Why AI growth could wreck your sustainability goals

Welcome back to Neural Notes, a weekly newsletter where I look at some of the most interesting AI news of the week. In this edition: A look at the sustainability challenges posed by AI data centres off the back of findings in a recent report by the UTS Institute for Sustainable Futures.

The Drivers of Change: Meeting the Energy and Data Demands of AI Adoption in Australia and New Zealand, released earlier this week, highlighted significant energy consumption concerns around the mass adoption of AI.

In particular, it found a disconnect between business uptake and practical plans to address energy consumption.

This is critical considering there are estimates that by 2026, AI data centre energy usage could increase to between 620 and 1,050 TWh which is roughly the same as powering the entirety of Germany.

The report found 72% of IT managers surveyed have either adopted or are piloting AI technologies, with only 5% having no plans to deploy AI in the next three to five years.

It also comes at a time when the government is pushing towards ethical deployment and use of AI and AI regulation.

However, despite the race to implement AI, the report found 68% of IT managers expressed concerns about the increased energy consumption, but 53% were unsure of the extent of the increase.

The AI energy use knowledge gap

According to Gordon Noble, a co-author of the report and the Director of Business, Economy and Governance at the Institute for Sustainable Futures at UTS, the lack of knowledge regarding AI’s energy intensity is a critical issue.

“Surveying IT managers, there’s a huge amount of uncertainty from their perspective as to what AI means from an energy perspective,” Noble said to SmartCompany.

The institute also released a report regarding this last year and found sustainability professionals had AI on their radar but only 6% were confident about the data they were receiving in regards to data centres.

A year later things have progressed, but there are still many improvements to be made.

“AI has exploded here and we’re starting to get the disclosures to understand what it actually means in terms of the intensity of the energy,” Noble said.

“We’re calling this the big reveal – we’re starting to understand and from that you get action, so that’s a positive.”

And it’s certainly necessary. Despite the increasing focus on sustainability in the business sector in recent years, these concerns seem to be somewhat pushed to the side when it comes to AI adoption.

“There needs to be a bridge between IT managers and sustainability professionals,” Noble said.

He went on to say IT managers should focus on business growth and implementing solutions for the business, while a sustainability expert can focus on hitting those environmental targets.

“At an institutional level they need to be able to talk to each other, and then we also need to do that at an industry level,” Noble said.

While this may be achievable at the larger end of town, it can be particularly challenging for smaller companies. The vast majority of companies in Australia – which make up 97% of our business landscape – don’t have an in-house IT professional, let alone a sustainability expert.

“We really need to make it easier [for small businesses] to be able to understand sustainability impacts,” Noble said.

“We shouldn’t need a small business to employ a sustainability professional to be able to do that. We should be able to have information coming in a way that is easily digestible so they can make informed decisions.”

Businesses are trying to keep up

But it’s not all doom and gloom, there are organisations exploring different measures to offset AI’s energy usage, such as optimising equipment use and investing in green energy sources.

There are also startups such as alleged OpenAI rival Sakana AI which are looking for more sustainable ways to roll out LLMs.

“There are solutions around renewable energy, so basically looking at your energy consumption and trying to ensure that you’ve got renewable energy to match that,” Noble said.

That being said, the scale-up of AI and subsequent energy consumption has been incredibly fast so it’s been difficult for businesses to keep up.

Google proved this just this week in its 2024 Environmental Report, which revealed a 17% increase in its data centre electricity consumption in 2023.

Google’s greenhouse gas emissions also increased by 13%, which the tech giant said highlighted “the challenge of reducing emissions while compute intensity increases and we grow our technical infrastructure investment to support this AI transition”.

But this isn’t isolated to Google, a large portion of companies – both big and small – are facing this issue.

“When we look at the recent disclosures of Google, Meta and Microsoft – the growth has been so quick in terms of their emissions in the last couple of years. They haven’t been able to match that in terms of the renewable energy investment,” Noble said.

Small businesses, which often lack dedicated sustainability professionals, need accessible and digestible information to make informed decisions about their AI and energy use. Noble emphasised,”we really need to make it easier to understand your sustainability impacts. We shouldn’t need a small business to employ a sustainability professional to be able to do that”.

The role of policymakers

AI regulation has been on the lips of the federal government for a while now, with a focus on balancing safe and ethical implementation with innovation. But perhaps it’s worth considering whether sustainable implementation should be an equally important pillar in this rollout.

Noble says this is being addressed to a certain extent through protocols surrounding the government’s own procurement and some green rating standards for data centers.

“There are some good things. There are some positive things happening. And obviously, the challenge, even from the federal government perspective, is that the growth of AI has probably surprised them because the data center industry is not included in terms of the big focus areas of decarbonisation in Australia,” Noble said.

Noble also said that while some Australian data centres have green ratings, this is currently not the case. For Noble, setting green standards for data centres and including these facilities in Australia’s broader decarbonisation strategies would be an obvious win.

“That’s something I think the federal government can nudge to make sure that anyone they’re procuring from does have those ratings,” Noble said.

Noble proposed the concept of a “data budget,” similar to financial budgets, to help businesses understand their data usage and make informed choices about AI applications. This approach could prevent unexpected costs and support sustainable AI deployment.

“We need some tools that can help a business to understand what their overall data budget is. How much are they spending on data so that they can actually, not so much in dollar terms, but in terms of what they’re actually utilising around data,” Noble said.

Noble also stated the government should be supporting small businesses not only with AI upskilling, but sustainably implementing AI.

“In order to support small businesses it’s [about asking] how they’re going to need to be supported and how can the federal government support small businesses around understanding the sustainability disclosures,” Noble said.

He also said the government could help in skilling up small businesses on how to make their own energy disclosures when this is requested by larger businesses they’re working with.

“I think there’s quite a bit that the government could do in that space,” Noble said.

Other AI news this week

The Tech Council of Australia released a report projecting 200,000 AI jobs for Australia by 2030… but only if we upskill our workforce
Atlassian alumni just scored $11.5 million in seed funding for Redactive AI which aims to address the lack of AI engineering and security skills within enterprise software teams
Images of Australian children were used to train LLMs without permission
You can now tell YouTube to remove videos where your voice or face has been AI deepfaked
Can AI be used to help diagnose dementia?