news 
The Insatiable Hunger For AI Power Continues To Grow
As we continue to rely increasingly on artificial intelligence (AI) in our daily lives, it’s no secret that the growing demand for power is putting a strain on the electrical grid. According to SAP estimates from 2022, it takes an astonishing 200 terawatt-hours of electricity per year to power all data centers around the world – and this number is only expected to grow.
Despite the challenges, it seems that the trend is far from slowing down. In fact, U.S. President Donald Trump recently announced a $500 billion joint venture to create new data centers to fuel AI services, further exacerbating the issue. The Biden administration has also taken notice, launching an AI Data Center Task Force in 2022 to address these growing needs.
The numbers are staggering – in 2023, data centers were already consuming approximately 4% of the nation’s total electricity, and the Electric Power Research Institute (EPRI) predicts that this number will skyrocket to a whopping 9% by 2030. It’s clear that we must find innovative solutions to meet these increasing demands.
Enter nuclear power – long criticized but gaining popularity among Americans. Microsoft is even sponsoring the reopening of the notorious Three Mile Island nuclear plant in Pennsylvania, promising to purchase its entire output to fuel their data centers and AI services. Google and Amazon are also investing in small modular reactors (SMRs) that can be built near tech companies’ facilities.
But it’s not all bad news – some companies are taking matters into their own hands by improving efficiency. SAP boasts that their data centers operate entirely on renewable electricity, sourced from solar panels and purchased certificates. By optimizing machine learning models and selecting the most suitable options for specific tasks, they’ve managed to reduce energy consumption and costs.
Other innovators are adopting similar strategies – using smaller AI models when possible, capping processing power during certain phases, and even halting underperforming models altogether can significantly cut down on unnecessary energy waste. The industry’s shift towards renewable energy sources is also gaining momentum, with companies like Google reporting a 67% increase in efficiency with their latest Tensor chips.
In an effort to curb the problem, some organizations are exploring alternative solutions such as fusion power and hydrogen-powered data centers. In a bold move, startup ECL announced plans to build just that – a hydrogen-powered AI data center in Texas, set to begin construction in 2024.
While the future may seem bleak, it’s clear that companies and governments alike must work together to find sustainable solutions. By leveraging innovations like these, we can continue to harness the power of AI while minimizing its devastating impact on our planet’s resources.
The clock is ticking, as is the demand for energy – it’s time for us to take control of this insatiable hunger and forge a path towards a more sustainable future.
Source: http://www.forbes.com
