How data centers accelerate climate change and why this needs to change
Artificial intelligence (AI) has drastically transformed the way we work, create and make decisions. On the one hand, it is bringing significant benefits to our lives, but on the other, its advancements come with a cost to the environment.
Behind every AI query, image generation and data analysis is a data center that consumes more energy with each request. As AI adoption surges, so does the energy demand. Despite promises to curb carbon emissions, data centers worldwide, including hyperscalers like Google or Microsoft, are reporting a rising carbon footprint.
If you’re an IT professional rather than an environmental activist, you might legitimately ask: “How does this concern me?” Unfortunately, the record-breaking summer temperatures, floods and wildfires not only affect the planet but also operational costs and business efficiency.
This article explores the commonly overlooked environmental impact of AI and offers actionable steps for decision-makers. Whether you’re a data center operator, cloud service provider, or a business looking for hosting solutions, you have to understand how to decrease your costs and future-proof operations.
To start with, it is important to understand the AI’s effect on energy consumption, which is the core cause of climate change. There have been numerous reports by Goldman Sachs, McKinsey and others that prove that. In its report, Goldman Sachs, in particular, quotes the International Energy Agency (IEA) that states that a GPT request needs ‘2.9 watt-hours of electricity, compared with 0.3 watt-hours for a Google search’. Another report by EPRI indicates that a single ChatGPT query consumes nearly ten times the electricity of a Google search. With over 100 million weekly users of Open AI’s ChatGPT, this demand adds up quickly.
As a result, this drives electricity consumption worldwide. According to Gartner, by 2030, AI could account for up to 3.5% of global electricity demand, a figure comparable to the current energy consumption of the global agriculture and forestry industries combined. In January, IEA also forecast that global data center electricity demand will more than double from 2022 to 2026, with AI playing a major role in that increase.
The electricity demand itself might not be such a problem if the supply comes from a renewable source. However, most data centers continue relying on fossil fuels, which translates into carbon emissions. According to this year’s reports, Microsoft’s carbon footprint has risen by nearly 30% since 2020 and Google’s emissions in 2023 were almost 50% higher than in 2019.
Recently, the Guardian has released a comprehensive analysis showing that ‘from 2020 to 2022 the real emissions from the “in-house” or company-owned data centers of Google, Microsoft, Meta and Apple are probably about 662% – or 7.62 times – higher than officially reported.’
On the surface, the situation might look much more positive, with widespread claims by hyperscalers like AWS or Google about having achieved or almost having achieved “net-zero carbon emissions.”. However, this mostly comes from renewable energy certificates (RECs), which allow companies to invest in green energy without actually using it instead of fossil fuels. While this might facilitate green energy production, it has zero impact on the actual carbon footprint reduction.
Therefore, despite the claims and promises about net-zero neutrality, in reality, the carbon footprint is large.
So far, we have established two links. Firstly, AI has a significant impact on data center energy consumption. Secondly, despite the promises, this energy consumption translates into growing carbon emissions.
Now, let’s explore how these things affect data centers and operational costs beyond the climate impact itself. The logic is straightforward. Firstly, larger demand for electricity drives bigger operational costs. But there is more. The more intensively servers work, the bigger the need for cooling systems to prevent hardware from overheating, which also requires electricity. According to McKinsey, cooling systems account for at least 40% of data center energy consumption, which further drives operational costs up.
Since the power mostly comes from non-renewable sources, the increase in AI adoption can directly be linked to the increase in carbon footprint, leading to rising temperatures. These, in turn, lead to other unfortunate events for businesses. In 2022, extreme heat in California already caused operational challenges in X’s (ex-Twitter) data centers. That same year, Google Cloud’s data center in London went offline for an entire day due to cooling failures.
Therefore, as temperatures rise, we can expect more frequent service disruptions, downtime and higher maintenance costs. In addition, as people are becoming more affected by the effects of climate change, we might anticipate calls for stricter regulations on emissions or carbon taxes, which eventually will also affect data centers and businesses.
The above conclusions demonstrate that genuine sustainability is in the best interest of any business dealing with IT. For data center operators, achieving sustainability leads to lower operational expenses, while for service consumers, it means lower service costs, more stability and less downtime.
Legislators and large market players have to act on a global level. However, several actionable tips can help reduce the environmental impact of data centers right now. Data center service providers could consider implementing these, while businesses looking for colocation can use them in their decision-making process of choosing an optimal data center service provider.
One of the most effective and cost-efficient ways to reduce your carbon footprint is to choose data centers in regions that rely on renewable energy or natural cooling. In this case, your or your provider’s energy costs will be close to zero, as well as the carbon footprint.
Iceland, for example, is probably the world’s leader in green energy. All 100% of its electricity comes from renewable sources, primarily hydro and geothermal power, accounting for 70% and 30% respectively. Moreover, Iceland’s cold climate reduces the need for artificial cooling, since data centers like atNorth or Borealis, rely on natural cooling. This provides both environmental benefits and saves costs. Another leader in this is Switzerland with their renewable sources and strict regulations.
Choose Hardware Parts with Lower Power Consumption
Innovative hardware can also seriously impact energy consumption. This may require larger upfront investments, but these expenses will pay off with lower energy use, higher efficiency and improved sustainability.
The latest innovations include ultra-low-power AI semiconductor chips that are being developed in South Korea, that are designed to deliver high performance with minimal energy use. Other energy-efficient options, AMD’s EPYC CPUs and Instinct GPUs, also claim to consume less energy per operation. Therefore, it’s worth spending time and effort researching the hardware components that you or your potential service provider plan to use.
Power Usage Effectiveness (PUE) is a key metric that measures the data center’s energy efficiency. The closer a PUE is to 1, the less energy is wasted. Therefore, checking the provider’s PUE is an easy way to contribute to lowering the carbon footprint.
Unfortunately, many data centers are quite far from that standard. A 2023 survey by the Uptime Institute reflects the industry’s average PUE at 1.58, but the more efficient ones are not so hard to find. For example, data centers in regions like Iceland can offer PUE as low as 1.05 or 1.03.
As AI and data center demand continues to rise, the environmental impact of these technologies cannot be ignored. Decision-makers must take a proactive approach when choosing data center services or building their own infrastructure for consumers. By choosing sustainable DC locations, opting for energy-efficient solutions and advanced hardware, businesses can not only protect the planet but also secure long-term cost savings and regulatory compliance.
Sustainability and profitability are not mutually exclusive. Companies that act now will be better prepared for the future, while those that wait risk paying a higher price, both financially and environmentally.
You May Also Like
