Victoria Serghievici of Coherent Solutions discusses the growing energy demands of AI and how to mitigate them.
The rise of artificial intelligence (AI) applications and the pursuit of artificial general intelligence (AGI) are creating demands for massive increases in energy resources. According to the International Energy Agency, data centers feeding the massive amounts of data required for AI are expected to consume around 3-4% of the world's electricity by 2026.
To put this into perspective, the entire global electricity supply in 2023 was around 29,734 TWh, but if data centers were to soon consume around 1,189 TWh (4%) per year, this could be more than four times the total annual electricity consumption of the UK. Most of the energy used to generate electricity is derived from fossil fuels, which has serious environmental impacts.
Because AI advances require so much energy, “there's no way we're going to get there without breakthroughs,” OpenAI CEO Sam Altman said, noting that more clean energy is needed to meet AI-driven demands. The solution is to focus on renewable energy (mainly solar, and potentially fusion, though that's uncertain at this point) and adopt practical ways to reduce emissions.
Businesses need effective strategies for decarbonization
The world will see a 50% increase in renewable capacity in 2023 compared to the previous year. The AI industry is united in a broader call to balance technological advancements with sustainable practices. Organizations that develop and leverage AI solutions are prioritizing green initiatives today for a more sustainable future and to enable the expansion of AI without increasing carbon emissions.
Today's energy suppliers are increasingly developing farms that aim to provide 100% clean, renewable energy from wind and solar. These farms help homes and businesses make a positive impact on the environment by switching from fossil fuels to sustainable energy, reducing their carbon footprint and contributing to a cleaner future.
There are several key strategies that businesses of all sizes can adopt to significantly reduce their environmental footprint. Together, these strategies can help organizations move their operations forward in line with global environmental goals.
Use energy-efficient AI hardware
A focus on sustainability and reducing energy consumption requires the right tools to measure and manage your environmental impact. Accurate data and insights are essential to make informed decisions.
Most companies that don't operate their own data centers can use public cloud providers like AWS or Microsoft to run their AI workloads, and advanced tools like AWS Carbon Footprint can help companies measure their estimated carbon footprint and develop sustainability goals.
Deploying Green Data Centers
Data centers put a heavy burden on the environment due to their high energy requirements. To mitigate this impact, large enterprises with on-premise data powerhouses can take advantage of renewable energy sources such as solar, wind and hydropower.
Businesses can further reduce energy consumption by using liquid cooling and free cooling technologies instead of energy-hungry chillers. Machine learning algorithms can help improve cooling systems and workload distribution, minimizing energy waste and making data centers more sustainable.
Dynamically adjust computing power
Modern servers and storage devices are designed to consume less power, and companies can combine them with software that dynamically adjusts computing power based on real-time demand: Kubernetes and Apache Hadoop YARN, for example, automatically scale resources for large-scale data processing, and cloud tools like AWS Auto Scaling and Azure Automation optimize resource allocation.
The use of AI itself contributes to carbon reduction. AI helps identify areas where emissions can be reduced, streamlining logistics and improving resource utilization. It is critical to maintain the accuracy and efficiency of machine learning models, optimize computing resources and ensure AI operations run smoothly without consuming excessive energy.
The Future of Fusion
Powerful large-scale language models (LLMs), data analysis modes, and advanced AI applications require large amounts of energy that are currently difficult to predict: training a single model such as GPT-4, for example, can generate around 300 tonnes of CO2.
A theoretical clean energy solution, nuclear fusion technology can generate nearly limitless power without producing long-lived radioactive waste. Nuclear fusion is still an area of research, with breakthroughs but far from practical application. A team from Princeton University's Plasma Control Group used AI to help stabilize plasma, a key step in sustaining safer and more efficient fusion reactions. In the future, this development could make it feasible to use fusion energy as a practical power source.
Accelerating change to reduce emissions
As AI advances, so will energy consumption. Companies can significantly reduce AI's carbon footprint by adopting a multi-pronged approach. The key to change is to foster a conscious culture and drive AI development with sustainability in mind. This strategic decision is an ethical choice for the planet and will also help strengthen your brand's reputation and customer loyalty.
Victoria Sergeevich
Victoria Serghievici is a Machine Learning Engineer at Coherent Solutions. She develops scalable models that improve operational efficiency and help energy suppliers enable a cleaner world. She is passionate about continuous learning and keeps abreast of the latest ML and AI advancements.
Find out how emerging technology trends will reshape the future in our new podcast, Future Human: The Series. Listen now. Spotifyupon apple Or wherever you listen to podcasts.
