Chris Sharp, Chief Technology Officer

This is part 2 of our 3-part blog series about Artificial Intelligence (AI) and AI-ready data centre infrastructure.

Read Part 1: Harness AI’s Potential and Navigate Disruption with Digital Realty
Keep reading for Part 2
Read Part 3: Integrating AI with Legacy Infrastructure

Designing data centres for the future

In my role as CTO, I’m often asked how Digital Realty designs our data centres to support new and future workloads, both efficiently and sustainably.

How do we ensure that data centres built for the simple enterprise use case of web hosting can support the next generation of AI workloads and the huge volumes of data that must be stored, transmitted, and processed?

My answer? Digital Realty first presented publicly on the implications of AI for data centres in 2017, but we were tracking its evolution well before that. We’ve built the required variability and modularity into our data centre designs, anticipating the tipping point that resulted in a surge of AI applications.

The AI-ready infrastructure you need to power AI innovation isn’t just about building one data centre – learn why it requires a modular global data centre platform that can support your needs at scale.

IT infrastructure needs are always evolving

The internet, the things we use it for, and the infrastructure that supports it are always changing.

AI isn’t the first new technology to impact IT infrastructure requirements, and it won’t be the last.

Through the emergence and maturity of technologies like email, web browsing, social media, and streaming video, we’ve experienced massive changes in the requirements for scale and complexity when it comes to data centres and the infrastructure within our data centres.

AI’s unique IT infrastructure needs

AI is pushing a new generation of chips, network equipment, and other infrastructure into the data centre. We’ve witnessed great changes in the size and number of networks that connect into our data centres and the amount of data stored by a typical enterprise. This is why we’ve been sharing our insights about Connected Data Communities and the impact of Data Gravity for years.

AI requires more computing, network, and storage infrastructure than any shift on the internet since the cloud, and in many cases, we believe it will be even more intense.

We’ve been an active player in this evolution: Digital Realty’s global data centre platform, PlatformDIGITAL^®, was chosen to be the home of many groundbreaking AI training sets and applications that showed what was possible. For example, we’ve supported these customers, such as Castle Globalⁱ , SURFsaraⁱⁱ , and Graphcoreⁱⁱⁱ, with the infrastructure they needed to power their innovation, and we’re using our experience, expertise, and key partnerships to help enterprises that are ready to begin their AI journey today. Most recently, we announced that we are one of the first data centre providers globally to be NVIDIA DGX H100-ready as part of NVIDIA’s DGX-Ready Data Centre Program for our newest data centre in Osaka, Japan, KIX13.

System of systems: modularity is the key to a timeless design approach

Digital Realty data centres are comprised of many individual systems, down to the cooling technology used for one rack, on one floor, in one suite, in one building, and on one of our campuses.

By designing the data centre as modular, interconnected systems, we can evolve just one or a small number of those systems at a time without needing to redesign or retrofit everything at once. This is called a "system of systems" design approach. This approach is cost effective and operationally efficient. We expect our customers to benefit from modularity because it reduces the time and cost required by them to support new infrastructure.

Modularity enables customised AI deployments

With this design approach, we can also solve a customer’s bespoke requirements without impacting others. This is vital to meeting the needs of AI, as new deployments are often needed quickly and have specialised needs when it comes to power consumption and cooling.

AI applications change the size and density of the computing infrastructure deployed on PlatformDIGITAL^®, which directly impact the design of our data center facilities to accommodate the needs of our customers.

For example, some AI infrastructure will draw 10 times the industry average power per rack but may still need to be deployed in the same physical space as other, lower energy usage infrastructure. Our highly modular approach easily enables this level of flexibility.

Modularity enables sustainable cooling technologies

Another example of Digital Realty’s modular design philosophy in action is our ability to deploy liquid cooling in our data centres.

The cooling requirements of AI infrastructure are driving significant new interest in using liquid cooling in the data centre. This method of cooling differs from traditional air cooling by using a loop of coolant pumped around the facility, connecting to the heat-generating chips in the racks, and then moving that heat to a large radiator where it can be removed.

Computer CPU cooled by liquid cooling on a motherboard

^{Computer CPU cooled by liquid cooling on a motherboard}

Liquid cooling is a highly efficient way of cooling AI infrastructure because liquid is 800 times denser than air, making it much easier for liquid to move excess heat from IT equipment. It also enables data centres to use power from utilities more efficiently, reducing overall power usage.

Liquid cooling is a notable example for our modular design philosophy. Even five years ago, using liquid cooling in the data centre was seen as esoteric outside of edge cases – the infrastructure used by the applications of the day simply didn’t generate enough heat to warrant its use. Today, it’s increasingly seen as a hard requirement for dense AI infrastructure deployments.

Many of Digital Realty’s data centers have a chilled liquid system that can be used for this purpose, and some don’t – but this doesn’t stop us from providing a liquid cooling solution for our customers.

We’ve spent years researching, developing, and operating many ways to support new cooling demands. For example, we use high-efficiency air cooling and other alternate liquid cooling technologies that can be fitted to a floor, a suite, or down to even an individual rack. We can accommodate these alternatives because our designs are modular.

Rear door heat exchanger cooling a High-Performance Computing (HPC) deployment in a Digital Realty data center facility

^{Rear door heat exchanger cooling a High-Performance Computing (HPC) deployment in a Digital Realty data center facility}

Not only does this modular approach let us meet customer delivery time requirements and performance goals; it also allows us to be cost-effective and highly sustainable. We aim to use the most efficient technology that fits a specific customer's need, mixing-and-matching across even a single floor, on a global scale.

This flexibility creates complexity: there’s no one-size-fits-all solution

Operating at scale with many unique systems at over 2.4 gigawatts of power may seem daunting, but it’s what we do best at Digital Realty. PlatformDIGITAL^® is the world’s leading global data centre platform with the biggest multi-tenant data centre footprint, and we’ve honed our expertise and capabilities by partnering with more than 5,000 of the largest and most innovative companies in the world.

It’s not just about what we can do today: we’re ready to support our customers for decades to come. Digital Realty is expected to have more than 400 megawatts of additional data center capacity coming online in 2023 and 2024, designed in the same future-proof manner as our other facilities.

IT infrastructure should begin with the end in mind

Digital Realty began evaluating the future impact of AI on the internet and on the data center over six years ago.

Although we can’t predict when an application will mature or an AI service will become the next big thing, we’ve seen repeatedly with our customers that the key to successful digital transformation is to plan your infrastructure deployments with the end in mind as much as possible.

Going slow now to design and deploy infrastructure that’s future-proofed will save you time and money down the road.

Stay innovative, reach out to us, and let’s deploy AI in a way that’s scalable and most effective for you.

This is Part 2 of a 3-part series about AI-ready data centre infrastructure:

Part 1: Harness AI’s Potential and Navigate Disruption with Digital Realty
Part 2: Are Data Centers Obsolete in the Age of AI? Not On Our Watch
Part 3: Integrating AI with Legacy Infrastructure

Are Data Centres Obsolete in the Age of AI? Not on our Watch