Artificial Intelligence: Economic Revolution or Energy Sink?
Artificial intelligence (AI) has established itself as an indispensable technology, not least in energy and building management, where it helps to improve sustainability and efficiency. However, the proliferation of AI also has a downside: It's driving up energy consumption in data centers. According to the International Energy Agency, data centers already account for about one percent of global electricity consumption, and this figure could double by 2030. This raises important questions about the sustainability and efficiency of AI applications.
#Artificial Intelligence
The energy consumption of AI can vary greatly and is difficult to measure. However, there are concrete examples that demonstrate that making AI more energy efficient is crucial for the future. Training AI models is extremely energy-intensive and consumes much more power than conventional data center activities. For example, a ChatGPT query consumes ten times more energy than a regular Google query, and researchers estimate that the creation of GPT-3 consumed almost 1300 megawatt hours of electricity and generated 552 tons of CO2. The same amount of energy could power around 105,000 average households in Switzerland for 24 hours.
It is well known that AI makes an important contribution to energy management and sustainability - but also to energy consumption. Measures are therefore needed to counteract this. These include sustainable AI practices, innovative solutions for data centers and the use of renewable energy sources to power AI technologies.
Energy demand becomes a bottleneck
Experts predict that by 2028, AI will consume about 20 percent of data center electricity. According to a forecast by the International Energy Agency (IEA), this demand will already amount to around 1,000 terawatt hours (TWh) in two years. Technological efficiencies will partially offset this enormous energy consumption. However, AI and the rapid growth of other workloads (think cloud) are driving a steady and inexorable increase in energy consumption. According to Meta CEO Mark Zuckerberg, this fact means that energy constraints are the biggest obstacle to the expansion of AI data centers. Such constraints have hindered the expansion of Meta's own data centers. According to Zuckerberg, energy, not computing power, is the biggest bottleneck to the advancement of AI.
AI cools data centers
AI offers both a challenge and an opportunity for the energy industry. Although it increases demand, in some cases massively, it also has the potential to improve efficiency in various sectors and thus offset at least part of the energy demand. Numerous companies are working on solutions to reduce energy consumption. DeepMind, an artificial intelligence company operated by Google, has developed a machine learning system that optimizes the cooling of data centers. The system uses real-time data from sensors in the data center to monitor temperature and humidity and adjust cooling systems accordingly. The DeepMind system recognizes complex patterns in the sensor data. These patterns enable the system to predict when demand for cooling will rise or fall and adjust the cooling systems accordingly.
The implementation of the DeepMind system in a Google data center in Singapore led to a 40 percent reduction in energy consumption for cooling. This translates into millions of dollars in annual savings. Another important approach is the development of energy-efficient hardware. NVIDIA, a leading manufacturer of graphics processing units (GPUs), has made tremendous strides in this area. The latest GPUs from NVIDIA are up to 25 times more energy efficient than previous generation models.[i] This means that they can deliver the same performance with a fraction of the energy consumption. NVIDIA achieves these energy savings through a number of innovative techniques, including AI-powered optimization. This allows the performance and energy consumption of the chip to be adapted to the respective requirements in real time.
From boom to more ecological integrity
The AI boom of recent months and years has contributed to the emergence of a sustainable AI movement focused on promoting AI products towards greater environmental integrity.[ii] Overall, the movement is an informal joint effort. Researchers, companies, policy makers, non-profit organizations and individuals are participating. By implementing concrete measures, AI can actually be made more sustainable and at the same time be used as a powerful tool to promote sustainability efforts. For example:
- In smart grid management, AI is used to optimize energy distribution in power grids, reduce energy waste and improve the integration of renewable energy sources.
- In sustainable building design, AI is used to design buildings with higher energy efficiency, use natural light and ventilation, and optimize heating and cooling systems.
- When it comes to efficient training of AI, techniques are being developed to train AI models with less data and fewer computing resources.
- Model sharing and model reuse will promote the sharing and reuse of already trained AI models in order to avoid redundant training measures and the associated immense energy consumption.
AI can drive sustainability
AI can accelerate progress towards sustainability.[iii] Companies of all industries and sizes are increasingly recognizing the potential of AI to drive their sustainability efforts while improving their profitability. For example, AI-powered systems can collect and analyze vast amounts of data from sensors, satellite imagery and other sources to monitor changes in climate, environmental conditions and resource use. These insights can help to identify environmental problems at an early stage, understand trends and make informed decisions. AI can be used in various sectors to optimize processes and improve resource efficiency. For example, AI can be used in agriculture to optimize irrigation, fertilization and pest control to increase yield and reduce negative impacts.
UMB, Equinix, and KI
UMB works closely with Equinix[iv], one of the leading companies in the field of data center and enterprise network colocation.[v] Equinix operates data centers that enable companies to securely store and distribute data and is becoming increasingly important as a location for the provision of AI infrastructures. Equinix's cloud proximity, ecosystem connectivity, and extensive global reach make it an ideal platform for private AI deployments, benefiting both enterprises and service providers. Private AI provides risk protection while delivering the benefits of trained AI models, enabling organizations to realize the full potential of AI.
UMB AG and Equinix are working together to provide robust and secure cloud services. AI plays an important role in enhancing these services.
Contact us for more information.
[i] GeForce RTX 40 Series Graphics Cards | NVIDIA
[ii]The Imperative for Sustainable AI Systems (thegradient.pub)
[iii] The role of artificial intelligence in achieving Sustainable Development Goals | Nature
[iv] Data Centre Company & Enterprise Network Technologies | Equinix
[v] Hybrid-Cloud, Multicloud, and Colocation for Your Connection to the World. (umb.ch)
![[Translate to Englisch:] Temi, der Roboter, im Einsatz](/fileadmin/_processed_/0/1/csm_UMB_Paraplegikerzentrum_Temi_864700ecb9.jpg)
