As artificial intelligence transitions from an experimental novelty to the backbone of global corporate operations, a quiet crisis is brewing in the shadows of the digital revolution. While the efficiency gains of AI are widely celebrated, the physical reality—the massive, power-hungry, and water-intensive infrastructure required to sustain it—has remained largely obscured from the corporate balance sheet.
In response to this growing disconnect, two prominent voices in the climate and technology sectors have joined forces. Boris Gamazaychikov, a former AI and sustainability manager at Salesforce, and Dr. Sasha Luccioni, a renowned climate scientist and the former AI and climate lead at Hugging Face, have launched the Sustainable AI Group. This new research and advisory firm aims to move beyond the industry’s hype, providing sustainability professionals with the tools to demystify, measure, and mitigate the environmental footprint of their AI initiatives.
The Genesis of a New Mission: The Chronology of an Urgent Need
The formation of the Sustainable AI Group is the culmination of years of individual advocacy and collaborative research. The seed of the project was planted long before the recent generative AI explosion, rooted in the realization that the corporate world was sprinting toward automation without a map for its ecological consequences.
Early Observations
Dr. Luccioni’s journey began at financial services giant Morgan Stanley, where she observed a widening chasm between the rapid, aggressive adoption of AI technologies and the firm’s commitment to climate targets. The disconnect was not just a failure of policy, but a fundamental lack of visibility. Driven by a desire to research how AI could be leveraged in a "climate-positive" manner, she eventually left the financial sector to join Hugging Face. There, she became one of the first researchers to rigorously quantify the outsized energy consumption of large language models.
The Salesforce Connection
Concurrently, Boris Gamazaychikov was navigating the complexities of implementing AI within the enterprise environment at Salesforce. He witnessed firsthand the confusion among stakeholders. While companies were eager to integrate AI into their day-to-day processes, their sustainability departments were left in the dark, struggling to calculate emissions or assess water consumption. "It became apparent that unlike other sustainability topics that I’ve touched in the past, this is moving at such a rapid speed and customers are feeling really disempowered," Gamazaychikov noted.
Collaboration and Formalization
Before launching the firm, the pair collaborated on a series of critical industry resources. Their partnership produced the AIEnergyScore, a resource hosted on Hugging Face that allows for the comparison of energy efficiency across different models. They also authored essential primers for procurement teams, detailing the specific questions organizations must ask vendors regarding the power and water resources required to fuel their digital ambitions.
Demystifying the "Saber-Toothed Tiger" of Data Centers
A central tenet of the Sustainable AI Group’s philosophy is the need to bridge the psychological distance between the end user and the data center. Dr. Luccioni often uses a powerful analogy to explain why the environmental impact of AI is so easily ignored.
"I think that most people don’t realize to what extent the AI that they use, that we use, doesn’t run locally," Luccioni explained. "All of this is running in data centers, and all these data centers are so far away from us. As human beings across the millennia, we’ve been focusing on immediate threats to our safety, like the saber-toothed tiger that jumps out of the bush. Data centers are the saber-toothed tigers that are very, very far away from us."
This distance is not merely geographical; it is operational. Most SaaS (Software-as-a-Service) providers do not explicitly report the carbon intensity of the specific AI queries they perform, leaving corporate sustainability leads unable to account for their "Scope 3" emissions—those indirect emissions that occur in a company’s value chain.
Supporting Data: Why Visibility Matters
The environmental cost of AI is multifaceted, involving three primary resource drains:
- Energy Consumption: Training a single large language model can consume as much electricity as hundreds of homes do in a year. Beyond training, the "inference" phase—where the model processes daily user queries—creates a continuous, compounding demand for electricity.
- Water Usage: Data centers require immense amounts of water for cooling systems to prevent hardware from overheating. In drought-prone regions, the water consumption of a single major data center can compete with the needs of local communities.
- Hardware Lifecycle: The rapid obsolescence of high-performance GPUs (Graphics Processing Units) creates significant e-waste, adding a layer of material impact that is often omitted from standard sustainability reporting.
The Sustainable AI Group intends to address these metrics by creating standardized reporting frameworks that force transparency. Their goal is to transition the conversation from "AI is magic" to "AI is industrial."
Implications: The Path Toward Accountability
The Sustainable AI Group is positioning itself as a vital consultant for enterprises reaching "mainstream maturity" in their AI adoption. As investors and employees demand greater transparency regarding corporate ethics and sustainability, organizations can no longer afford to treat their digital carbon footprint as an externality.
Empowering Sustainability Professionals
Gamazaychikov believes the most effective path forward is for sustainability departments to become active participants in the technology procurement process. "That actually helps aggregate the signal and push the software-as-a-service providers into really demanding this from their AI vendors," he argues. By asking pointed questions about where AI models are hosted and how their energy consumption is audited, firms can create a market incentive for cleaner, more efficient AI infrastructure.
A New Standard for ESG
The firm plans to expand its library of resources, focusing on sectors where AI is deeply embedded in workflows. This includes everything from supply chain management to investor analysis. By providing a clear methodology for assessing the sustainability of AI, they hope to integrate these concerns into the standard Environmental, Social, and Governance (ESG) criteria used by major corporations.
The Road Ahead: Industry Integration
The movement toward sustainable AI is gaining momentum beyond the work of the Sustainable AI Group. Events like Trellis Impact 26 have begun to place the intersection of AI and sustainability at the center of their programming. The upcoming event will feature key industry leaders—such as Microsoft’s Jim Hanna—to discuss the evolution of sustainable data centers.
Furthermore, the "AI x Sustainability Program" will delve into the practical applications of this research, including how AI can be used to monitor supply-chain deforestation and improve the quality of carbon credit verification. These sessions reflect a nuanced reality: AI is not inherently the enemy of sustainability; rather, it is a tool that requires rigorous, science-based management to avoid doing more harm than good.
As Gamazaychikov and Luccioni move forward with their advisory firm, their message remains clear: the era of "black box" AI operations is ending. The future of technology will be defined by those who can successfully marry the power of artificial intelligence with the planetary boundaries of our physical world. For the sustainability professional, the time to start asking questions is now—before the "saber-toothed tiger" of digital demand becomes impossible to ignore.
Heather Clancy is a veteran journalist whose reporting on technology and sustainability has been featured in The New York Times, Fortune, and The International Herald Tribune. Her work continues to highlight the leaders and innovators shaping a more transparent, sustainable future.
