COVER STORY: The role data centres play in an organisation's AI implementation

Data centres are known for storing an organisation’s data and hardware but what some executives are unaware of is how pivotal they are in powering AI-based applications.

With 87 percent of Australian C-Suite executives leveraging generative AI across some, or all business functions, some of these leaders are unaware of the important role data centres play in powering their AI programs. 

A JLL report showed that as AI adoption continues to grow, a data centre will serve as the "backbone" of its growth. 

Guy Danskine, managing director Australia at Equinix said the relationship between AI and data centres is “symbiotic”.

“AI needs huge amounts of data in order to build and train effective models and a great place to host a lot of data and manage it securely and distributed is in a data centre,” he explained.

“And certainly, as AI continues to grow, we expect there to be some growth within our space as well.”

Danskine said the emergence of generative AI has come at the perfect time when a number of other factors have helped.

“Things like cloud infrastructure and how available and cost-effective that can be. That has certainly helped a number of these models accelerate and hit the mainstream,” he said.

“It's a confluence of factors where a number of the technologies have been around for a long period of time, but making it as accessible as it is today, there's some timing, which is what's going to drive the growth we're seeing now.”

The impact of generative AI

Danskine at Equinix said they are figuring out how to take advantage of generative AI’s popularity as it is still in its early days.

“ChatGPT is what's driving this mainstream awareness of generative AI and so now it is in every conversation,” he said. “Customers we're speaking to now, it has transformed from the nascent parts of some technology companies to now boardrooms and enterprises across Australia.”

Danskine said he is discussing with these enterprises and boardrooms whether generative AI is a threat or an opportunity.

“In many cases, it's both. How do we react? How do we invest? How do we resource behind it? That is the kinds of early conversations we're having with our customers because there's a lot of unknowns about how to capture the opportunity,” he added.

With more organisations using generative AI, the hardware itself has changed to keep up with the demand.

Steve Martin, head of sales, major markets Australia at NextDC said what has changed in the data centre space is density, both in terms of power, cooling and weight.

“One of the things that a lot of organisations don't necessarily think about is when you're building a generative AI platform, you might typically be looking at racks that are individually more than a ton in weight,” he said.

“When you add 10 or 20 racks into your generative AI platform, that's a huge amount of weight that only a very, very small number of data centre operators can handle.”

Martin said the weight does not impact the power consumed per se, but it does impact where you can physically place the equipment.

“For example, since our generation two data centres and generation three now, the floor ratings have all been in excess of a ton per square meter to allow for these types of deployments to be made, A lot don't have that rating, aren't able to physically hold the racks that are required,” he added.

Data sovereignty

When using generative AI, one of the big challenges organisational leaders need to understand is data sovereignty, according to Martin at NextDC.

“If you're going to be massively successful with your own generative AI platform, whether it's in the cloud or whether it's your own build-out, taking control of your data and understanding where it is at all times is one of the more critically important components,” he said.  

Martin said there is no major AI platform, or generative AI platform operating in Australia today.

“ChatGPT, all of the stuff from the hyperscalers is running overseas mostly out of the US,” he said.

“If I'm an Australian-based medical research company, and I want to use generative AI, I've got to ship all of my data overseas to do that today. If I want to use generative AI deeply to maximise the return, I need to feed it every piece of my intellectual property across my business.”

Martin said this is a very different risk profile to what organisations might traditionally be used to.

“Tracking that data sovereignty, managing how you retain control of your IP is one of the bigger issues that are on the top of the list for people to deal with,” he said.

The beginning of a new era

As many organisational leaders look to transition to AI-based software, Chris Thorpe, founder and CEO at Leading Edge Data Centres explained this is the beginning of a generational shift within the data centre industry towards better cooling processes.

“[AI] requires a lot more power, a lot more compute capacity means a lot more power means a lot more cooling,” he said.

A high-powered rack of storage would be somewhere between 15 to 20 kilowatts and Thorpe said it's moving towards 50 to 60 kilowatts, even higher-powered racks.

“We're moving into liquid cooling, liquid immersion, I was at a conference recently, and there was talk of 100, 150 kilowatts, even 200 kilowatt racks being developed,” he said.

“This is moving into an area, which is highly specialised. You think about the power requirement you need, you can't have that in a normal office, it needs to be in a data centre environment.”

With the growth in power the design in data centres will evolve, Thorpe added.  

“The thinking of data centres will evolve with this new high-power compute. But with that high performance compute as the AI engines are built, AI will effectively start to live out on the edge as it becomes distributed, being close to where the workloads are and close to where it's needed.”

Over the next five years, Thorpe said data centres will see a transformational shift as technology and AI continue to advance.   

“You're going to come out the other side and feel like it's 50 years that we've gone through an evolution of technology and the data centre needs to underpin this shift and this evolution,” he said.

To help drive this growth and evolution in data centres, Thorpe said there needs to be a large amount of investment made.

“Whether that comes from this local market, I think a lot of [capital] is going to come from overseas, but there's going to be some huge capital investments to make this happen and planning, you've got a few key players in this market,” he said.

“There's going to be a lot of pressure on those players to keep up with the demand.”

He said everything is getting bigger and the fact today there are 300-megawatt data centres is considered “crazy”.

“It's just a crazy number, 30 to 50 megawatts was a huge data centre five years ago. So, it's a huge shift that is happening,” he added.

Opportunities await

Harry Jensen, senior director of operations at Equinix said AI will be a growth area for the data centre company.

“Certainly, the operations team is excited by that and we've seen the advent of cloud interconnectedness and our data centres grew and our customers grew. This is just going to be another flavour of that,” he said.

Jensen said he is waiting in anticipation of what opportunities the AI boom will bring.

“There's different types of technology that are going to be deployed with this, whether it's immersion cooling, there's a lot of different types of cooling methods that we use with hot oil containment, cold oil containment,” he said.

“We're looking forward to this new chapter and the new technologies that will be coming out of it.”