Is AI sidelining sustainable data center development?

With data center investments and expansion plans pouring into Southeast Asia, is sustainability data center development still a part of the conversation, given the increasing demand for compute to support AI use cases?

The data center industry in Southeast Asia is rapidly expanding in both scale and sophistication due to the region's increasing demand for AI, digital services, cloud computing, and internet infrastructure. Over the past two years, the growth in data centers increased its pace as the need for more compute power to support the increased usage of AI applications in the region.

Tech giants from the US and China continue to pour investments in the region, with Malaysia, Singapore, Indonesia, Thailand as well as Vietnam being the prime markets recording growth. Malaysia in particular continues to witness record breaking data center investments with AWS, Google, Microsoft and Oracle planning to set up cloud regions and data centers in the country.

Singapore’s lifting of a data center moratorium in 2022 has also resulted in more data center providers expanding their current infrastructure in the region. Singapore is also witnessing increasing growth in AI adoption, creating more demand for compute to support the applications.

While the growth in data centers brings in economic growth for countries in the region, there are concerns on where sustainability will fit in with all these developments. According to a report by IDC, IT power capacity for data centers in Asia Pacific is expected to double in 2024, primarily due to the emergence of hyperscale data centers. In Southeast Asia, the drive in infrastructure modernization efforts across the region and the AI-boom will see demand for more energy and power as well.

Sustainable data centers

For Rich Farrell, Regional Director for Cybersecurity, Digitalization, Data Center Businesses in APAC at Eaton, sustainability still needs to be on the radar for data center development, especially with data centers using as much power as what the airline industry does.

“It’s tipped to double the amount of electricity and power that we need by 2030. However, if we look at what's currently driving the growth, it's all about AI, and sustainability has taken a bit of a backseat, to be honest. An interesting stat from EY states that 70% of CEOs have deprioritized sustainability to number two in favor of AI and tech investments. So, sustainability is still a priority, but it's not a priority. For me, that's a bit of a concern,” said Farrell.

Judging by the data center expansion in the region, Farrell explained that if there isn’t a focus on how renewables can be integrated and how data centers can be built with sustainable designs, then the current boom is unsustainable itself.

“We can't actually keep up with all the power requirements that are needed. I think you're going to find that power usage efficiency (PUE) and water usage efficiency (WUE) are going to be used in the same breath in sustainability very soon. Both water and power are scarce resources, and without the focus on sustainability, there's going to be no way that we could continue this current growth,” explained Farrell.

At the same time, Farrell also pointed out that the industry has not reached its critical mass. However, he feels if the current growth trajectory keeps continuing, it will be at that critical mass.

“If you look at brownfield data center sites, a lot of them are going into retrofitting cooling services. We're all aware of immersion cooling and direct-to-chip cooling, which seems to be the preferred method to cool down the enormous chipsets that are needed for the LLM on AI as well. At the moment, it's sustainable, and with proper R&D, it drives the cost down as well. But we're not at that point yet,” he said.

Farrell highlighted that exorbitant costs to retrofit a brownfield data center is a challenge as data center operators are essentially changing the whole design topology of the data center. As such, the greenfield data centers that are now coming online, are all built around AI purpose-built.

“If the data centers are designed with AI in mind from day one, they won't have issues on expansion at growth. The problem happens when data centers keep adding more power to support increasing demands. How do they actually get enough power in that building? How do they still cool down these devices? How do they actually meet their sustainability metrics?” said Farrell.

An interesting point Farrell mentioned is that a lot of these sustainability metrics are self-imposed by data center providers. He said that every single hyperscale company globally that had their original sustainability projects in annual reports, have all shifted and changed them as its all driven by AI currently.

“While it's manageable currently, the current growth will be harder to manage. There's got to be that tipping point where something must give or something must change,” Farrell mentioned.

Designing and monitoring data centers

According to Farrell, alternative design in data centers is the way forward for sustainable data center development. This means having the ability monitor the capabilities and outputs that are coming from data centers. Apart from that there is also the use of efficient materials in the design of products.

“Our digital platform Brightlayer pretty much encompasses our entire software portfolio. The Brightlayer Data Centers suite is a portfolio of digital solutions that will enable you to efficiently manage your increasingly complex ecosystem of IT. We're now working with AI companies to try to integrate AI monitoring and provisioning into our software for the data center itself,” said Farrell.

Data center teams can measure PUE and monitor it in real time as well as check their water usage effectiveness, carbon emissions monitoring and discover data holes that are using more power. Farrell also mentioned that some of the large hyperscale companies are using Eaton’s Electrical Power Monitoring System (EPMS) and Data Center Infrastructure Management (DCIM) platforms.

“We are taking that software and then using the capabilities of AI organizations, integrating these together, and then actually rolling them out to our customers. An example is Red Dot Analytics, a local startup in Singapore. They're already working with a couple of co-location organizations in the country and we’re working together with them to see how we can integrate our software and their AI engine together,” explained Farrell.

For Farrell, at the end of day, it all goes down to the design of the data center. If a data center is designed properly to cater to the needs, including the access data, there will be much better output and sustainability.

“But the question that remains is which data center design are we using? Will data centers continue with their traditional designs or pivot to an AI data center? And how do we make sure that we're not sucking up all the power in the entire region? How do we know two or three years from now that there is going to have to be separate power grids for AI?”

“Actually, you're going to have sub-power grids just for technology out there. Apart from the design, it's also about how do these organizations work? It's about partnerships between people. We actually have a very robust channel structure that we use as well,” said Farrell.

Farrell highlighted as well that the majority of data centers used are usually directly to the organization themselves, which is the grey space. At the same time there is all the infrastructure which is the white space. These are the racks, servers, networking gear and security gear that goes in as well. And this is where Farrell believes having the right network of partners is key.

“While the design of data centers is one part of the solution, the construction and installation of the data center is just as important as well. Only then, can data center sustainability be something possible,” he concluded.

As governments in the region push for sustainable data center development, meeting these requirements can be challenging especially with increasing AI workloads demanding greater compute power. As such, sustainable and green data centers are a critical response to the growing environmental impact of global data infrastructure.

By incorporating energy-efficient technologies, renewable energy, innovative cooling methods, improved data center designs can help reduce carbon footprints while meeting the increasing demands of the digital world. As regulations tighten and public expectations rise, the trend toward sustainability in data centers will likely continue to accelerate.