As the AI Energy Council gathers, the questions hanging in the air are: How can you move the future without blowing the grid?
The large data centers needed to train and run modern AI are craving electricity. Data centre power usage in the UK is on track to increase six times by 2034. At that point, you can smoke almost a third of the electricity in the country. This is a huge tension to have a system built for a completely different world, with predictable one-way power flow.
The AI Energy Council, a team-up with Tech Giants, Energy Firms, OFGEM Regulator and National Energy System Operator, has an important job of predicting how thirsty this AI Beast is. Their work is happening just as the government is putting £2 billion into AI’s Opportunity Action Plan. This is an epic vision for weaving AI into hospitals, classrooms and businesses.
Peter Kyle, UK’s science and technology secretary, said: “Ensuring researchers and innovators access to the processing power they need will not only maintain their position as the world’s third largest AI force, but will also place UK expertise at the heart of AI breakthroughs, improving our lives, modernizing public services, and inducing economic growth, the basis of planning for change.
“We are looking clearly on the need to confirm that British AI can strengthen this golden age through responsible and sustainable energy sources. Today’s talk will help us to drive that mission and provide AI infrastructure that benefits the country’s communities as generations come without drawing up the ambitions of clean energy superpowers.”
It is difficult to exaggerate the enormous scale of the energy problem. Globally, data centers are expected to double the amount of electricity they need in just five years, and ultimately the entire UK is demanding three times the amount of electricity they currently use. AI is the main culprit.
One rack of AI servers may require 120 kW of power. This is a big leap from 5-10 kW, where you need a regular rack. These do not drink steady power either. AI workloads surge unpredictably, creating sudden, large-scale power surges that threaten the stability of the entire grid.
In response, the UK is planning a monumental overhaul. The centerpiece is a “Great Grid Upgrade” and a £58 billion investment designed to be a “generational expansion” for the power network. This includes building new, large-capacity electrical highways running from north to south, and expanding offshore grids to bring about a huge amount of new wind power.
Ed Miliband, secretary of Energy Security and NetZero, commented:
“To bring together the biggest players in AI and energy will help us to discuss the role that AI will play in creating a new era of clean electricity for our country and meeting the power demands of new technologies in building clean electricity systems for families and businesses.”
However, there is a major obstacle. Even if you build wind farms and solar panels, connecting them to power grids to address the current surge in AI demand is another story. The current process is slow, with over 600 renewable energy projects (billions of worth) queued. Some people say they could be waiting for 15 years.
Emergency reforms are being pushed to try to clear this backlog. This is an important step if the future of AI is to have green energy. The government is also trying to speed things up by declaring data centers as “critical national infrastructure” and setting up “AI growth zones” that allow for rapid tracking of plans and power connections.
The data center industry is moving from part of the problem to part of the solution. Not only is it a pig of power, it is also becoming an active partner in the energy grid. Many are chasing net zero targets, investing in their own on-site renewable power and taking part in the “demand side response” program. This means that you can intelligently pause non-urgent AI tasks when the grid is under stress and fire them again when green energy is abundant, balancing the entire system.
The AI itself is also useful. You can also use the same complex algorithms that require so much power to make your grid smarter, predict energy spikes, and optimize power flow in real time.
It’s obvious from now on, but it’s not easy. The UK has the right idea and puts serious money on the table to address the power grid demands of AI, but it all depends on speed and execution. You need to break the grid connection jam and great grid upgrades need to occur at the pace.
(Photo: Andreas Jabusch)
reference: Artificial Test AI Running a Real Business with Strange Results
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London. The comprehensive event will be held in collaboration with other major events, including the Intelligent Automation Conference, Blockx, Digital Transformation Week, and Cyber Security & Cloud Expo.
Check out other upcoming Enterprise Technology events and webinars with TechForge here.
