Connect with us

World

AI Is Pushing The World Towards An Energy Crisis

Published

on

AI Is Pushing The World Towards An Energy Crisis

The dramatic resignation of Ilya Sutskever, the chief scientist of OpenAI, which is behind Artificial Intelligence (AI) and large language models like OpenAI’s ChatGPT, has reinvigorated public debates on the future of AI and its exorbitant costs. Beyond the many acknowledged concerns, such as AI safety or the future of work and creativity, there is a trade-off that will be no less transformative. AI is one of the most energy-intensive modern IT undertakings. The world, concerned with carbon emission, may not be ready.

The data centers used to train and operate these models require vast amounts of electricity. GPT-4, for example, required over 50 gigawatt-hours (GWh), approximately 0.02% of the electricity California generates in a year, and 50 times the amount it took to train GPT-3, the previous iteration. As AI proliferates across industries, this energy demand will only grow. When AI is used to optimize energy-intensive manufacturing, which will need experimentation and more data, this problem will only grow.

Data centers and their associated transmission networks have become a primary driver of global energy consumption. At present, this accounts for 3% of global consumption, emitting as much CO2 as Brazil. Increasing energy requirements show no sign of slowing down either, as consumption could grow from 460 Terawatt-hours (TWh) in 2022 to 1000 TWh in 2026. In the United States alone, the increase in power demand due to data center demand is expected to rise from 200 TWh as of 2022 to 260 TWh in 2026, equivalent to six percent of all power use across the country. Now, data centers’ energy demands are expected to double by 2030.

Further compounding difficulties is the geographic dispersion of data centers. Data centers, like many industries, cluster. The rapid speed of their construction combined with their ability to be economically viable almost anywhere means that new energy problems are rapidly emerging in unexpected locales. Northern Virginia has become the largest hub of data centers in the world. The region has 51 million square feet of dedicated space for data centers, and it consumes electricity equivalent to 800,000 homes. This strain on electricity distribution creates dangerous swings in power demand that threaten energy infrastructure.

As companies begin to utilize AI for more than just large language models, we can expect many individual firms’ electricity usage to increase. This does not simply drive more data centers to be built but for data centers to use more power as well. A rack of traditional servers in a data center runs on 7 kilowatts of electricity, while a rack of AI servers with increased processing power uses 30-100 kilowatts. Nvidia, the current leader in the AI server market, shipped 100,000 units last year expected to consume 7.3 times of energy annually.

Computational efficiency may follow Koomey’s Law, where the amount of energy necessary to do a set amount of computing falls by half every two and a half years, but data centers present their own obstacles even in this scenario. For example, as data processors become more efficient, they naturally run hotter. This requires additional energy and water resources that must be devoted to cooling systems, with Google
Google
and Microsoft
Microsoft
alone consuming 32 billion liters of water in their data centers in 2022.

To meet this level of consumption, the United States must carefully use its resources. Currently, large enterprises responsible for the proliferation of data centers, such as Microsoft, are attempting to offset their electricity usage by constructing wind and solar farms to provide them with power. While this is a good start, these sources of energy are intermittent, and data centers will still require energy to continue processing when the wind is not blowing and the sun is not shining.

Any solution to the usage problem posed by data centers must address the strain that will be put on the grid to which they are connected. The American electrical grid is already strained due to increased industrial production, buoyed by more manufacturing returning from overseas, and will not slacken. Electric vehicles, which are growing in popularity in the United States, necessitate heavy electricity use to charge. EVs are set to help push US domestic energy demand up by 38% by 2035, with half of all vehicles sold then being EVs. The risk these factors pose in conjunction has already shown its face, as Texas, a state rife with technology companies and data centers, has seen price spikes and blackouts partially as a result.

To enjoy ample, affordable, and consistent energy, the AI industry and utilities that serve it need to look toward abundant natural gas and the potential of nuclear energy to meet high energy demand, in addition to constructing solar and wind farms. AI data centers abroad may look towards electricity powered by U.S. liquified natural gas. Those closest to the issue agree. OpenAI CEO Sam Altman has expressed concerns about the capacity for current clean energy and battery technology to allow AI’s growth, stating that breakthroughs would be necessary to consider them as an option.

America can also update its outdated grid infrastructure to keep up. The construction of new transmission lines has decreased from 4,000 miles in 2013 to about 1,000 miles annually today, while the country’s industries and households need more electricity to operate. Extra energy generation capacity means very little when utilities cannot transmit an increased power supply.

AI represents the future of humanity and economics. If the United States is unable to accommodate its development, it will inevitably fall behind peer competitors, such as China. While this technology poses a new challenge to the country’s ability to generate and transmit energy, nuclear energy, natural gas, and renewables all have a part in developing the nation’s infrastructure. To do so, policymakers and business leaders must shift their focus from the short-term fixes and lower the barriers to using all the energy resources at the country’s disposal.

Continue Reading