Every AI task uses power. That power is electricity. Tech companies keep finding new ways to integrate AI into every facet of our lives. ChatGPT often provides better search results than Google, which displays ad after ad. This takes energy. So how worried should we be about AI’s electricity demands?
The Energy Challenges Posed by AI, Crypto Mining, and Data Centers: A Deep Dive
Around the world, thousands of data centers host computers processing our requests. As these tasks become more sophisticated and AI evolves toward superintelligence, the demand for electricity will increase significantly. Companies like Platocom are closely monitoring this digital infrastructure revolution, specializing in data center migration, audits, compliance, and deployment. Governments, tech companies, and data centers have only one way forward: a commitment to sustainability and innovation, driving the development of green data centers. These are essential as AI's energy demands continue to grow.
"Electricity consumption from data centers, AI, and cryptocurrency could reach double 2022 levels by 2026." ~ International Energy Agency (IEA)
Similarly, the rise of cryptocurrency mining presents another significant energy challenge. Mining operations require vast amounts of computational power (and water!) to solve complex algorithms, leading to high energy consumption.
"AI is already using somewhere between 8-10% of the world's electricity budget." ~ Kate Crawford
In Ireland, about 30% of the whole country’s electricity supply is used for data centers.
The AI Revolution and Its Energy Demands
In our increasingly digital world, even seemingly simple tasks on modern laptops demand significant energy. Now, imagine a global network of millions of interconnected computers, not just creating spreadsheets but developing advanced AI algorithms. These algorithms could potentially combat global warming, create new jobs, and even develop cures for diseases like cancer. The pressing question is: How much power will be needed to operate such a massive network, along with the cooling systems required to keep it running?
We at Platocom are avid listeners of tech podcasts and radio programs, and especially BBC’s The Inquiry. So we thought we should share what we learned with you.
The Power-Hungry Nature of AI
"Artificial intelligence is the name we give to super-advanced computer programs that can do much more than what we're used to using computers for," says Mark van Riemann, a strategic futurist based in Sydney, Australia. "They can process huge amounts of data and work out ways to do things far beyond the human brain's capability, and they're already delivering amazing breakthroughs."
For instance, Google launched Med-PaLM, akin to a healthcare-focused ChatGPT, providing doctors with a digital assistant to aid in patient diagnosis. Similarly, DeepMind's AlphaFold, which predicts protein folding, is revolutionizing personalized medicine. Van Riemann highlights that AI can transition healthcare from a reactive to a proactive system, emphasizing the transformative potential of AI technologies.
But the capabilities of AI come with a significant energy cost. Kate Crawford, a professor at the University of Southern California, notes that AI's computational demands are immense. "One estimate suggests that by 2027, AI could use as much electricity per year as a country the size of the Netherlands," she explains. "AI is already using somewhere between 8-10% of the world's electricity budget, which is enormous."
To Sustain AI's Growth, Renewable Energy Sources Are Ideal, but the Transition Is Challenging
Presently, our energy grids struggle to meet existing demands, and non-renewable sources like coal and nuclear power often fill the gap. Sam Young from the UK consultancy Energy Systems Catapult discusses the increasing global energy demand driven by rising standards of living and technology usage.
Sean notes, “We see in Southeast Asia a lot more demand for electricity because more people want air conditioning. Everyone’s having to build more and more power transmission capacity to cope with that.”
Generating and transmitting energy efficiently is a complex task. Energy generation varies widely across the globe, and each region has unique challenges. Sean explains, “There's no such thing as a global power grid. Each country generally has its own power grid. In some parts of the world, like Europe, national grids are connected, so electricity generated in Sweden can be sent to Italy. But in other parts, there’s no reliable grid at all.”
The more the world uses AI, the more energy it needs. Sean says, “There are two reasons we’ll need more energy for AI. The first is that some AI models will get bigger and need more energy. The second, and probably more important, is that we’ll use AI models more and more, increasing the total energy needed.”
This increased demand could lead to bottlenecks. In some countries, data centers are already consuming a significant portion of the electricity supply. Sean gives an example, “In Ireland, about 30% of the whole country’s electricity supply is used for data centers. That means you can’t use that electricity for something else.”
Even if a country’s grid can cope now, future demands might be overwhelming. Sean continues, “Data center energy use might increase by 50% in the next five years and could double over longer time frames. It’s important to remember that this isn’t just driven by AI, but also by our demand for streaming video, loading websites, and storing data on the cloud.”
The Path Forward: Sustainable Solutions
The key to addressing these energy challenges lies in innovation and sustainable practices. Platocom’s is increasingly focused on green data centers. By focusing on energy-efficient technologies and renewable energy sources, these centers aim to reduce their environmental impact while meeting the growing demands of AI and data processing.
Africa Rising: A Potential Energy Solution
While the energy demands of AI are daunting, Africa presents a unique and promising opportunity to meet these needs sustainably. Rose Matiso, Research Director of the Energy for Growth Hub, sheds light on Africa's role in this equation. "Africa accounts for less than 1% of global data center capacity," she says. "The total number of data centers in Africa is roughly equivalent to those in a single city like Mumbai."
Despite this, Matiso sees an opportunity. "You can use electricity-hungry data centers as anchor customers for utilities," she explains. "This drives revenues and improves the quality of power for the entire network. For poor countries that are energy-starved, the power requirements of AI and digital infrastructure can be less of a threat and more of an opportunity to build out this infrastructure and start closing energy access gaps."
Furthermore, Africa's potential for renewable energy is vast. "There's a massive clean energy opportunity in Africa," Matiso emphasizes. "Countries like Kenya are actively pitching themselves as destinations for data center companies. With abundant solar, wind, and geothermal power, African countries can meet the stringent carbon targets set by digital companies."
The Promise of Renewable Energy in Africa
Africa's renewable energy resources offer a promising solution to the growing energy demands of AI. "On the African continent, there's a lot of renewable potential, whether it's solar energy, wind power, or geothermal energy," Matiso notes. This aligns perfectly with the sustainability goals of many tech companies, making Africa an attractive location for new data centers.
Not only does this approach benefit the tech companies, but it also has significant positive impacts on local economies and communities. "Having more local data centers will benefit businesses and users alike," Matiso explains. "African users will experience faster, more reliable internet services, and businesses will have better access to cloud services, which can drive economic growth and job creation."
Conclusion: A Bright Future with Africa at the Helm
In conclusion, will we have enough energy to power AI? In the short term, it seems likely. Africa's renewable energy potential and strategic positioning could play a crucial role in meeting these demands. The challenge lies in ensuring that the electricity used by AI data centers comes from sustainable sources.
As AI continues to develop, its energy consumption will rise, but Africa's growing infrastructure and renewable energy capabilities offer a promising solution. By leveraging Africa's potential, we can turn the energy demands of AI into an opportunity for growth and innovation on the continent.
Credits
The Inquiry from the BBC World Service was presented by David Baker, produced by Louise Clarke, researched by Katie Morgan, and edited by Tara McDermott. It was mixed by Craig Boardman.
The esteemed guests were:
Dr Mark Van Rijmenam, a strategic futurist
Kate Crawford, research professor at the University of Southern California in Los Angeles and senior principal researcher at Microsoft Research in New York
Rose Mutiso, research director of the Energy for Growth Hub
Comments