AI consumes a large amount of electricity and water to cool data centers. Gartner predicts that by 2030, AI could consume up to 3.5% of the world’s electricity. Obviously, this negative impact must be mitigated. We must not forget, however, that if AI is used correctly, it can help enterprises reach their ESG goals by accurately predicting demand in manufacturing and reducing waste, for example.
Orange Business has developed consulting offerings to help organizations leverage the most suitable AI for their needs while optimizing energy consumption, reducing greenhouse gas emissions, and integrating sustainability-focused governance.
Doing more with less
Frugal AI revolves around designing, developing and deploying AI systems that meet business outcomes and answer clients’ challenges while being more resource efficient and cost effective. The idea is to achieve high performance using less processing power and data, which means creating models with smaller datasets and prioritizing algorithms that consume less energy during the training and inference phases.
Alongside improving resource allocation and sustainability, one of the other key benefits of frugal AI is enhanced cost efficiency. By reducing energy consumption, computational power for AI operations and data in terms of scaling, maintenance, and data storage requirements, enterprises can reduce their operational expenses by adopting these leaner models.
Adopting frugal AI helps organizations comply with emerging legal requirements, such as the Corporate Sustainability Reporting Directive (CSRD), an EU directive on ESG reporting. While it doesn’t spotlight AI’s energy consumption, organizations deploying AI technologies will be expected to include associated energy usage in their reporting.
Controlling AI’s energy demands
With AI adoption growing rapidly and set to drive a 160% increase in data center power demand by 2030, it looks like frugal AI has arrived at just the right time. To put this in perspective, a single ChatGPT query necessitates 2.9 watt-hours of electricity, compared to 0.3 watt-hours required for a Google search, according to a recent International Energy Agency report.
While some tech giants like Google, Amazon and Microsoft have committed to powering their data centers with 100% renewable energy by 2025, AI operations still leave a significant carbon footprint. So, there is still a lot of work to be done.
Many organizations, however, lack awareness about frugal AI and how it can help them reduce carbon emissions. Most AI discourse centers around cutting-edge, high-performance models rather than cost-effective alternatives. There is also a misconception that bigger AI models yield better results, even though smaller models are acceptable for many applications.
Learning how to work with AI minimalism
Orange Business and its consultants support clients in developing and implementing frugal AI. We have developed an AI methodology that prioritizes ethical and environmental responsibility. From the start of each project, Orange Business minimizes data usage, chooses sustainable data centers for infrastructure and refines large language models (LLMs) into smaller, more efficient, specialized models (SLMs). We ensure AI deployment is controlled to reduce energy consumption by processing less data on lower-power infrastructure.
Orange Business aims to educate companies on choosing the best AI for the task with the least environmental impact. Our consultants can help organizations develop a frugal AI strategy across the entire AI lifecycle. First, they will determine whether AI fits the problem correctly. AI should only be adopted if it adds value over alternatives, such as analytics. Where it can be used, it should be used as efficiently as possible and where it will improve carbon footprint.
We recommend finding the best AI solution for your business needs, avoiding large language models (LLMs) where possible and opting for small language models (SML), which are more energy efficient. The most successful frugal AI strategy follows a minimalist approach: only use AI where necessary, keep models small and efficient and optimize infrastructure to see cost and energy savings. Identifying AI use cases that reduce environmental impact can help balance AI's energy consumption with its potential for supporting sustainable initiatives. We are very much practicing what we preach within Orange Business. We use an AI tool internally, for example, alongside Live Intelligence, which shows employees how much CO2 they are going to use by choosing a particular AI search. Most people don’t consider the computational costs of AI queries and are often shocked at what they see.
Now is the time to make AI more efficient
Beyond rising energy costs and performance demands, sustainability and regulatory pressures are driving the need for organizations to adopt a frugal AI mindset.
The move to smarter, leaner AI is a strategic imperative for long-term business resilience and success. Learning about the benefits of frugal AI today means organizations can deploy AI strategically and build sustainable, impactful AI solutions for the future.
To find out more about the frugal AI consultancy offer from Orange Business, contact our team.
Aurélie Gonçalves, Strategy and Innovation Manager at Orange Business, leverages emerging technologies to develop AI, data strategy and Green Data solutions. Her expertise in innovation and marketing helps maximize client value.