STT GDC turns to AI to optimise its data centres

STT GDC is turning to AI to improve the sustainability of its data centres, beginning with a pilot in Singapore.

The plan is to trial an AI-based autonomous control system from Phaidra to optimise data centre cooling in STT GDC’s facilities in Singapore.

Phaidra who?

Phaidra was founded in 2019 by ex-DeepMind executives and an HVAC industry veteran. So this is a team that knows exactly what it is doing.

STT GDC has 6 data centres here, supporting both air- and liquid-cooling with diverse computing workloads.

• STT S'pore 1, S'pore 2, S'pore 3 at Defu.
• STT S'pore 4 at One-North.
• STT S'pore 5 at Tai Seng.
• STT S'pore 6 at Loyang.

STT GDC is the first operator to pilot Phaidra's AI control systems in Asia. If it works, the pilot will serve as a model for deployment across its 95 data centres globally.

Traditional AI

It's worth noting that "AI" here isn't generative AI, but what I call "traditional AI" or machine learning (ML).

To my knowledge, this was first pioneered by Google over a decade ago. And won't you guess, Phaidra CEO Jim Gao was the technical lead responsible for it then.

I recalled how the VP of Data Centre at Googe, Joe Kava, lauded Jim at an event many years back. From memory:

• Jim took an online course to learn ML.
• Used it to optimise a Google data centre.
• Was eventually tasked to roll it out globally.

Apparently, some of the recommended settings were counterintuitive even to data centre experts at Google. And it also took the weather into consideration.

But it worked.

Better cooling energy efficiency

So how will Phaidra's solution do its magic?

• Agents analyse thousands of sensors in real time.
• AI learns and adapts to changing facility conditions.
• Incorporates external factors such as weather changes.

STT GDC says this initiative aligns with its commitment to be carbon neutral by 2030 and Singapore's vision of a green data centre industry.

• Will test them in a hybrid cooling environment.
• Baseline savings of 10% expected initially*.
• Potential increase to 30% savings*.

*Cooling energy savings.

Leaders in sustainability

As the power needs of AI workloads continue to grow, data centre cooling and operational infrastructure must evolve to keep pace.

However, it takes considerable leadership and technical efforts to move the needle for more sustainable data centres.

This is particularly the case for existing data centres. So far, I've only seen a handful of operators like STT GDC and Digital Realty announce tangible initiatives.

Hopefully, more data centre operators will step up to the challenge.