Sustainability has been a buzzword in the digital economy for years now. Data centers have been using Power Usage Effectiveness (PUE) to measure energy efficiency since 2006, according to The Green Grid. Google followed this introduction of PUE becoming the first major cloud player to report carbon neutrality in 2007.
The past year, however, has helped clarify just how much sustainability really matters to the market. The explosion in AI applications has pushed the demand for compute — and thus, the demand for energy — higher than anyone anticipated, thereby creating challenges for datacenter operators to meet published sustainability goals, particularly related to reducing Green House Gas (GHG) emissions by 2030, and for many, achieve pledged Net-Zero emissions by 2050.
With the ramp of generative AI, 2024 has demonstrated that consumption of electricity in data centers isn’t slowing down. And with forecasts of AI compute infrastructure scaling at 23.5% through 2030 according to Grand View Research, hyperscalers, enterprises and other stakeholders are scrambling to recalibrate their expectations for energy efficient data center management.
Even so, sustainability commitments have been made by many operators — and the pressure to live up to promises across energy utilization and carbon emissions remains. Regulators like the European Union are applying more pressure, already implementing new rules that set higher sustainability standards for data centers placing pressure on infrastructure providers of every type to improve efficiency and sustainability.
With the growing demand for AI at odds with sustainability goals, we’re left asking: Now what? The impact of AI
It’s first worth assessing just how much AI-driven electricity demands are truly impacting the feasibility of various sustainability measures.
AI-powered applications for advanced Large Language Models (LLMs) use a significant amount of energy in both training and inference. In fact, in a report published earlier this year, the International Energy Agency (IEA) noted that ChatGPT 4o, OpenAI’s massively popular LLM virtual assistant, uses on average 2.9 watt-hours of electricity every time it answers a user request. By comparison, a typical Google search takes just 0.3 Wh.
On a global scale, the impact of this change in technology is daunting. In 2022, data centers around the globe consumed 460 terawatt-hours (TWh) of electricity, accounting for 2% of the world’s energy usage. But by 2026, that figure could more than double, surpassing 1,000 TWh,
the IEA forecast in its report. That sharp increase is attributed in large part to the growth of power-hungry AI applications as well as cryptocurrencies.
Unless data centers can significantly increase their use in renewable energy sources, that increase in energy consumption ultimately equates to an increase in carbon emissions. The major hyperscalers have already acknowledged as much.
In 2021, Google committed to reaching net-zero emissions across all of its operations and value chain by 2030. However, Google’s 2024 environmental report says that, due to the energy demands of AI, the company now views this goal as “extremely ambitious.” Due largely to the increased energy usage of its data centers, Google’s total greenhouse gas emissions have spiked by 48% since 2019, the report says. In 2023, Google emitted 14.3 million metric tons of carbon dioxide pollution, 13% year-over-year increase.
Similarly, Microsoft reported in its own annual environmental sustainability report that its greenhouse gas emissions increased by nearly 30% from fiscal year 2020 to FY 2023. Other companies with large data center footprints have a similar story.
The response so far: Stay the course
Statements from major firms suggest that, while sustainability remains a strategic objective, they will not let it impede the growth of AI workloads.
In its environmental report, Google said it expects its total greenhouse gas emissions to continue to rise before falling again.
“As we further integrate AI into our products, reducing emissions may be challenging due to increasing energy demands from the greater intensity of AI compute, and the emissions associated with the expected increases in our technical infrastructure investment,” the report said.
Earlier this year, Microsoft president Brad Smith took a similar tone in an interview with Bloomberg: “In 2020, we unveiled what we called our carbon moonshot. That was before the explosion in artificial intelligence. So, in many ways the moon is five times as far away as it was in 2020, if you just think of our own forecast for the expansion of AI and its electrical needs.”
Meanwhile, regulators around the globe are taking note of the way AI is impacting sustainability efforts. For instance, earlier this year, US lawmakers in the House and Senate introduced legislation that would set clear standards and voluntary reporting guidelines to measure AI’s environmental impact.
“There is a Dickensian quality to the use of AI when it comes to our environment: It can make our planet better, and it can make our planet worse,” Sen. Ed Markey, a Democrat from
Massachusetts, said in a statement. “The development of the next generation of AI tools cannot come at the expense of the health of our planet.”
In the European Union, the Energy Efficiency Directive (EED) is finally set to go into effect, requiring data centers to monitor and report their energy consumption and emissions. In Singapore, the government last year launched new standards for optimizing data center energy efficiency in tropical climate countries. Meanwhile, a group of data center companies this year formed the Asia-Pacific Data Centre Association (APDCA) to support sustainable growth in the industry.
Building for sustainability without slowing down AI
For data center developers and operators, the challenge now is to build and maintain facilities that can handle powerful AI infrastructure without backsliding when it comes to sustainability. Data centers are already increasingly leveraging renewable energy sources, but stakeholders will have to do more. This will require a multi-faceted, strategic approach and some innovative thinking.
First, data center operators are considering new sources of electricity to fuel their environments including expanded use of green power, captured power like nuclear generation, and even future technologies like fuel cell deployments. With grid availability limited across many regions of the world, large scale operators are taking upon themselves to create power at the data center site and manage out the risk of sourcing from utilities.
Power sourcing is not the only focus on data center operator minds. Many organizations are also pursuing strategic approaches to circularity. The concept of a “circular economy” refers to a holistic system intended to minimize waste and pollution. In a circular economy, products can be refurbished and kept in use, while natural resources are used in a sustainable manner. There are resources, such as tools from the CEDaCI (Circular Economy for the Data Centre Industry) project, that can help data center stakeholders make sustainability-focused decisions, such how to refurbish or properly dispose of servers.
Keeping the principles of circularity in mind, data center developers can implement sustainable design and operational practices into their facilities from the start. For instance, ZincFive nickel-zinc (NiZn) batteries offer a sustainable option for backup battery power. NiZn batteries are highly recyclable, making them a logical component of a “circular economy.”
NiZn batteries also provide an option for data center operators who want to replace existing lead-acid batteries with a safe, longer-lasting and more sustainable option. ZincFive offers a NiZn drop-in replacement for lead-acid UPS batteries. It is adapted to use the same charging system as lead-acid batteries, making the replacement process seamless. NiZn batteries have an operating life up to 3x that of lead-acid batteries, thanks to the stable, non-corroding positive nickel current collector in nickel-zinc batteries.
Additionally, ZincFive NiZn batteries’ lifetime greenhouse gas emissions are 25-50% lower than lead acid or lithium values. Furthermore, nickel-zinc batteries use common, widely available, materials, most of which can be recycled.
Meanwhile, even though AI has put sustainability goals farther out of reach, AI can also be a part of the solution. For instance, several years ago, Google said that by applying its DeepMind machine learning to data center operations, the tech giant managed to reduce the energy used for cooling by 40 percent. The right algorithmic tools can help an organization optimize its data center operations, and they can help operators predict problems before they arise.
Creating sustainable industry is never easy, but it is achievable. In the data center industry, the advent of AI has made the challenge even harder. Industry stakeholders are obligated as leaders to rise to the challenge. By leveraging renewable energy, new and innovative infrastructure solutions, and comprehensive strategic design principles, they can make it happen.
