Siemens: Can AI Solve Grid Problems with Data Centres?

The energy demands of the data centre industry are reaching unprecedented levels as AI and cloud services expand rapidly.

AI workloads regarding power density are now blazing past the 100kW per server rack threshold, leaving traditional infrastructure struggling to keep up..

For technology companies like Siemens, this creates both opportunity and pressure. 

Ciaran Flanagan, Vice President and Global Head of Data Centre Solutions and Services at Siemens, tells Energy Digital's sister title AI Magazine about the current data centre market and AI’s role in it.

What about Siemens’ approach to the data centre industry are you most proud of?

We’re in the most transformative and exciting period in data centre history – and demand has never been greater.

But this demand brings with it immediate challenges that the industry needs to tackle. 

These challenges include:

• The need to ramp up data centre capacity rapidly to deal with increasingly intense cloud and AI workloads
• The responsibility to meet sustainability regulations while bringing more electricity generation online
• Rising operating costs and capital intensity
• Speed pressures on the construction pipeline
• Digital skills gap

The good news is that this convergence is creating opportunities for collaborative innovation that we couldn’t have imagined just a few years ago. 

For Siemens, the focus has been on how quickly we can adapt to the challenges and support the evolution of our industry. 

We believe that digital technology is fundamental to this. Integrating software and smart technology across the data centre infrastructure will boost operational effectiveness and help us all to become more efficient, resilient and sustainable. 

We are working tirelessly to develop tech solutions that solve those very specific challenges of our data centre customers.

That innovative, collaborative approach is what I’m most proud of, we can’t do this alone.

How are power demands changing the economics of data centre operations?

Advances in cloud and AI workloads and high performance computing (HPC) are reshaping data centre power infrastructure.

Specifically in the domain of AI, with power densities reaching unprecedented levels – often exceeding 100kW per rack – traditional distribution systems are having to be reimagined. 

Designers are dealing with sub-second spikes in power consumption, new space and weight requirements for power distribution and perhaps even changing risk profiles when it comes to fire safety due to unprecedented thermal loads, all very diverse topics. 

These new requirements demand innovative and integrated approaches to power delivery and operations. 

Besides these new challenges within a data centre, externally, “time to power” has become a critical bottleneck in the industry, especially when deploying new facilities. 

Grid connection backlogs are stretching into years, with many operators facing MW/GW-scale connection delays or outright rejections.

In response, leading operators are taking control of their destiny by either developing on-site generation capabilities with microgrids or forming strategic partnerships with utility providers. 

At some point, data centres will become part of the grid ecosystem. 

So instead of a utility company having to build another gas fired power plant, perhaps, they can use excess energy storage capacity from a set of data centres that sit on the grid. 

Data centres are becoming prosumers, by offering balancing services and demand-side response to help network operators better manage the grid with a greater penetration of renewable energy sources. 

So when we look at the grid and the power system in its entirety, we will see: everybody’s got a part to play.

Again, digitalisation is key to making this a reality. 

How can data centres meet both AI compute demands and sustainability targets?

We have to be realistic. 

Given the growth, it is difficult for data centres to be or become net zero at the moment. 

Everything we do as humans has an impact and data centres are no exception. 

As an industry, what we can do is to make sure that the net benefit of having a data centre is greater than its impact – and so far, operators have been quite aggressive and innovative in adopting sustainability measures. 

For example, the data centre industry has been one of the driving forces behind the adoption of renewable PPAs (power purchase agreements). 

But with the limited availability of power, as mentioned above, it’s not enough to simply purchase renewable energy anymore – operators are moving towards generating their own energy onsite, from renewable sources like wind or solar farms, feeding it directly into their operations. 

Plus, they’re working tirelessly to squeeze out every watt of efficiency they can find. 

Somewhat ironically, AI technology is also part of the solution to making this a reality and can counter its own increasing power demands by helping data centre operators to achieve their sustainability targets. 

For example AI apps, like our Building X Suite, includes Operations manager which uses fault detection diagnosis to predict patterns and enable preventative maintenance, which helps drive efficiency. 

There’s also AI-driven software to optimise white space cooling. 

AI can help industry drive out energy waste and optimize the grid to manage demand effectively. 

Especially when you consider the additional pressure that the integration of renewable energy can put on the grid. 

This is where smart technology is needed to manage the added complexity and ensure a consistent supply.