Every day, we hear how artificial intelligence is changing our lives. It promises to save effort and time on a scale never before witnessed in human history. But to quote Serena Williams: “Everything comes at a cost. Just what are you willing to pay for it?” In the case of AI, the cost to all living species on planet earth may be more than we anticipated.

Technological advances have historically led to significant, and perhaps unforeseen, negative impacts on society and the environment. The Industrial Revolution, while a period of immense technological and economic progress, brought about poor working conditions and child labor. The introduction of the automobile revolutionized transportation but also resulted in urban sprawl. The development and widespread use of plastic resulted in massive pollution problems, with plastics now contaminating oceans and affecting marine life on a global scale.

The exponential growth of AI applications has catalyzed a surge in energy consumption and carbon emissions, primarily stemming from the intensive computational requirements of AI processing. AI algorithms, particularly deep learning models, demand vast resources, necessitating the operation of data centers equipped with high-performance servers and cooling systems. The continuous training, inference and optimization processes associated with AI tasks result in prolonged utilization of server infrastructure, driving up electricity consumption and carbon emissions. According to Professor Scott Galloway, professor of marketing at NYU’s Stern School of Business, one ChatGTP request requires 10 times the energy of a Google search. In five years, the incremental energy demand of artificial intelligence will be equivalent to 40 million homes — more than California, Texas, Florida, and New York combined.

As organizations harness AI to analyze vast datasets, automate tasks and optimize operations, the environmental impact of AI processing cannot be overlooked. Efforts to mitigate the emissions increase from AI processing are essential to reconciling technological advancement with environmental sustainability.

The proliferation of AI comes at a time when laws aimed at limiting carbon emissions in supply chains are increasing. The European Union's Green Deal includes regulations such as the Corporate Sustainability Reporting Directive (CSRD), which mandates large companies to disclose their environmental impacts, including carbon emissions generated by supply chain partners, also known as Scope 3 emissions. 

Scope 3 encompasses indirect greenhouse gas emissions that occur throughout a company's supply chain, including both upstream and downstream activities. For companies relying on AI, a significant portion of Scope 3 emissions originates from data centers and server infrastructure used to support AI applications. The energy-intensive nature of AI algorithms and data processing tasks results in heightened electricity consumption, driving up carbon emissions associated with server operations.

As companies strive to meet sustainability targets and address climate change, it becomes an imperative to understand and mitigate Scope 3 emissions from server suppliers. Failure to account for these emissions not only undermines corporate sustainability goals, but also exposes companies to reputational risks and regulatory scrutiny.

Businesses should engage with their server suppliers to gain transparency into their energy sources, infrastructure efficiency and carbon footprint. Open dialogue and collaboration can promote the sharing of best practices and encourage suppliers to adopt energy-efficient technologies and renewable energy sources. Contractual agreements should prioritize sustainability metrics and incentivize emissions reductions to align suppliers with corporate sustainability goals. 

By conducting lifecycle assessments of server infrastructure, companies can pinpoint inefficiencies and emissions hotspots across the supply chain, focusing on hardware production, data center operations, and end-of-life disposal. Optimization and emission-reduction opportunities can be identified, such as server consolidation, virtualization and data center efficiency improvements.

Going all-in on AI is a good bet for just about every company. Become aware of the environmental impact of your server infrastructure, and take steps to manage the Scope 3 emissions they will generate. When signing up new server vendors, dig into how they’re addressing their emissions. Include questions about emission-reduction plans in your requests for proposals. Responsible practices such as engaging with suppliers, optimizing lifecycle processes and investing in renewable energy can reduce carbon emissions, boost innovation and create long-term value. 

Justin Dillon is founder and chief executive officer of FRDM.