AI is a sustainability tool. It is also a sustainability cost.

When Mabel Chan introduces AI to the students and executives she works with, she reaches for the image of a brand-new Ferrari: a beautiful, high-performance machine built to move fast. The metaphor captures both the appeal and the catch. Speed is useful, but it arrives with a cost that is easy to miss.

That tension framed The Impact of AI on Sustainability, a LinkedIn Live session hosted by Rita Tsui, Founder of AsiaHRM, as part of its Sustainability on Business Series, supported by HRM Asia. Chan, Co-Founder of the green edtech organisation Carbon Linking, joined to unpack a question that is becoming harder for leaders to sidestep: when an organisation adopts AI, is it building a more sustainable future, or simply shifting the environmental burden to a less visible place?

The enabler

Chan was clear that the upside is real, and she drew on her own work to show it. At Carbon Linking, which has reached more than 6,000 students through partnerships including Ocean Park Hong Kong, AI has become, as she calls it, a “super engine” for climate education. The organisation designs interactive formats such as a Sustainable Development Goal (SDG) escape room and campus-wide treasure hunts, where students absorb sustainability concepts while feeling as though they are playing. AI now speeds up the research, visual production, and storytelling behind those experiences, and supports more personalised analysis and coaching, allowing learners to work with real data rather than “dry climate textbooks.”

The same pattern, she said, plays out across other industries. After a recent visit to a Chinese robotics company, Unitree, Chan pointed to AI-trained robotics as a way to make physical work more precise – in agriculture, for instance, machines can read weather data, sow and farm more accurately without a human directing each decision, using water and resources more efficiently. In supply chains, she added, AI can surface more optimised delivery routes and help organisations identify suppliers with lower carbon emissions, doing in moments what a team might once have pieced together manually.

The hidden bill

The counterweight, Chan stressed, is that AI does not simply “sit on the cloud” as many users imagine. Every request travels from a device to a data centre, where the algorithm runs and the result is returned. Those data centres need energy to operate and water to cool them, much as a laptop heats up during a demanding game and has to be cooled down. As adoption scales, so does that footprint.

To illustrate the trajectory, Chan cited Ireland, a European tech hub where, by her account, data centres already consume around 22% of total electricity and are on track to reach 32% by 2026. She also noted that countries and organisations often site these hubs offshore to manage operating costs, which can place the environmental burden well away from where the efficiency gains are booked.

Counting both sides

Asked how leaders should judge whether their AI use is genuinely net positive, Chan framed it as a cost-benefit exercise – a reading Tsui echoed, describing the cost as the carbon footprint AI generates and the benefit as the efficiency it delivers.

READ MORE: Sustainability isn’t optional: The growing stakes of corporate climate action

Chan’s advice was to start with a clear picture of where an organisation’s emissions actually come from, since AI may be a relatively small slice of the total, and each industry differs. Only then can leaders weigh the efficiency AI delivers against the energy it consumes. She also made the point that most organisations do not own their data centres; they subscribe to cloud services, which limits how far they can improve the underlying infrastructure themselves. There is a lifecycle argument too: if AI compresses the time needed to deliver the same output, she suggested, the associated operational and environmental cost can be compressed with it.

A governance and literacy question

For Chan, governance begins with literacy – and this is where the leadership and workforce dimension came through most strongly. Management teams, she argued, should audit which parts of the business drive emissions, then extend that awareness to employees, who use these tools daily. Once teams understand that AI carries an environmental cost, they can actively favour suppliers that disclose their emissions and commit to renewable energy. Major cloud providers increasingly publish carbon dashboards, she noted, and pointed to Canva, which has set a net-zero target for 2040 and is working to decarbonise its supply chain, as an example of the kind of commitment buyers can look for within the same tier of efficient, cost-effective vendors. For the minority of organisations that do run their own infrastructure, the conversation shifts to the energy mix itself – options such as on-site solar and more efficient temperature control.

Tsui’s closing read captured the throughline: education is key. Not only owners, she said, but employees too need to understand the relationship between AI and sustainability – a reminder that for leaders, the question is less about the technology than about the people deploying it.