How Salesforce is Tracking AI's Impact with AI Energy Score
As artificial intelligence (AI) becomes increasingly integral to global systems, the pressing question of its environmental toll arises.
At the heart of the debate is AI's extensive use of power-hungry data centres, raising concerns about sustainability.
However, as AI continues to develop, it's clear there is a vital need for a more standardised way to understand and measure these impacts explicitly.
Enter Salesforce's AI Energy Score, a pioneering tool intended to precisely gauge and benchmark the energy efficiency of AI models.
The tool's official release is set for the AI Action Summit on 10 February 2025, where more than 100 country's representatives gathered to deliberate on utilising AI for beneficial outcomes.
Supported by the French Government and the Paris Peace Forum, this summit and the introduction of the AI Energy Score mark significant moments in striving for more sustainable AI practices.
“We are a critical moment with the rapid acceleration of both the climate crisis and the AI innovation," explains Boris Gamazaychikov, Head of Sustainability AI at Salesforce.
"These two trends are interlinked, as we can see exemplified by new gas development slated to support AI data centers. However, there has not been much transparency on the link between AI models and their environmental impacts to date.
"That’s why we created the AI Energy Score - to have a standardised, comparable framework to disclose and compare the impacts of AI. We hope that this remove a key blocker to future where sustainable AI is the norm.”
Understanding the AI Energy Score
Conceived in collaboration with entities like Hugging Face, Cohere and Carnegie Mellon University, the AI Energy Score aims to bring much-needed transparency and uniformity in measuring AI models' energy consumption.
“The AI Energy Score represents a major milestone for sustainable AI,” says Dr. Sasha Luccioni, AI & Climate Lead, Hugging Face.
“By creating a transparent rating system, we address a key blocker for reducing the environmental impact of AI. We’re thrilled to launch this project and look forward to seeing wider adoption.”
The AI Energy Score introduces several key features to promote transparency and sustainability in AI development:
- Standardised energy ratings: A framework that evaluates AI models' energy efficiency, requiring developers to measure energy consumption.
- Public leaderboard: A comprehensive ranking of more than 200 models across 10 common AI tasks, including text and image generation.
- Benchmarking portal: A platform for submitting and evaluating both open and proprietary AI models.
- Recognisable energy use label: A 1 to 5-star rating system for easy identification of energy-efficient models.
- Label generator: A tool for AI developers to create and share standardised energy efficiency labels for their models.
The introduction of the AI Energy Score is expected to have far-reaching implications for the AI industry:
- Driving market preference: By enhancing transparency, the score can drive market preference towards more efficient models.
- Incentivising sustainable development: The public nature of the ratings is likely to encourage AI developers to prioritise energy efficiency in their model designs.
- Facilitating informed decisions: AI users and developers can make more informed choices about which models to use or build upon, based on their energy efficiency ratings.
Salesforce's investment in green AI
Salesforce is pioneering this initiative, becoming the first to disclose energy data for its proprietary models under this new system.
This aligns with their broader sustainability goals and responsible AI development approach.
The company's Agentforce platform, launched in 2024, utilises autonomous AI agents to enhance various business functions while optimising for energy efficiency.
“The AI Energy Score builds on Salesforce’s Agentforce innovation, the agentic layer of the Salesforce Platform for deploying autonomous AI agents across any business function," explains Boris.
"Agentforce offers tools to build and customise agents, as well as a library of ready-to-use skills for sales, service, marketing, commerce, Tableau, Slack and more.
"We’re already seeing some incredible innovations on how organizations use Agentforce. For example, Good360 connects donated goods with people in need. Each year, six billion pounds of goods go to landfills instead of communities. With Agentforce, Good360 now allocates donations more efficiently, cutting waste, reducing emissions — saving 1,000+ employee hours annually.
"Agentforce is built with sustainability at its core, delivering high performance while minimising environmental impact.
"Unlike DIY AI approaches that require energy-intensive model training for each customer, Agentforce is optimised out-of-the-box, eliminating the need for costly or carbon-heavy training."
“Reducing AI energy consumption lowers operational costs, optimises infrastructure and enhances long-term sustainability and profitability,” explains Suzanne DiBianca, EVP and Chief Impact Officer at Salesforce.
“We are proud to work with industry leaders to build a more transparent AI ecosystem.”
Confronting the AI energy challenge
This initiative is timely, especially as concerns over AI's increasing energy demands grow.
A survey by Salesforce highlights that 65% of sustainability professionals are weighing the environmental costs against the benefits of AI as its data centre energy use is set to double by 2026.
"The AI Energy Score benchmarking tool is more than just a energy use metric — it can be a strategic business advantage," Boris says.
"By helping organisations assess and optimise the energy consumption of their AI models, it can contribute to lower operational costs, improved infrastructure efficiency and long-term profitability."
Explore the latest edition of Energy Digital Magazine and be part of the conversation at our global conference series, Sustainability LIVE.
Discover all our upcoming events and secure your tickets today.
Energy Digital is a BizClik brand
