor industry, and the global data network. But it also promises a potentially hostile global trading climate.
Gridlocks and global partners
Apart from the race for natural resources, the question of where and how to build data centers is also of strategic importance. Building infrastructure and upgrading power grids is crucial to facilitate the AI revolution. However, highly urbanised countries in Europe have an almost completely occupied electrical grid, not to mention that the average age of both North American and European grids is often two times that of China’s. New grids could take up to 10 years to construct, potentially blocking AI innovation in the West.
To resolve the ‘bottleneck', Foreign Policy suggests that countries like the US need to consider other partners where data centers can be reliably and safely built. This does not only require financial but also security guarantees, narrowing down the scope of countries that may be eligible for this. These ‘partner’ states will need to have cutting edge infrastructure to support data centers, availability of natural resources to guarantee energy and mineral supply and would have to align with the geopolitical ambitions of the US. The question is can the US make ‘partners’ without coercion, so that we do not see an amplification of neocolonialism to facilitate the AI revolution.
Essentially, the AI revolution is opening up pressing questions that if not addressed with adequate care could jeopardise international relations, as well as our planet’s health. In an increasingly turbulent world, the data center wars could deepen the North/South divide. From the acquisition of mineral and natural resources, to the establishment of infrastructure, countries may focus overwhelmingly on competition and chauvinism. However, winning the data center wars should not be done at all costs.
Comments