Artificial intelligence is currently the most potent force transforming how we process, analyze, and use data. Predictions say that the artificial intelligence market will reach a whopping $407 billion by 2027. We are talking about an expected annual growth rate of around 37 per cent until 2030. Our interlocutor, Jelena Pejković, Sales Director of Secure Power at Schneider Electric, points out that it is crucial to understand that we are not only talking about generative AI here; this technology will revolutionize numerous industries.
Q: What is artificial intelligence?
A: Artificial intelligence is much more than ChatGPT and similar Generative AI applications intended for the broadest audience. AI already has applications in industry, medicine, education, science, autonomous driving systems, and many other fields. Its exponential growth brings along challenges, such as a lack of new locations and new mega data centers that require large amounts of energy and, therefore, will be responsible for a higher percentage of CO2 emissions, which brings us to the question of sustainability. However, the answer to sustainability challenges is digital transformation, of which AI is an integral part, so this must not be an obstacle in their construction because it would slow it down significantly.
This story is fascinating, like the eternal dilemma – which came first, the chicken or the egg. Honestly, I’m thrilled that I’m dealing with this segment right now when this topic is hot. Perhaps it is easiest to start with Generative AI because everyone has already encountered it. Whether talking about Chat GPT or Microsoft Copilot, as many as 97 per cent of company managers already see the benefits of using AI in their daily business, generating reports and presentations, translating information, developing websites, etc. Still, this peak in the adoption of new technology will undoubtedly require more significant investments in infrastructure.
In Serbia, the topic of renewable energy sources, solar power plants and wind farms is topical. Here, AI algorithms also play a crucial role in optimizing energy distribution through the network in real-time. These algorithms continuously analyze data from the grid and adjust the flow of electricity to meet demand while at the same time ensuring the stability and efficiency of the entire grid. With these two examples, I want to illustrate how broad the field of AI application is.
IN FOCUS:
- PROCREDIT BANK – THE NEW FACE OF DIGITIZATION
- CEEFOR IS DESIGNING ANOTHER POWER PLANT
- SUPPORT FOR THE IMPLEMENTATION OF INNOVATIVE SOLUTIONS FOR A SMOOTH GREEN TRANSITION
Q: How can we meet the demands of this new world driven by artificial intelligence?
A: Data centers represent the critical infrastructure that supports the artificial intelligence ecosystem. Here, it is perhaps important to emphasize, without going deeply into the topic, that we distinguish training and inference models. So, there are data centers where training clusters are located, where training or model training is carried out, and data centers where AI applications are located that we, as users, use for decision-making.
However, whether we are talking about large clusters for training or smaller edge servers that run applications, artificial intelligence is becoming an increasingly large percentage in the data centers themselves, and this is currently affecting a significant increase in consumption per river, i.e. density. These new, significantly increased densities further influence the design and management of data centers. Here, we talk about four key AI attributes and the latest trends that answer the challenges of the physical infrastructure of data centers: power supply, cooling, placement in racks, and management software.
Perhaps a more significant issue than the increase in energy consumption is its more efficient use. Schneider Electric offers numerous solutions and continues developing new ones independently and with numerous partners, such as the recently announced cooperation with NVIDIA.
Q: How can we provide the necessary electricity for Al?
A: Regulatory requirements are strict, but despite that, Internet giants are leading the way when it comes to sustainability and corporate social responsibility goals, pushing the entire industry forward. The world’s leading data center operators largely purchase energy from green sources, introduce a circular approach to energy use, hand over waste heat to heating plants in local communities, limit water use, and recycle.
Although artificial intelligence requires large amounts of energy, data analytics based on AI algorithms can help data centers move closer to net zero and address sustainability issues. So, AI is both a challenge and a solution.
Interestingly, 1,287 MWh of electricity was consumed to create GPT3, and 552 tons of CO2 were produced, equivalent to the emissions that 123 gasoline vehicles would produce in one year of driving.
Q: How can we meet the increasing demand for AI power while minimizing the impact on the planet?
A: Data centers are constantly evolving to adapt to the demands of AI. Improving power distribution systems and energy efficiency within the data center helps minimize losses and deliver power to servers most efficiently. Operators are focusing on more energy-efficient hardware and software and diversifying power sources. Advanced power distribution units, intelligent management, high-efficiency power systems, and renewable energy sources enable data centers to reduce energy costs and carbon emissions. However, the extreme densities of AI servers lead to challenges related to cooling methods.
The transition from air to liquid cooling is imperative, primarily from a sustainability perspective. Liquid, Direct-to-Chip cooling, where coolant is circulated through servers to absorb heat, is rapidly gaining popularity. The advantages are numerous: from increasing the reliability and performance of the processor, saving space, increasing energy efficiency, improving the PUE coefficient, and reducing water use.
Data center operators can also use automation based on AI algorithms, analytics, and machine learning to find new opportunities to increase efficiency and decarbonize. By using data insights more effectively, we can drive new, more sustainable behaviors.
Here, I am primarily referring to DCIM, EPMS, BMS, and similar applications that reduce the risk of unexpected behavior and provide a digital replica of the System, which makes decisions easier.
One example is Equinix, which improved the energy efficiency of its data center by nine per cent using AI-based cooling. The company reduced energy consumption by regulating the cooling system more efficiently.
So, it is clear that AI applications are leading to a significant increase in electricity consumption in data centers at a time when they need to become more sustainable. However, AI simultaneously provides us with intelligence, with the help of which we will manage those same data centers more thoughtfully.
To conclude, by combining quality and efficient equipment with the contributions made by a monitoring system based on AI algorithms, data center owners, operators, and users can respond more efficiently to the demands of high-density AI clusters without risking energy efficiency, reliability, and sustainability.
Schneider Electric
Read the story in the new issue of the Energy portal Magazine NATURE CONSERVATION
