When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
The dawn of generative AI has been a springboard for wild AI adoption among businesses and individuals at various scales.
While we can argue for artificial intelligence‘s positive impacts and growing adoption, the same cannot be said about its environmental impact.
The amount of energy needed to train and run leading AI systems has increased by a factor of 350 million in the past thirteen years, according to research by the Center for the Governance of AI.
This compute demand has enabled major AI advances such as large language models (LLMs), protein folding, and autonomous vehicles, the research found.
The immense computational power required by AI translates to massive data centers, guzzling electricity and leaving a worrying environmental footprint.
Recognizing this environmental concern, Salesforce has called for AI emissions regulations that would force companies to disclose their AI emission. The San Francisco-based software maker believes the growing adoption of AI is pushing tech sector’s carbon footprint to a disturbing level.
They advocate for the establishment of standardized metrics for measuring and reporting the environmental impact of AI systems.
Techopedia sat with Salesforce’s Director of Climate & Energy at Salesforce, Megan Lorenzen and Senior Manager of Emissions Reduction, Boris Gamazaychikov, to discuss the environmental and financial challenges posed by AI’s energy demands, along with possible mitigation strategies.
About Megan Lorenzen and Boris Gamazaychikov
As Director of Climate & Energy at Salesforce, Megan Lorenzen leads Salesforce’s global climate policy program, advocating for policies that set the world on a path to a just transition to a 1.5-degree future. Megan also leads the company’s power sector decarbonization efforts.
This includes catalyzing new, high-impact, renewable energy projects to maintain Salesforce’s global 100% renewable energy commitment and deploying the company’s cross-functional energy transition strategy.
Boris Gamazaychikov is the Senior Manager of Emissions Reduction at Salesforce. His work involves implementing carbon reduction strategies across Salesforce’s value chain.
He works at the intersection of sustainability and innovation, having developed and executed Salesforce’s sustainable AI strategy.
He also works closely with suppliers, customers, and partners on their journeys, publishing tools and resources to help accelerate climate action.
Prior to Salesforce, Gamazaychikov consulted diverse global clients, including the world’s largest technology companies, on their decarbonization strategies. Before this, Gamazaychikov spearheaded the Pentagon’s sustainable construction program. He received his degree in Environmental Engineering from the University of Maryland.
The Environmental Cost of AI
Q: What are the specific concerns around AI’s environmental impact, and what’s Salesforce’s role in all of it?
Gamazaychikov: Generative AI models are powered by algorithms that are orders of magnitude larger than traditional machine learning models. Training and deploying these massive models require immense compute resources and massive data centers powered by electricity, often not strictly from clean energy sources.
Data centers currently account for a few percent of global electricity consumption and emissions, but research suggests AI could potentially triple or even quadruple the energy needs of data centers.
In certain regions with concentrated data center construction, the environmental impacts could be even more severe.
That is why we are calling for transparency and sustainable integration of AI. At Salesforce we have “AI + data + trust framework” that ensures ethical considerations and environmental stewardship are paramount in AI development.
Minimizing AI Environmental Footprint
Q: Data centers are a significant energy consumer for AI. What strategies can be implemented to minimize the environmental footprint of AI infrastructure?
Gamazaychikov: At Salesforce, we start by ensuring our models are right-sized for their intended use cases, avoiding unnecessarily large and energy-intensive models. We’ve developed domain-specific models tailored for our CRM use cases, which are far more efficient than off-the-shelf massive models.
Additionally, we consider the energy efficiency of the hardware and data centers used for training and deployment.
GPU efficiency has improved 36-fold in seven years, so using the latest efficient hardware is crucial. Mapping the carbon intensity of data center electricity sources and favoring lower-carbon options can significantly reduce emissions.
For companies not directly training models, the key is demanding transparency from AI providers on energy efficiency and carbon footprints – treating it like fuel efficiency for a new car purchase.
Standardizing AI Sustainability Metrics
Q: What are the standardized metrics for measuring the sustainability of AI operations?
Gamazaychikov: There aren’t uniform standards yet, but efforts are underway. The AI firm Hugging Face, a Salesforce investment, recently proposed an “Energy Star for AI models” rating system.
We’re exploring piloting this with our models to contribute to developing a simple, universal efficiency metric that could guide adoption across industries and potentially inform regulations.
Lorenzen: In addition to that, the lack of consistent standards is why we’ve advocated for legislation like the U.S. AI Environmental Impacts Act, which would establish a consortium to build guidelines for transparency in measuring and disclosing AI’s environmental impact.
Beyond the Tech Sector’s Responsibility
Q: Beyond tech companies, what other sectors can play a huge role in the push for a more sustainable AI ecosystem?
Lorenzen: This is a global opportunity to ensure AI breaks down climate barriers and drives equitable outcomes. While minimizing AI’s environmental impact is critical, we must also unlock its potential for innovation and advancing sustainability objectives.
Salesforce recently concluded our first AI Accelerator, supporting NGOs leveraging AI for increased efficiency and climate action.
All industries can play a role, whether through policy advocacy, trailblazing use cases, or prioritizing transparency from AI providers.
Balancing Costs and Sustainability
Q: Sustainable solutions sometimes come at a premium. How can organizations find a balance between financial considerations and environmental impact when deploying AI?
Gamazaychikov: Interestingly, in AI, sustainability and cost-efficiency are often aligned. Smaller, more efficient models reduce energy consumption and carbon emissions while also lowering computational costs and improving response latency – critical for product experience.
The one potential trade-off is the carbon intensity of a data center’s electricity grid, which varies by region.
Lorenzen: We frequently see efficiency improvements that deliver both sustainability and cost benefits.
However, we must consider the full system, like enhancing grid transmission infrastructure to support more low-cost renewables.
Alleviating grid congestion can reduce consumer costs while enabling greater renewable penetration and emissions reductions. It’s an interconnected challenge requiring a holistic approach.
Q: Megan, can you share what your average day looks like?
A: My days are quite full, often waking up to the beautiful Colorado mountains that inspired my passion for sustainability work. I split my time across the energy transition – thinking about our 100% renewable energy portfolio deployment, what’s next beyond just purchasing renewables to fully decarbonize the grid, and the policies and standards needed to enable further progress.
This global work is underpinned by supporting teams engaging lawmakers in different regions on crucial enabling policies.
While no two days are alike, they are packed with efforts to leverage Salesforce’s full influence to accelerate the just energy transition to a 1.5°C future through sustainable operations, systems change advocacy, and unlocking climate innovation opportunities. The views are nice, but the work is demanding yet impactful.
Q: If you could have a drink with Mark Zuckerberg, Elon Musk, or Jeff Bezos, who would you choose?
Lorenzen: I’d choose Jeff Bezos for both personal and professional reasons. Personally, as an Amazon customer concerned about the environmental impact of e-commerce deliveries, I’d love to discuss how consumers can reduce their footprint.
Professionally, given Amazon Web Services‘ massive data center footprint, engaging with Bezos on their renewable energy strategy could have a tremendous global impact on decarbonizing the cloud computing industry.
The Bottom Line
The rise of generative AI poses sustainability challenges and opportunities. As Salesforce’s energy and climate leaders emphasize, proactive strategies and collaboration across sectors are critical to curbing AI’s environmental toll while tapping its potential as a catalyst for climate solutions.
Embedding sustainability into AI’s core is paramount — right-sizing models, prioritizing energy efficiency, leveraging renewables, and mandating emissions transparency. Meanwhile, standardized metrics are crucial for guiding eco-conscious adoption.
However, the onus extends beyond tech, with policymakers cultivating an enabling environment, businesses advocating sustainable practices, and all stakeholders embracing a planet-friendly digital transformation mindset, we can come close to reducing AI’s footprint on the environment.
