Raphael Thys
When data was stored on paper, and then on punch-cards, tape, floppy disks and whatever other medium long since faded from memory, it wasn’t hard to recognize the resources needed to keep, process and access it.
But now that data is mostly out there, in the “cloud,” it seems easy to imagine that it’s “clean” and comes at little cost of resources other than the monthly bill from the company that runs your servers. Time to think again. As it turns out, data storage, access and processing is emerging as one of the key environmental bad actors of business, including supply chain management.
The trouble is that technological innovation just seems to increase the amount of data we generate and store. Sure, it’s fantastic that there’s more computing power in your washing machine than on the Apollo 11 spacecraft that landed on the moon. But it comes at a cost. Cryptocurrency activity, which is pure data, consumes the same amount of electricity as the country of Argentina, and in the U.S. alone is estimated to emit from 25 to 50 million tons of CO2 each year, on a par with the annual emissions from diesel fuel used by U.S. railroads.
Sometimes the costs are surprising. A researcher at the University of California, Riverside, recently calculated that, every time you ask ChatGPT a question, it gulps up 500 milliliters of water (close to what’s in a 16-ounce water bottle). And in its latest environmental report, Microsoft disclosed that its global water consumption spiked 34% from 2021 to 2022 (to nearly 1.7 billion gallons, or more than 2,500 Olympic-sized swimming pools), a sharp increase compared to previous years that outside researchers tie to its artificial intelligence research.
And, while AI offers some compelling tools for battling climate change, such as monitoring deforestation, identifying pollution leaks and tracking extreme weather events, a new report by green groups has cast doubt over whether the AI revolution will have a positive impact upon the climate crisis. Instead, they warn, the technology will spur growing energy use from data centers (and also the proliferation of falsehoods about climate science).
They’re probably right. According to Egnyte, a vendor of cloud-based content security, compliance and collaboration software, data storage requirements in the architecture, engineering, and construction industry have risen from an average of 3.34 terabytes in 2018 to 25.64 terabytes in 2023, representing a compound annual growth rate of 50.3%.
Further, the explosion in data is bumping up against the ability to store it. Newmark’s 4Q23 U.S. Data Center Market Report, which dissects AI’s role in fueling all-time highs in data center demand and development, highlights major gaps in existing infrastructure as data center power consumption is expected to reach 35 gigawatts (GW) by 2030 — up from just 17 GW in 2022. Rather than taking a rocket to the moon, it seems we’re heading for a data overflow Niagara Falls.
“It’s a maturity problem,” says Kayla Broussard, chief technology officer — consumer & travel market at Kyndryl, which bills itself as the world's largest provider of IT infrastructure services. Kyndryl has its own sustainability goals (to be carbon net-zero by 2040). But they also help users of their services pay heed to what data is costing them and the environment. “It’s a lack of employee education and engagement. Most people know about their company’s sustainability goals, but what’s missing is the employee’s direct linkage to them as an individual. And it’s not just IT. It’s HR. It’s accounting. It’s everything.” Ironically, this is partly due to employees not having access to the data that would allow them to drill down into the issue at a sufficiently granular level, Broussard argues. Companies need to implement real metrics and deploy dashboards that their actual employees can understand. “You want them to know: What are the consequences of me initiating that generative AI project?” says Broussard. Kyndryl offers tech tools that show employees such costs. “A side benefit is that the organization can now actually quantify the positive impact driven by their workforce,” adds Broussard. “Because, in the end, sustainability goals fall on the shoulder of all employees.”
Lack of awareness, or a strategic plan, is notable. While 85% of organizations place a high strategic level of importance on achieving their sustainability goals, only 16% have integrated sustainability into their strategies and data, according to The Global Sustainability Barometer study commissioned by Kyndryl and Microsoft, from research firm Ecosystm, released in December, 2023.
While CEOs and boards have made sustainability a priority, they need help with execution. For example, the Barometer study found that 61% of organizations are using AI to monitor energy usage today, and only 34% use current data to predict future energy consumption. “That’s not enough,” says Broussard.
The good news is that there are ways to drastically streamline data storage and management. Fundamentally, most companies’ data storage is a mess, with duplications, unnecessary silos and software system integrations held together with multiple work-arounds that just aren’t efficient. Kyndryl knows only too well how easy it is to carry on oblivious to these issues. When the company spun off from parent IBM in 2021, they found they had data scattered over 100 disparate warehouses, Broussard laughs. “And we didn’t realize, until we had a hard deadline to get out.” The experience was typical. Multiple teams had their own copy of the data. “We didn’t just unify it. There was a critical focus on sustainability,” says Broussard. Putting a bead on preserving resources allowed Kyndryl to reduce their use of data warehouses by 90%. “It went hand in hand – our consolidation of data centers and a focus on green energy, and green IT in general.”
Sometimes, it’s as simple as figuring out what data you’re not going to need immediate access to any time soon, and put that in a more efficient form of data storage that means it takes longer to summon, but uses a lot less electricity and water. “You should consider taking data you haven’t touched in five years to offline storage rather spinning-disk (storage),” Broussard says.
But awareness and accountability at the granular level are absolutely critical, says Broussard. “It’s not enough to do a quarterly emissions report. You should report emissions per gigabyte of data and even per employee,” Broussard advises. “Let your employee see their own metric. Let them see the cost of generating a prompt, or making a copy of some data.”
“Data is exploding. It’s the new natural resource, and AI is exacerbating this rise,” says Broussard. “By understanding the carbon footprint of data, you can start decoupling carbon and growth.”