Generative AI (Gen AI) has undeniably revolutionised industries, with manufacturing standing out as a sector poised to reap the rewards of this transformative technology. From predictive maintenance to supply chain optimisation, manufacturers are rapidly adopting Gen AI to streamline operations, enhance product design, and drive efficiency.
However, the environmental cost of this energy-intensive technology cannot be overlooked. Our recent global research revealed that nearly half of executives reported a 6% average increase in their organisation’s greenhouse gas (GHG) emissions over the past year, with Gen AI adoption as a key factor. And whilst organisations are increasingly aware of the sustainability impact of Gen AI, only 12% are tracking the carbon footprint of the technology and most are not prepared to address the issue.
To tackle this, it’s essential to prioritise responsible Gen AI adoption and an environmentally conscious approach to scaling the technology. The first step, however, is getting a deeper understanding of the challenges.
The environmental cost of Gen AI
The critical aspects of Gen AI's environmental impact lie in its huge energy consumption. To train large language models (LLMs) like OpenAI’s GPT-4 there is a staggering amount of energy needed. It is estimated that a single training cycle can consume between 51,772 and 62,319 MWh of electricity, which equates to enough power for least 5,000 homes for a year. Even after training, the inferencing phase, where the model is deployed and utilised, can often require equal or greater amounts of energy. The energy demand for data centres, where these operations take place, is expected to double from 460 TWh in 2022 to 1,000 TWh by 2026, according to the International Energy Agency.
In addition to energy use, water consumption is another major concern associated with Gen AI. Data centres require vast quantities of water for cooling to prevent overheating during intensive computational processes. For instance, a single inference query – or a series of a few prompts - on an LLM-based chat bot requires about 500 millilitres of water. ChatGPT’s 200+ million weekly users underscorethe significant environmental impact of Gen AI.
Limited commitment to environmental standards
Despite these challenges, our latest data shows that businesses are not prioritising sustainability when it comes to their Gen AI investments. Only 21% of executives rank environmental footprint among the top five factors considered when selecting or building Gen AI models, compared to 78% who prioritise performance.
Executives from manufacturing organisations stated that a significant barrier to addressing the environmental footprint is the lack of transparency from Gen AI providers. 70% of business leaders cited limited disclosure from hyperscalers and Gen AI model providers as a key reason for not measuring the environmental impact of their AI initiatives. This lack of visibility may impede efforts to incorporate sustainability into AI adoption strategies.
The good news is that Gen AI can also be a powerful tool for aiding organisations’ sustainability efforts. 45% of manufacturing organisations have started using Gen AI for sustainability initiatives, making it the second most active industry in our research after high-tech. Additionally, 61% of manufacturing organisations are piloting Gen AI initiatives, signalling a growing recognition of the need to balance innovation with environmental responsibility.
Striving to balance innovation with sustainability
Addressing this clear hurdle requires proactive measures. To ensure sustainable Gen AI adoption, manufacturers must integrate sustainability throughout the AI lifecycle. This includes evaluating environmental costs, adopting energy-efficient solutions like model compression and using greener data centres powered by renewable energy.
Smaller, tailored language models for specific industrial IoT devices can reduce computational intensity while maintaining performance. Customising data sets for older equipment enables Gen AI integration without overhauling infrastructure. Additionally, Gen AI improves system interoperability, paving the way for adaptive factory systems that respond to consumer demands, supply chain shifts, and regulatory requirements.
Transparency is also an essential component, so organisations should monitor AI carbon footprints, set sustainability targets, and report progress.
We’ve seen great examples of this success in the real-world. Walmart’s Gen AI-powered waste-management system reduces waste and optimises donations, while battery manufacturer EnerSys leverages Gen AI to align with climate goals by analysing emissions, travel, and waste data.
Gen AI can be also used to reduce time spent on tasks like coding and optimising assembly line configurations. It enables faster design cycles by modelling material behaviours, predicting bottlenecks, and informing future product designs based on past data.
By prioritising sustainability, manufacturers can harness Gen AI’s capabilities while mitigating its environmental impact.
The path to sustainable innovation
While Gen AI introduces undeniable opportunities, its deployment must be balanced with environmental stewardship. Manufacturers can unlock significant advantages, including reduced costs, enhanced regulatory compliance, and stronger market positioning, by integrating sustainable practices into their AI strategies.
The future of manufacturing will be defined by those who embrace both the transformative potential of Gen AI and the responsibility to manage its environmental impact. By investing in greener infrastructure, optimising AI models, and fostering transparency, manufacturers can lead the way toward a sustainable industrial revolution.
Mike Dwyer is UK Head of Intelligent Industry at Capgemini
Will AI end or enhance a career in engineering?
<blockquote>‘Go and get me a number of feasible designs that will satisfy these constraints so that I can add my greater creativity.’ </blockquote>If...