AI’s Power Play: Is Your Data Center Ready for the Carbon Cost?
- By Winston Thomas
- April 24, 2024
It was not long ago when there was an uproar over Bitcoin mining. We were bombarded by news reports claiming its power draw rivaled entire nations like Ireland.
Who would have guessed that was merely a warm-up round? With the rise of AI, especially GenAI, we’re about to witness a true energy revolution in the data center—and the infrastructure teams need to be at the forefront of handling the fallout.
“You’re really going to see a hockey stick [scenario] here,” says Matthew Oostveen, chief technology officer for Asia Pacific & Japan at Pure Storage.
He noted that companies need to start addressing limited floor space, power restrictions, and maxed-out cooling systems as they adopt AI.
And guess what? Pushing the problem to the public cloud won’t save you, adds Oostveen.
The metrics that really matter
Building a robust AI infrastructure isn’t just about throwing servers into a room. Data centers face hard limits:
• Floor Space: Powerful AI systems often pack a surprising compute density into small footprints. Can your racks handle it?
• Kilowatts: AI processors are notoriously power-hungry. Is your power grid or contract up to the challenge?
• Cooling: Those powerhouse systems generate heat—lots of it. If your cooling system can’t keep up, you’ll have throttling (or worse).
• Water (The Overlooked Factor): In climates like Singapore or Australia, water availability for cooling becomes a critical and often expensive factor.
Calculating the hidden cost of a query
“Did you know that for every 10 ChatGPT searches you do, you use about a liter of water?” Oostveen reveals. That’s due to the cooling demands of the massive data centers powering these AI models.
This means if you do 30+ searches daily, you’re outpacing your doctor’s hydration recommendations without even taking a sip.
This highlights AI’s hidden environmental cost. Every query, every generated image, and every line of AI-written code comes with a tangible carbon footprint.
Oostveen notes that companies touting ambitious ESG (Environmental, Social, Governance) targets can’t afford to ignore this reality.
Beware the cloud illusion
Outsourcing your AI to public clouds might seem tempting. After all, it’s their problem to handle the power and cooling, right? Wrong.
While those costs may not appear directly on your balance sheet, they still fall within your company’s Scope 3 emissions—indirect emissions throughout your value chain.
More spend generally equates to more power consumed, resulting in a higher environmental impact, says Oostveen. And with AI’s unpredictable cost fluctuations, you could be unwittingly torpedoing your ESG goals.
So, what are infrastructure teams to do?
It’s not all doom and gloom. Here’s where savvy infrastructure heads and storage admins can make a difference:
• Get proactive: Don’t wait for your data center to hit its limits. Audit your infrastructure now and start planning for expansion needs.
• Efficiency is king: Storage optimization takes on a whole new meaning. Look for technologies that maximize data density, lower power footprints, and integrate with smarter cooling systems like liquid cooling.
• The hybrid option: A mix of on-premises and targeted cloud use can help balance workloads and environmental impact.
• Advocate and educate: Be the voice of reason in your organization. AI is incredible, but its actual cost needs to be part of any strategic conversation.
AI is transforming industries, and data centers are ground zero for this revolution.
While the power and environmental challenges are real, ignoring them isn’t an option. By embracing a proactive and informed approach, infrastructure and AI teams can play a crucial role in ensuring we harness the power of AI responsibly.
This article is part of a CDOTrend eGuide on AI and Storage. To follow the various insights, trends and latest solutions, download the eGuide here.
Image credit: iStockphoto/VanderWolf-Images
Winston Thomas
Winston Thomas is the editor-in-chief of CDOTrends. He likes to piece together the weird and wondering tech puzzle for readers and identify groundbreaking business models led by tech while waiting for the singularity.