AI’s rapid adoption comes with a steep energy cost. Popular large language models like ChatGPT are thought to consume roughly 2.9 watt-hours each use— almost ten times more than a standard Google search. As AI becomes more integral to global strategies, its energy demands will soar. By 2027, its energy consumption could match that of entire countries, challenging industries, including shipping, to think hard about AI’s role in sustainability objectives.

AI’s double-edged sword

Data centres are the backbone of AI operations, hosting the computational resources to process massive datasets. These centres consume enormous amounts of energy. In 2020, global data centres consumed about 200 terawatt-hours (TWh) of electricity, accounting for about 1% of the world’s total energy use—a figure expected to rise as AI adoption accelerates.

Professor Thomas Nowotny, Head of the Sussex AI Research Group at the University of Sussex, told ICS Leadership Insights: “AI has an energy problem. Training models like ChatGPT have been estimated to consume as much energy as driving a car to the moon and back. The technology behind such models is increasing in size and energy consumption by 750 times every two years. At this rate, we might use all the energy on the planet to build chatbots within a decade.”

Nowotny explores alternatives inspired by the human brain, which operates on about 20 watts of power. He envisions a future where AI systems utilise ‘neuromorphic’ computing, achieving low-power, sustainable AI. However, he cautions that its current pace of adoption has reached a critical point, and that energy infrastructure cannot keep up with demand. “The free lunch of AI is over”, he stressed, adding: “Companies need to avoid getting locked into solutions that may not be optimal in the long run.”

For shipowners this could have a significant impact on Scope 3 emissions, those indirect greenhouse gas emissions that are incurred elsewhere in the value chain. When any maritime business uses AI-powered services provided by cloud providers or third-party data centres, the emissions from the energy used to power them become part of their Scope 3, whether they like it or not.

Mitigating AI’s energy footprint.

The challenge lies in balancing AI’s energy demands with the broader sustainability benefits it generates. WiseTech Global creates software solutions that it claims will be able to help optimise complex supply chains.

Richard White, CEO and Founder of WiseTech Global, told ICS Leadership Insights that while Generative AI’s energy consumption is a concern, its potential for optimisation in operations and compliance justifies the trade-off.

“We started work on GenAI in secret and very early on when it became obvious that it was going to be important,” he said. “We have been using and growing Machine Learning, Big Data, Natural Language Processing and a variety of automation tools for more than a decade and increasingly, we are blending GenAI into the functionalities and automations that we create.”

WiseTech Global operates three data centres in Sydney, Chicago, and Hamburg and is increasingly turning to sustainable energy for answers.

“Our own data centres are very energy efficient – as are most major cloud providers – that said, GenAI has become a major consumer of power and has the potential to dramatically increase energy consumption as demand grows,” he said. “We already have mitigations in place and a significant proportion of energy in our largest data centre (Chicago) is already from non-GHG (greenhouse gas emissions) sources.

“The carbon emissions saved by our AI-driven optimisations far outweigh the additional energy consumed.” White said they are constantly working to encourage customers to use its centralised, more efficient data centres, removing inefficient self-hosted and energy-intensive environments.

“The carbon (GHG) that can be removed from the burning of fossil fuel by using GenAI and related optimisations is orders of magnitude larger than the rather minor increase in GHG caused by the additional processing power demanded by GenAI to create these optimisations,” he added.

While the benefits may indeed outweigh the energy cost, data centres still use a colossal amount of power and water for cooling systems. Shipowners implementing AI into their strategies will need to carefully ensure their partners have a robust strategy and commitment to renewable energy options and are not relying on electricity generated by fossil fuels.

Making sure AI works for you

In order to feel the benefits of a sustainable AI strategy in environmental social governance, companies must be careful when selecting their data centre partners, ensuring they are equally committed. Major tech companies like Amazon Web Services (AWS), Microsoft Azure, and Google Cloud are all investing heavily in renewable energy at their data centres and Google has publicly committed to net-zero carbon emissions across all of its business and value chain by 2030.

Among its many projects, Maersk Innovation Center has developed AI-based solutions such as digital twins and computer vision to enhance visibility and optimise drayage movements. Erez Agmoni, Global Head of Innovation for Logistics and Services at Maersk, told ICS Leadership Insights that maritime leaders need to start with clear problem definitions when adopting AI and warns against them jumping on the bandwagon without fully understanding the environmental implications.

He said: “You should always start with ‘what problem are we trying to solve here?’ and whether AI will solve it for you. If what you’re looking at is going to solve it great, run a proof of concept test on a small scale and see if you like it, then expand further but don’t just be following trends and buzzwords out there.”

He added that many of the latest AI developments won’t immediately add value, and businesses must work hard to understand how to shape technology to meet specific challenges.

Balancing innovation with sustainability

As AI’s popularity grows, maritime businesses must collaborate closely with data centre partners to balance their drive for innovation with sustainability.

Ravindra Rapaka, Director of AI and Product Manager at Aquasight, a US-based technology firm specialising in enhancing water systems, sums it up. He says a truly green AI is not here yet, but companies should enshrine sustainability in any longterm AI strategy. He told ICS Leadership Insights: “Although the transition from classic AI to Green AI is a new topic, it’s both on time and important, especially for global logistics providers. In the short term, demand for mainstream AI will remain strong because of its entrenched nature, but demand for Green AI will gain because of increasing awareness of the environmental impact and because it’ll be the next big thing for organisations keen to address sustainability.”

He added: “Achieving a completely Green AI landscape will take time.” This was due to more progress needed in energy-efficient hardware and computing technologies and the high initial investment costs of alternative energy. He said that through better regulatory support, international policies and greater incentives to innovate, Green AI would indeed be the next big thing.