Amazon Web Service (AWS) has unveiled new sustainable data center components with a focus on energy efficiency, bolstering its ongoing commitment to sustainability across the board.
Among these is a new mechanical liquid cooling solution that will provide liquid-to-chip cooling in both new and existing data centers, AWS said. Liquid cooling is a technique that uses liquid to absorb heat and reduce the energy needed to cool.
As some AWS technologies do not require liquid cooling, there will be integration between air and liquid cooling for powerful AI chipsets such as AWS Trainium. This multimodal cooling design will provide maximum performance and efficiency, AWS said.
The firm is also paying attention to how its racks are positioned in data centers.
Using generative AI, AWS can predict the most efficient way to land servers and thus reduce ‘stranded power,’ energy that is available but unused or underutilized.
Customer trends were telling AWS about data center usage, AWS’ global tech lead for sustainability Margaret O’Toole told ITPro, so it needed to think critically about what its move would be.
“We worked with some chip manufacturers in order to enable liquid cooling, we were able to rethink how we actually land servers in the data center to increase the density,” O’Toole said.
“The net result of all of those optimizations is that we now have 12% more compute for customer workloads per data center,” she added.
AWS’ sharpened sustainability focus
AWS’ latest announcement builds on a sharpened focus on sustainability at the firm through which the company has delivered several green-focused, energy-efficient offerings over the past months and years.
Amazon set a goal to reach 100% renewable energy by 2030 and has already hit that target, according to a 2023 report. The firm also plans to be ‘water positive’ by 2030 according to a commitment made in 2022.
“We’ve always had this frugality principle, and if we look at technology and how fast technology moves, we’ve always had to kind of rethink how we do our business to align to tech trends,” O’Toole said.
O’Toole specifically highlighted Amazon Nitro, which she said was about increasing the density of customer workloads, as well as MicroVMs on Firecracker to further pursue this aim.
“We knew that we needed it, but we didn’t know what it was going to look like, and so having had those challenges, actually, I think set us up well for how we could develop in the future,” O’Toole added.
“We’re comfortable stepping back and going, ‘okay, we got to rethink a bunch of things here in order to move forward,’ and I think that served us pretty well.”
AWS isn’t worried about generative AI sustainability
AWS is approaching generative AI with a similar focus on sustainability, and O’Toole said the firm is mindful of the surging demands placed on infrastructure by the technology.
“It’s a new approach, and we were pretty thoughtful in saying, ‘Okay, this is a new way of customers behaving – let’s act accordingly,’” she said.
Asked whether she felt generative AI raised issues around sustainability, O’Toole said the company isn’t saying it’s “business as usual” but that it is “pretty confident in [its] approach at this point”.
The environmental impact of generative AI has become a key talking point over the last 18 months, with surging power demands placing significant pressure on data center infrastructure on both sides of the Atlantic.
Research from Gartner in November found nearly half (40%) of data centers could face power availability issues by 2027 as AI demand ramps up.
Similarly, a report from McKinsey in October called for sweeping investment across Europe to contend with growing power demands. The consultancy warned that data center energy demands are set to triple across the region by the end of the decade.
As such, McKinsey called on governments across Europe to boost investment in sustainable power sources and accelerate efforts to upgrade grid infrastructure.
AWS’ industry competitors have reported issues with meeting environmental goals due to their AI-related activities. Microsoft’s 2024 sustainability report, for example, revealed its carbon emissions had grown by 29% on the back of its AI-fueled data center expansion program.
The situation at Google is much the same, according to its own reporting. The firm’s greenhouse gas emissions have increased by nearly half in five years, with the steep increase attributed to AI-related activities.
