Balancing ESG with the power needed for AI

As the world races to embrace AI, are ESG considerations stalling? For companies opening the throttle on their engines to power AI ambitions, failure to place ESG at the heart of their strategy could cause them bigger challenges down the road, with the risks of spiraling energy costs, ESG fines, and even reputational damage. A recent survey of 1,200 IT leaders globally highlighted the issue, with 76% of organizations rolling out AI for widespread or critical functions. Still, only a third (33%) place ‘impact on the environment’ among their top three priorities – showing a clear chasm between those two crucial areas.¹

The pressure for power to feed AI

According to a study by Goldman Sachs, AI will drive a 160% rise in data center power demand; unsurprising when it’s also claimed that a single ChatGPT search can use ten times more electricity to process than a Google search. Depending on where you look, the figures vary. Still, the trends are all pointing in one direction: the rise of AI-driven processes, particularly in data-intensive industries, will amplify the pressure on computer processing requirements and on the grid. It’s why many large technology leaders are already turning to nuclear power in a bid to find green power sources.

The opposing force of ESG

While data centers gobble up ever-increasing amounts of power, ESG reporting is slowly becoming mandatory around the globe – with organizations facing hefty fines and brand damage should they fail to comply.

So, with the benefits of AI undeniable (we’re already seeing it make room for innovation, provide deeper insights, and drive highly personalized services) how can you balance AI adoption with the urgent need to mitigate the impact on our environment? 

It’s a push and a pull, as AI pushes innovation forward while ESG responsibilities stand to pull it back. How can businesses manage the two and fly freely, with ESG and AI both wings of the same plane?

With the right approach, it is entirely possible to create an energy-efficient AI architecture and sustainable data center, enabling IT leaders to achieve operational goals while meeting ESG mandates.

Data is, of course, the key here; first, optimizing data models through classification and data duplication to ensure everything is being stored in the right place – and that everything being stored is necessary. Some data can be moved off-premises and into the cloud. But, of course, as more computing moves to the Edge (and for other mission-critical data like ERP), demand for data stored on-premises to ensure resiliency and security will rise. This is where modernization comes in, with new, vastly more efficient servers and storage devices (that offer advanced file compression, for example) available on the market to reduce the overall energy needs of your data center. Look to sources like Energy Star to find solutions that rate highly (our ongoing focus in this area has paid off, with our Virtual Storage Platform (VSP) One coming first, second, and third in recent evaluations by the EPA Energy Star® rating system).

It's important to stay abreast of developments in processing capabilities too; take the ARM processors as an example. By combining the ARM Neoverse-based Grace CPU with their Blackwell GPU architecture, NVIDIA has reduced energy consumption by 25x and boosted performance by 30x per GPU compared with NVIDIA H100 GPUs.

Other physical changes can be made, too. For example, addressing the cage layout of data centers to optimize the arrangement of cabinets, racks, and servers and introduce advanced cooling systems to reduce the need for traditional, power-hungry air conditioning. These factors alone can reduce the overall data center footprint significantly, helping to bring down power consumption. You can see an example of how we did it ourselves in our Denver data center, resulting in a reduction in the data center footprint from 180 to 74 cabinets, and operating cost savings of around 40%.

Looking beyond the data center to power providers, a move to renewable energy sources will also immediately boost ESG credentials.

A long-term outlook is essential

A long-term outlook will set organizations on a winning path when setting out their AI strategy, where sustainability considerations are central to every decision. This will avoid the risk of needing to completely rearchitect the data center further down the line when ESG regulations and mandates come more fully into force. Not to mention, the cost savings of a modernized data center are significant, along with learnings made through the modernization process that will speed up problem-solving and make employees more empowered and knowledgeable moving forward.

Conclusion – three steps to unlocking optimal, sustainable AI

1. Data optimization to ensure efficient storage without compromising access
2. Data center modernization, reducing the physical footprint, and swapping in efficient solutions
3. Consider your energy sources, looking to renewable energy and working with sustainable, environmentally conscious suppliers and partners where possible

As Joe Morris, CIO of Hitachi Vantara says of unlocking greater sustainability: “Don’t accept the status quo. The only way you can invoke change is to challenge the way things have been done in the past. Get outside your comfort zone.”
 

SOURCES: ¹ State of Data Infrastructure Global Report 2024 - How AI is Shifting Data’s Foundation