Will AI innovate itself out of its sustainability problem?
Artificial intelligence is now in full bloom, driving transformative change across nearly every industry.
But as its presence and impact on the way we do business scales up, so does its environmental footprint.
According to Goldman Sachs’ estimates, AI will push data center power demands up by 160% by 2030, consuming up to 2-3% of overall power globally. Columbia University attributes 2.5 to 3.7% of global greenhouse gas emissions to data centers already today, with the figure poised to skyrocket in the coming years.
That is, unless we do something about it by looking at innovative options to power AI.
The connection between AI and sustainability is thankfully no longer an afterthought. Instead, it’s become a defining factor for responsible deployment of AI by companies that are leading the charge on AI, including AT&T, IBM, Salesforce and Microsoft.
The stakes are high, and AI’s future may rest on how swiftly the industry can build sustainable solutions and ensure innovation doesn’t come at an irreversible cost to our ecosystem.
The Growing Carbon Footprint of AI: Why It’s a Problem
There’s no putting the genie back in the bottle when it comes to AI.
Businesses across all sectors are pressing forward with AI applications that promise to redefine everything from customer service to logistics and management consulting.
The generative AI market is projected to grow from $40 billion in 2022 to over $1.3 trillion by 2030, a feat few other technologies have managed to achieve in the past. However, with a level of demand that is surging at these rates comes intensifying environmental costs.
We’ve become so accustomed to using services such as search and text generation over sleek user interfaces that few even consider what each click or prompt entails in real terms.
Google has estimated that each online search takes up 0.3 watt-hours worth of electricity, and the latest estimates on generating images with services like DALL-E peg one image at the same energy requirement as charging up your mobile phone.
Whatever data point we examine, one thing is clear; AI is hungry for energy, and the environment seems positioned to bear the brunt of the cost.
At the same time, competitive pressures are forcing the hand of virtually every CEO to deploy AI across their value chain, whatever the externalities involved. For many, this spells out disaster both for the long-term viability of AI and the environment.
Some, such as Saleh ElHattab, CEO of Gravity, a carbon and energy management platform, see the rise of AI as an opportunity for the grid.
“Historically, the largest energy consumers have been organizations without climate commitments. With AI, an epicenter of energy consumption will be one of the sectors most dedicated to sustainability: the tech industry,” Saleh explains.
“AI’s energy requirements will require that tech companies launch energy solutions to help reduce their consumption. Some of these solutions exist already, can be implemented in data centers, and are incredibly cost-effective. Others will require more investment and exploration. Broadly, these investments will accelerate the greenification of the grid,” he continued, before explaining how companies like Gravity are coming up with new ways to help clients navigate the complexities inherent in carbon and energy budgets and save on energy costs.
As fans of history know, this isn’t the first time the economy has had to innovate its way out of the environmental consequences of economic growth that is too lucrative to pass up.
For example, London’s notorious smog in the 19th century stemmed from coal combustion that powered the rise of the British Empire itself, while leaving a dark mark on public health and the environment.
Will history repeat itself as businesses embrace AI, or can we find solutions that allow AI to thrive without pushing the planet to its limits?
The AI Dilemma: How Energy Demands Could Threaten Sustainability
There are a growing number of experts who are ready to answer in the affirmative.
Just as deployment is picking up speed, efforts to curb AI’s environmental toll are gaining traction simultaneously, led by both innovative tech giants and fresh startups focused on sustainable models.
One giant that is taking action on both fronts is Salesforce, as Boris Gamazaychikov, Salesforce’s Head of AI Sustainability, knows well.
“We know the solutions we can deliver for customers are better when they’re more efficient and curated,” Boris began, noting that AI will play an increasingly important role in their customer-facing offering as well as internally. Salesforce and others see hope across a multitude of approaches, including in deploying smaller models that are fit for purpose rather than all-encompassing in their reach.
“With Agentforce, we deploy an ensemble of efficient, purpose-built models that deliver high performance without the heavy energy costs of monolithic general-purpose models,” Boris adds.
He also notes that Salesforce is among those pushing for “low-carbon data centers” and has even launched an educational initiative, Trailhead, to address knowledge gaps in the industry.
Boris continues on the importance of education, where he sees “a huge untapped role lies in educating companies on their AI-related emissions, ” an area where giants like Salesforce have their role to play.
Meanwhile, other industry players are investing heavily in renewable energy sources to power their data centers. Amazon Web Services has committed to 100% renewable energy by 2025, and Microsoft has implemented water positive cooling systems that drastically reduce water consumption in its servers.
One energy source that is seeing a resurgence in interest is nuclear power, since it is touted as a clean energy alternative that promises a smaller carbon footprint to offset AI’s high-energy needs.
As Deóis Ua Cearnaigh, CTO at Aeon Blue, observes, “While nuclear’s day may not be today, it is the inevitable destination. In fifty years, we’ll likely be talking about nuclear as the backbone of sustainable energy.”
Deóis’ work at Aeon Blue highlights that renewables alone may struggle to support consistent energy needs, especially as demand continues to rise thanks to AI. “The sun rises and sets, the wind waxes and wanes,” he explains, “and even with all the lithium ever mined in human history turned into batteries, we wouldn’t be able to store the U.S. grid for one hour. Nuclear offers a steady, resilient alternative to keep us powered without compromise. In the meantime, it’s carbon capture and efuels.”
Collaboration across industry players is also a solution that can no longer be avoided.
Hugging Face and Salesforce’s Energy Scores for AI Models project is a great example of exactly such collaboration in action.
“Our collaboration is about transparency,” says Boris. “Consumers need clear, simple and standardized ratings to gauge a model’s environmental impact. This is especially critical as LLMs race toward commercialization.”
The Role of AI in Solving Its Own Environmental Challenges
As the AI industry races ahead we all face a critical challenge: can we meet our growing demands for AI without compromising the environment?
The stakes are clear—deploying large-scale AI can bring transformative efficiencies and novel solutions, but not without considerable environmental costs.
Business leaders like Lan Guan, Accenture’s chief AI officer, emphasize that AI can be a source for solutions. “We’re seeing AI accomplish truly astonishing feats on behalf of our clients, just as it can also be energy intensive.”
Lan is among a growing number of experts that see AI itself as a part of the solution. “AI agents, in particular, can get to action faster and can enable more efficient energy use within an organization even when they are using non-negligible amounts of energy themselves.”One particular area where Lan sees potential for AI agents to do good is in streamlining supply chains, which often accounts for the vast majority of a company’s emissions.
“We’re seeing AI agents be able to cut down on procedural waste and reduce environmental impact while maximizing efficiency, with a net-positive impact on the environment,” Lan added.
Many others share this view.
In a survey of nearly 500 sustainability professionals earlier this year, Salesforce found that 58% believe the benefits of AI will outweigh its risks when solving the climate crisis. In October, the company launched the Salesforce Accelerator – Agents for Impact, the latest in a series of programs designed to help nonprofits deploy agents and other forms of AI to address environmental and social challenges.
The delicate balance between AI’s benefits and drawbacks, though promising, remains a work in progress. “Transparency, trustworthiness, and empathy are core tenets,” Accenture’s Lan notes, “and there’s work to do to make AI behave in ways that align with our human expectations, particularly when it comes to the environment.”
The future of AI hinges on innovation as much as companies’ commitment to sustainable growth strategies and responsible technology deployment. The journey toward sustainable AI may be complex, but with businesses focused on balancing innovation with responsibility, the industry is well-positioned to create solutions that work for both business and planet alike.
