Can AI help manage the new threat to our environment caused by AI?

Can we make data centers smart and green?

Warning about AI’s electricity consumption speed was being talked about a few months ago, but suddenly all is quiet on that front. Even as we worry and work towards a greener cleaner future, AI innovations are demanding huge amounts of computational abilities causing strain on the world’s green efforts.

In May, the MIT Tech Review revealed some heart-rending stats. The energy needed to create even a low-quality, five-second video is 42,000 times more than the amount required for a chatbot answer a question about a recipe, enough to power a microwave for more than an hour.

AI’s rapid growth has created an uncomfortable paradox, the very technology hailed as the key to a sustainable, intelligent future is now one of the biggest threats to that vision. Yet, within this tension lies opportunity.

According to CNBC, AI demand could strain electrical grid in coming decade, causing a power supply crunch for the US.

Till last year, investors for big tech, who manage hundreds of billions of dollars, were pressing Microsoft, Alphabet and others for more information on the power needed for AI and advanced computing, to help decide whether the sector should stay heavily represented in sustainable funds, investors said.

According to Google’s numbers, AI has vastly increased its emissions over the past five years.
Last year, the tech giant said its greenhouse-gas emissions totaled 14.3 million metric tons of carbon dioxide equivalent throughout 2023, 48% higher than in 2019. The reason is mostly due to Google’s enormous push toward AI, which will likely make it harder to hit its goal of eliminating carbon emissions by 2030.

In August, for the first time, Google released data on how much energy an AI prompt uses. Median queries consume 0.24 watt-hours, similar to a microwave’s one-second use. The company also provided estimates for water consumption and carbon emissions.

In the AI Action Summit in Paris, politicians and AI leaders from around the globe aired concerns about AI safety. Though the discussion was more about deregulation and energy.

We now know that generating one image with generative AI uses as much energy as charging a smartphone. AI is already wreaking havoc on global power systems.

OpenAI’s CEO Sam Altman has claimed AI will deliver an “Intelligence Age”, leading to “unimaginable” success and “astounding triumphs” such as “fixing the climate.” However, are tech breakthroughs all that can solve global warming? In fact, as it stands, AI seems to be making the problem much worse.

AI Boom = Data Center Boom

Meanwhile, the AI market is bursting at the seams. Google-parent Alphabet’s cloud sales last July-September quarter was rosy for top cloud providers Microsoft and Amazon.com, indicating that the market for AI-aided computing power is only growing.

What this means is that the need for bigger data centers, that drink electricity to run. According to Bloomberg, these data centers are starting to change the landscape of places like Northern Virginia, Loudoun County, where data center building has gone into overdrive because of AI. Data centers use more electricity than most countries. As Arm Holdings Plc CEO Rene Haas told Bloomberg, by 2030, the world’s data centers are on course to use more electricity than India, the world’s most populous country.

Microsoft, Alphabet, Amazon.com and Meta Platforms are set to spend $320 billion this year on data centers and other kit to power advanced chatbots. Last year, Microsoft made a deal to help restart Three Mile Island, a once-shuttered nuclear plant. The US$16 billion deal between owner Constellation Energy and Microsoft set to power the tech giant’s AI data centers.

Meanwhile, AI chips are not only expensive, but also we don’t have enough to cover today’s demand brought on by the AI explosion. Training an LLM takes the same amount of energy as the annual consumption of over hundred US homes, and generating an image with generative AI uses as much energy as charging a phone.

Ironically, AI Can Help

Are we looking to curb AI’s hunger for power? Will we find a solution? Ironically, AI could find that solution for us.

Gen AI can help in strategizing with energy storage. As per a study from Juniper Research, whilst AI is used extensively for grid automation processes, GenAI is providing additional features. Generative models trained on customer energy data can create scenarios for utilities to develop future grid strategies.

For instance, calculating energy output requirements based on houses adopting solar technologies allows utilities to plan future grid investments. GenAI allows utilities to enhance grid efficiency, so technology companies must integrate solutions into their offerings before market saturation occurs.

Efforts are on in inventing more sustainable ways of using AI. Sam Dillavou, a physicist at the University of Pennsylvania, is trying to find a low-power alternative to the energy-consuming GPU chips extensively used in machine learning. 

What if Data Centers Became Smart & Green?

In fact, as the market value of industrial smart building deployments is predicted to grow (95% to $14 billion by 2026 globally) AI-based building management solutions will be key to securing a return on investment in smart buildings platforms, whilst accomplishing sustainability and energy goals. The benefits that AI brings to smart building platforms is expected to enable global spend to exceed $42 billion by 2028.

One step in the direction of a solution could be making these data center buildings smart and green.

In fact, experts are saying that automation is the answer for cost reductions and sustainability, and service providers will do well to implement AI at the core of their portfolios to reap the maximum benefits of industrial smart building solutions. AI detects operative anomalies through the use of in-depth data analytics, resulting in increased safety and decreased operational costs, whilst reducing the extent of required human intervention.

Smart building solutions will also enable operations managers to meet sustainability goals. By integrating IoT networks and sensors with AI systems that automate building functionality in real-time, significant savings and emissions reductions can be achieved.

AI’s rapid growth has created an uncomfortable paradox, the very technology hailed as the key to a sustainable, intelligent future is now one of the biggest threats to that vision. Yet, within this tension lies opportunity. By applying AI’s predictive and optimization capabilities to its own energy challenge, from smarter data centers to AI-managed grids and low-power chips, we can turn the problem into part of the solution. The path forward will depend on how quickly innovation shifts from scaling intelligence to sustaining it. The question is no longer can AI help save the planet, but how soon it can learn to save itself first?