Building AI sustainably

Warning This is not a comprehensive guide, but a starting point to support your AI journey. This is not intended to be any form of legal guidance or advice, and should not be used as a substitute for independent legal advice.

Modern AI systems, especially generative AI, can have a large environmental footprint. They use significant amounts of energy, water, and critical minerals. But here's the thing: it's not about avoiding AI altogether. It's about making thoughtful choices that balance impact with value.

For example, generating videos can use many times more energy than generating images, which itself uses more than text. When deciding whether to use AI, consider what you're gaining and what the environmental cost is. It's about weighing the public service benefits against the impacts with the full picture in mind.

Check what you need from your solution

  1. Do you you need to use AI? Not every problem needs an AI solution. Consider whether AI is necessary for what you're trying to achieve, or if there's a simpler approach that would work just as well.

  2. What is the best AI approach? AI includes many options, each with different strengths and impacts. Foundation Models - the large AI systems that power popular chatbots and assistants - can handle lots of different tasks, but they have a high environmental cost.

  3. What is the most energy-efficient approach? Not all generative AI models use the same amount of energy. AI energy comparison tools, such as HuggingFace’s AI Energy Score Leaderboard, can help you choose models that have less environmental impact.

How to reduce your environmental impact

Several decisions can collectively make a real difference:

  • Choose green providers: Look for hosting providers that use renewable energy and hold environmental certifications (such as BREEAM for data centres or about ISO 14001 for environmental management), and publish transparent sustainability reports. These show a commitment to sustainable practices.

  • Use efficient hardware: Cloud providers offer different infrastructure choices. Where possible, select regions with renewable energy, use modern compute types, and right-size your deployments to avoid wasting resources.

  • Manage data thoughtfully: Data is stored in physical data centres that consume energy constantly. Set up efficient storage with clear retention policies, so you're not keeping data longer than needed.

How to monitor your environmental impact

You can use free online tools to calculate your environmental impact. Each tool serves different use cases:

  • CodeCarbon or Carbontracker are tools for tracking emissions when you control your computing environment

  • Ecologits is an online calculator for estimating impacts when using generative AI models through third-party services

  • Green Algorithms is a web-based tool for calculating the carbon footprint of various computational tasks

Monitoring your AI product's environmental impact throughout its lifecycle helps you spot issues early and make adjustments.

Back to top