Salesforce is spearheading a new AI Energy Score benchmarking tool.
Launched in collaboration with Hugging Face, Cohere, and Carnegie Mellon University, the solution allows AI developers and users to assess, compare, and understand the energy usage of different AI models.
The news was announced at the AI Action Summit in Paris, where the CRM giant confirmed that it will be the “first AI model developer” to publicly share energy efficiency data for its proprietary models under the new framework.
The initiative currently includes public ratings for 166 widely used AI models, with Salesforce claiming that the increased transparency will encourage market preference for energy-efficient models and promote sustainable AI development.
In a post on LinkedIn, Boris Gamazaychikov, Head of AI Sustainability at Salesforce, wrote that the “lack of transparency is a fundamental challenge to AI sustainability.”
Meanwhile, Bruno Bonnell, General Secretary for Investment for France, also praised the launch of the new benchmarking tool:
The AI Energy Score exemplifies the missions of the AI Action Summit by tackling a pressing societal issue – AI transparency and sustainability – through bold innovation and global collaboration.
In practical terms, the solution includes the following features:
- A standardized framework to measure and compare the energy efficiency of AI models.
- A public leaderboard that rates models across 10 common AI tasks, such as text and image generation.
- The capacity for AI developers to submit both open and proprietary models to the benchmarking portal. Open models are automatically tested, while closed models are evaluated in a secure sandbox environment.
- A new energy-use label, ranging from 1 to 5 stars, which helps users identify energy-efficient AI models. Once rated, developers can generate standardized labels for their models’ energy scores, with guidance on proper label display for visibility and impact.
AI’s Environmental Impact
When we think of environmental damage, it’s natural to picture oil spills and cars belching out fumes – but AI can have a deceptively heavy carbon footprint.
While one of the most impressive aspects of AI is its ability to make complex issues appear simple, in actuality, there is a tremendous amount of energy consumption that goes into powering these tools.
A recent report from the National Engineering Policy Centre (NEPC) emphasized the urgent environmental risks posed by AI’s increasing computational power demands.
One of the key findings of the report was the significant water consumption of companies like Google and Microsoft, with the NEPC recommending that UK companies should be required by law to report their water usage.
This consumption is due to the vast amounts of fresh water needed to cool AI servers.
Indeed, a separate report by Cornell University predicted that data centers housing AI are globally expected to consume six times more water than the country of Denmark by 2027, potentially using up to 6.6 billion cubic meters of water.
Interestingly, while Salesforce’s new solution will measure energy efficiency, it does not specify whether or not this will include water usage.
These misgivings around AI’s impact on the environment were also shared by French President Emmanuel Macron’s AI Envoy, Anne Bouverot.
“We know that AI can help mitigate climate change, but we also know that its current trajectory is unsustainable,” Bouvert said at the AI Action Summit in Paris.
Without worker representation, AI-driven productivity gains risk turning the technology into yet another engine of inequality, further straining our democracies.
Bouvert’s admission that AI can assist with climate change can be understood through a customer experience lens as the ability of AI assistants and chatbots to provide faster responses to customer inquiries.
In doing so, companies can reduce the time that customers spend on calls and navigating departments, which not only improves CX but also lowers energy consumption.
The longer a call or chat lasts, the more energy is used – making quicker, more efficient interactions environmentally beneficial.
Practicing What You Preach
When discussing the new AI Energy Score benchmarking tool, Salesforce was keen to highlight how it has already begun taking steps to reduce its environmental impact via its Agentforce platform.
The agentic AI offering used to build and deploy autonomous agents prioritizes sustainability by avoiding energy-intensive model training.
Rather than relying on a single large language model, the solution uses smaller, efficient models combined with advanced reasoning tools to reduce energy consumption.
As an example, the company pointed to its SFR-RAG model, which ensures accuracy and efficiency while minimizing resource waste.
Additionally, the feature leverages Salesforce Data Cloud and the Salesforce Platform for accurate, responsive outputs without excessive computational demand.
For more on the latest news coming out of Salesforce, check out our video: Salesforce Update – Agentforce 2.0, Atlas, & the New-Look Slack (January 2025)
