How Google Supports Demand Flexibility for AI Data Centres

The expansion of AI could present a major challenge to electrical power systems. 

Data centres are growing to support this and, therefore, using more energy.

The International Energy Agency (IEA) says that, in the United States, power consumption by data centres is set to represent almost half of the growth in electricity demand through 2030, caused primarily by AI.

In response, Google is looking to enhance the flexibility of its data centres. 

Grid operators typically maintain surplus power capacity to manage peaks in demand. 

In the US, for instance, approximately 50% of generating capacity is used on average, with the remainder held for high-demand periods. 

The construction of new generation or transmission infrastructure to support large AI workloads is a process that could be both expensive and lengthy.

Instead of building out the grid to accommodate larger peaks, Google is focusing on making energy demand more adaptable. 

This approach forms the basis of demand flexibility, a strategy to move or decrease energy use in real time, particularly for large-scale computing tasks. 

By aligning its compute demand with times of high clean energy availability or lower grid strain, Google can meet its needs while mitigating stress on power systems.

Integrating intelligent demand

 Google is now applying these principles to its ML workloads for the first time, developing a method that can support both the expansion of AI and the stability of its grid partners

“We’re sharing our advancements with new flexible demand capabilities in our data centres, now for the first time by targeting ML workloads,” explains Kate Brandt, Google’s Chief Sustainability Officer. 

“This new approach can support AI growth and our grid partners at the same time — helping utilities reliably and cost-effectively meet the electricity needs of all their customers. 

“While this is still early stages, we see demand response as a promising tool to move or reduce our power demand, providing flexibility when the grid needs it most.”

Demand response and collaboration

Demand response, where consumers adjust power usage based on incentives or price signals, is a key component of this strategy. 

Technologies like demand response are important for reducing the need for new transmission and distribution infrastructure according to the IEA.

By targeting the ML workloads that are central to modern AI, Google could make its data centres more flexible.

To implement these capabilities at scale, Google has formed partnerships with Indiana Michigan Power (I&M), Tennessee Valley Authority (TVA) and Omaha Public Power District (OPPD). 

In a pilot programme with OPPD, Google confirmed the ability to lower ML-related power demand during events of grid stress.

“I&M is excited to partner with Google to enable demand response capabilities at their new data centre in Fort Wayne, IN,” says Steve Baker, President and Chief Operating Officer of I&M.

“As we add new large loads to our system, it is critical that we partner with our customers to effectively manage the generation and transmission resources necessary to serve them.

“Google’s ability to leverage load flexibility as part of the strategy to serve their load will be a highly valuable tool to meet their future energy needs.”

The benefits of flexible demand

Making large-scale ML demand more flexible could provide several benefits, including:

• Faster AI deployment: New data centres can be connected to the grid without the need to wait for new power plants or transmission lines to be built
• Lower costs and carbon: Utilities can serve new demand with existing infrastructure, which reduces the need for capital projects and lowers the associated carbon footprint
• Grid resilience: Real-time load management contributes to a more robust power system, important to the integration of intermittent renewables.

“As AI adoption accelerates, we see a significant opportunity to expand our demand response toolkit, develop capabilities specifically for ML workloads and leverage them to manage large new energy loads,” Michael Terrell, Google’s Head of Advanced Energy, says.

“By including load flexibility in our overall energy plan, we can manage AI-driven growth even where power generation and transmission are constrained. 

“We believe this is a promising tool for managing large new energy loads and facilitating investment and growth.”