The human and machine learning technologies available today are unprecedented and they’re becoming ubiquitous. This rapid development is rapidly overextending the world’s energy supplies and leading to a sharp increase in carbon emissions. At this time, demand for supercomputers and advanced processors is at an all-time high. By 2030, experts predict that the energy demands for AI may increase by as much as 170-fold! Accordingly, the increasingly substantial energy use raises deep issues about whether these next-generation technologies themselves can be developed in environmentally sustainable ways.
Yet the chipmaking sector is one of the top contributors to this carbon footprint. So far in 2023, it has emitted over 99,200 metric tons of CO2. By 2024, that number had jumped to 453,700 metric tons. However, the majority of this production occurs in other countries, particularly Taiwan, South Korea and Japan. These countries are dependent on fossil fuel-dependent electricity grids to power their manufacturing processes.
Power Demands of AI Development
The fast growing AI industry requires building more supercomputers and a vast amount of processors to address its rising demands for computing. The production capacity for chips and semiconductors has increased by four times over the last year. The ecological toll from this expansion cannot be overstated. All those steps collectively eat over 10 times as much energy as the chip itself, and they’re all almost exclusively powered by fossil fuels.
In fact, Taiwan, South Korea, and Japan are already the worst offenders worldwide on a per-chip basis of carbon emissions from semiconductor manufacturing. Taiwan’s chip industry emitted 185,700 metric tons of carbon emissions just in 2023. Japan and South Korea’s semiconductor industries were largely responsible for driving up pollution levels. This bleak reality highlights the real imperative for both nations to accelerate towards greener energy economies.
Fossil Fuel Reliance and Its Consequences
Fossil fuels drive a substantial share of America’s advanced manufacturing centers. This extreme dependence greatly contributes to increasing emissions. Taiwan’s power grid is heavily centralized and dominated by the fossil fuel industry. This dependence leads to low carbon emissions and presents deep energy security risks. In the same way, neighbors South Korea and Japan continue to struggle against their energy infrastructures deeply rooted in non-renewable resources.
If the AI industry wants to chase new ideas and do big things, it has to face the environmental impact of its expansion. The increasing demand for energy-intensive technologies will only increase the already substantial carbon footprint of these countries. Real, far-reaching reforms need to be adopted to avert this detrimental effect.
Future Outlook and Sustainability Challenges
The anticipated growth in global energy needs for AI only adds further complexity to the narrative of AI’s sustainability promise. By 2030, the energy consumption by AI could increase fivefold, further stressing power grids that are already overloaded. Experts call on the chipmaking industry to make a clear pivot to renewable energy and eco-friendly production methods. Otherwise, the environmental impact of AI will quickly outpace its potential benefits.
