How to navigate net zero while accelerating AI

Historically, IT and technology teams have passed under the radar as organisations deploy their net zero strategies and look for opportunities to reduce their carbon footprint.

 

But as generative artificial intelligence (AI) becomes more prevalent in the charity sector, the potential for digital carbon emissions significantly increases. IT and tech teams become even more critical to enabling net zero.

 

AI has the potential to transform the way charities operate by boosting productivity and driving efficiencies but minimising its environmental impact needs to be a serious consideration. Below we explore how charities can embrace both generative AI and decarbonisation efforts.

 

 

What are the net zero imperatives for charities?

 

The Climate Change Act 2008 commits the UK to be net zero by 2050 across all sectors, which has led to a number of new requirements.

 

One requirement demands climate emissions disclosures, which directly impacts charities registered as companies and classified as large. Such charities may need to report under the Streamlined Energy and Carbon Reporting (SECR), which looks at financial risks and opportunities associated with the organisations sustainability plans.

 

Charities who do not fall under these mandatory requirements are still encouraged to adopt and embed the requirements before the 2050 deadline.

 

Finally, there are financial implications with getting to and retaining net zero. That’s the ‘net’ part. Charities in the target year will have unavoidable emissions that are critical to their operations and will thus have to consider the cost associated with their removal.

 

Prices vary depending on the type of removal solution – nature based or engineered – but reducing emissions before the net zero target is the best way to significantly reduce these costs.

 

 

Net zero inflection point: Carbon hungry AI

 

In the EU, the digital sector already accounts for approximately 9% of electricity consumption and just over 4% of greenhouse gas emissions. Generative AI models require more computational power than standard AI and in some cases in orders of magnitude.

 

Research shows that these models might use more than 30 times more energy when including data centres. As charities are likely to have several of their data models residing in the public cloud or hybrid colocation environments, the direct impact is likely to be seen in their supply chain scope 3 emissions.

 

The actual impact of AI on the digital supply chain is emerging as some cloud providers have already seen an increase in their overall emissions. Decarbonisation of the grid to power all these datacentres with clean energy is likely to be outpaced by AI growth.

 

This also excludes the embodied carbon emissions (material extraction, manufacturing, and end of life) associated with all the physical components across the digital value chain from networks to end point devices. Organisations including charities deploying AI will have to take on more responsibility to ensure carbon emissions across their value are managed sustainably.

 

So, what can charities practically do to build an AI strategy that doesn’t compromise on net zero?

 

 

How to build a net zero AI strategy

 

It’s important to start by building a carbon baseline by measuring the footprint across the full digital value chain for AI. Starting from the network through to data centres and into the cloud.

 

BT’s Digital Carbon Calculator provides customers with carbon footprint data across the lifecycle for networking devices to end point devices in accordance with ISO14040.

 

Charities can use this to get going while obtaining the same information from their data centre or cloud infrastructure provider for their services. This will help to understand where they’re at today.

 

The next step is to look at optimisation by designing networks to train and run AI models in the most efficient way from the edge and into the cloud.

 

A quick win but a highly effective one is to select hosting of AI models in data centres that are in low carbon regions or powered by 100% renewable electricity.

 

If there are opportunities to refresh devices, then selecting those with the highest efficiency or refurbed options can lower carbon footprint.

 

Finally, there is an opportunity for Trustees to lead from the top by setting an internal carbon price to drive more sustainable choices across the organisation influencing culture and driving down emissions while also helping to reduce future carbon removal costs.