At the end of last month, during a speech at the Bosch ConnectedWorld conference, the head of Tesla and SpaceX, Elon Musk, said that already next year the global shortage of electricity will be holding back the development of artificial intelligence systems. The power supply system will no longer keep up with the rapid growth in the number of data centers.
Growing Energy Demand for AI Systems
According to Boston Consulting Group, cited by Barron’s, data centers will consume 7.5% of all electricity available in the United States by 2030. Data centers built for AI systems can consume hundreds of megawatts of electricity per facility. According to analysts, from 2022 to 2030, data center energy consumption in the United States will increase from 126 to 390 terawatts per hour, this amount of electricity would be enough to supply 40 million American households.
Mitigating Energy Consumption Challenges
According to 650 Group estimates, supply volumes of server systems for AI needs from last year to 2028 will increase sixfold to 6 million units. According to Gartner forecasts, the average power consumption of one computing accelerator in their composition will increase from the current 650 to 1000 W. In other words, energy consumption will increase not only due to an increase in the number of servers, but also in specific terms. Attempts to optimize energy costs through the introduction of artificial intelligence in the field of energy consumption control will somewhat curb this trend, but will not completely solve the problem. The transition to liquid cooling of server systems will be inevitable in most cases. Data center operating costs can be reduced by more than 40% by eliminating traditional air cooling in favor of liquid cooling, according to Super Micro.
Challenges in Energy Distribution
The problem is aggravated by the uneven development of regional energy supply networks. In addition, it is not possible everywhere to efficiently transmit generated electricity to the location of large consumers. Experts say the US generates enough electricity to support the development of artificial intelligence systems, but there is a problem with distribution networks, notes NIX Solutions.
We’ll keep you updated on the evolving strategies and technologies aimed at addressing these challenges.